# AI agents
Source: https://docs.clairelabs.ai/documentation/claire-ai/agents
Build custom AI agents to conduct interactive interviews and assessments.
Claire's audio-native AI agents can maintain low-latency conversations and conduct entire interviews on behalf of educators. They can be used to add a conversational layer to new or existing coursework without sacrificing human oversight.
![Image]()
## How to build AI agents
Building an agent is straightforward and only requires a few lines of text. Our platform translates your instructions into an exhaustive playbook that you can review before sharing the agent with students.
Head to the Agents section on the platform. Here, you can view existing agents or create new ones.
* Click Create to start setting up a new agent.
* Give the agent a name and a short description (for internal organization).
* Decide whether you'd like to use your agent in an **Oral** (agent-led interview) or **Hybrid** (written submission followed by an interview) assessment.
* Add instructions that guide the agent and explain its role and objective.
Agents are highly capable and can call tools during the interview. Click Add context to attach media files or Add board to add a whiteboard (empty or with a drawn template). You can then mention the tools in your instructions by typing @ and selecting the item.
Click Update playbook to create detailed instructions that will be shared with the agent. This gives the agent detailed instructions on how to behave and conduct the interview, and makes it easy for you to review the interview script and guardrails.
Click Save changes to save your agent.
### Example instructions
Here's an example of an agent used for a business ideation assignment. It consists of only a few lines of instructions:
```text theme={null}
First, prompt the student to think out loud when answering your questions.
Task 1 - Identify business opportunities
- Ask the student to name 1 business idea based on their analysis of {Slide}
- If needed, prompt them to list their ideas clearly.
Task 2 - Sketch out a business model
- Tell the student to pick one of the ideas they came up with
- Then ask them to outline their business model onto the canvas: {Whiteboard}
Conclude the interview once the student has finished both exercises.
```
## Agent context and tools
Agents are extremely capable and can call tools during the interview with a student. Tools enable you to bring different media (images, lecture slides, etc.) or an interactive whiteboard into the conversation, requiring the student to demonstrate cognitive skills in real time.
To add tools to your agent:
1. Click Add context to attach media files, or Add board to add a whiteboard (empty or with a drawn template).
2. Once added, you can simply mention the tool in the agent's instructions by typing @ and selecting the item.
![Image]()
## Keyboard shortcuts
| Key | Action |
| ------------ | ----------------------------------------------- |
| @ | Mention a tool while writing agent instructions |
**Looking for more details?**
* To create assessments that leverage your new AI agent, see [Assessments](/documentation/course/assessments).
* To see what an AI agent interview looks like from the student's perspective, see [Student portal](/documentation/course/student-portal).
# Audit and compliance
Source: https://docs.clairelabs.ai/documentation/claire-ai/audits
Understand how Claire manages AI compliance, data privacy, and auditability.
When using AI (in any form or shape) for assessment or feedback, you should rely on systems that are transparent and allow you to trace and track the use of AI on a single-user level. We go beyond that and enable educators to export comprehensive audit reports, designed for internal and external stakeholders, that document the use of AI on a submission basis.
![Image]()
## Audit reports
Claire enables educators to generate and export audit reports for any AI-assisted grading or assessment process. These reports provide a traceable record of how AI was used, the feedback it generated, and how educators interacted with its suggestions.
## Regulatory compliance
We have aligned Claire with the European Union's GDPR and EU AI Act because we believe education requires a more conservative approach to AI. In assessment and learning, strong data protection, human oversight, transparency, and governance are essential.
Data stored on our platform is never used to train models or shared for commercial purposes.
* **EU AI Act:** Many AI systems used in educational assessment are considered high-risk and therefore subject to stricter requirements. Claire is designed to meet these requirements with robust human-in-the-loop controls.
* **GDPR:** GDPR adds further safeguards around lawful processing, data minimization, security, and student data rights.
***
**Looking for more details?**
* To see how student data privacy is explained to students, see [Data privacy and security](/tutorials/for-students/data-privacy-security).
# Insights: conversationally query your course data
Source: https://docs.clairelabs.ai/documentation/claire-ai/insights
Use the Insights feature in the sidebar to ask questions about student performance, feedback trends, and grading metrics across your courses.
The Insights section provides a powerful, conversational interface for querying your student performance and feedback data. Instead of manually exporting spreadsheets and running complex filters, you can simply ask Claire questions in natural language to uncover trends and metrics across your courses.
![Image]()
## Accessing Insights
1. Click on Insights in the main sidebar.
2. At the top of the view, you can optionally filter the data context to a specific [course](/documentation/course/assessments) using the dropdown menu.
3. Use the chat input field at the bottom of the screen to enter your query.
## Querying data
You can ask open-ended or highly specific questions about student progress, feedback sentiment, or rubric criteria performance.
Insights are generated based on the [grading remarks](/documentation/reader/annotate), highlights, and scores you have recorded across your submissions.
For example, you could ask:
* *"Which rubric criterion do students struggle with the most in the Midterm Assessment?"*
* *"Show me the distribution of grades for the Marketing Strategy course."*
* *"Are there common themes in the weaknesses I've highlighted this week?"*
Claire will analyze the available performance and feedback data and return a clear, contextual answer directly in the chat interface.
![Image]()
***
**Looking for more details?**
* For a class-level report that surfaces learning trends and actionable classroom recommendations, see [Course reporting](/grading/class-performance-reports).
# Assessments
Source: https://docs.clairelabs.ai/documentation/course/assessments
Create and manage written, oral and hybrid assessments using AI agents.
Claire allows you to easily create assessments that leverage agentic, conversational learning experiences. Assessments can be embedded into daily course work, tailored learning scenarios, or used as formal grading tools.
We use the term "assessment" rather broadly. You can use Claire for graded assessments, take-home assignments, casual and informal exercises, and role-plays or simulations.
## Assessment types
Our platform supports three different types of assessments:
### Written learning scenarios
Students submit written work without an interactive AI interview.
* Ideal for traditional assignments where AI can still be used for providing immediate feedback based on your rubric.
### Oral learning scenarios
Students speak with the agent directly, without submitting work beforehand.
* The interaction follows a custom interview playbook defined by the educator.
* Agents can call tools such as images, lecture slides, and whiteboards during the interview.
* Best suited for reflection exercises, knowledge and comprehension checks, role-play, and simulations.
### Hybrid learning scenarios
Students first submit written work, then continue into an interview with the agent.
* The interview still follows the educator's playbook, but can adapt dynamically to the student's submission.
* Agents can call tools such as images, lecture slides, and whiteboards during the interview.
* Best suited for assessing how much agency, understanding, and ownership a student has over submitted work.
## Creating an assessment
To create an assessment:
1. Select a course from the sidebar (or click to create a new one). Within the course view, click New assessment.
2. Give the assessment a title and select the type (Written, Oral, or Hybrid).
3. Select your agent from the list of available agents and click Continue.
4. Upload the **Instructions** that you'll share with your students.
5. Upload a **Grading rubric**. A rubric is mandatory regardless of whether you're planning on grading the assessment, as it will be used to provide students with immediate AI feedback.
Once created, the assessment will appear in your Assessments table.
![Image]()
## Publishing assessments
To publish an assessment:
1. Click the Publish button next to the assessment in the table.
2. In the dialog, set a **start date** and a **deadline**.
3. **Enable AI feedback** if you want students to access an AI-generated feedback report right after they submit their work or conclude the interview. For compliance reasons, if you want your assessment to be graded, you cannot use AI to give feedback.
4. Customize the feedback report in the Portal tab. This format will be used for both human and AI feedback.
![Image]()
***
**Looking for more details?**
* To learn how students access and view assessments, see [Student portal](/documentation/course/student-portal).
* To learn how to build AI agents for your oral and hybrid assessments, see [AI agents](/documentation/claire-ai/agents).
# Student portal
Source: https://docs.clairelabs.ai/documentation/course/student-portal
Manage how students access and view assessments.
The Student Portal is where your students will go to access their assignments, speak with AI agents, and review their feedback.
![Image]()
## Managing the portal
You can manage your published assessments in a simple Kanban view, accessible via the Student Portal tab at the top of your course page.
The Kanban view shows the status of each of your assessments. You can manually drag and drop assessments between the columns to change their state and visibility in the student portal:
* **Scheduled**: The assessment is not yet visible to students in the portal, but will go live on a defined start date in the near future.
* **Live in portal**: The assessment is live and can be taken by your students.
* **In review**: The assessment is no longer available to students and is awaiting your review (not applicable to assessments with AI feedback enabled). Assessments are still visible to students on the portal but labeled as "awaiting feedback". In this column, you can use the **Share grades in portal** toggle switch. When turned **on**, published feedback and grades will be live immediately in the student portal. When turned **off**, grades and feedback are staged, allowing you to release all feedback at the same time to avoid some students receiving their results early while others are still waiting.
* **Completed**: The assessment is completed. Feedback and grades are now available to students via the student portal.
## Accessing the portal
Click the Share course button in the page header on the right-hand side of the Student Portal tab to access a QR code and URL that directs you or your students to your course page.
![Image]()
### Student login experience
Students can log in to their account using their email address and an OTP (One-Time Password) code for enhanced security. After verification, they are automatically directed to your course page.
On the course page, students see:
* Their outstanding assessments.
* Completed assessments with available feedback.
To start an assessment, students simply click on Take assessment and follow the instructions.
Before starting an interactive oral or hybrid assessment, students will be prompted to perform a device check to ensure their microphone and camera are working properly.
If the assessment involves an AI agent interview (Oral or Hybrid), the student will be presented with a conversational interface where they can interact with the agent via chat or voice.
Once an assessment is completed and graded, students can return to the portal to view their comprehensive feedback, grades, and the full transcript of their session.
***
**Looking for more details?**
* To create and manage your assessments, see [Assessments](/documentation/course/assessments).
* To see what the student's feedback report looks like, see [Feedback reports](/grading/student-feedback-reports).
# Manage and track submissions
Source: https://docs.clairelabs.ai/documentation/inbox
The inbox is your central hub in Claire. See all student submissions at a glance, track their status, upload files manually, and manage flags and archives.
When you sign into Claire, the inbox is the first thing you see.
![Image]()
It gives you a bird's eye view of student submissions in your course — what still needs grading, what's been shared with students, what requires follow-up, and what you've put away.
You will only see student submissions that do not have AI feedback enabled. Submissions that receive AI instant feedback are handled automatically and will not appear in your inbox.
## Submission statuses
Submissions from the Student Portal that are due for review will automatically appear in your Inbox. You can open and view a submission by clicking on the respective submission item. This will open the submission (written work, transcript, or both) in the [Reader](/documentation/reader/annotate).
**Todo** submissions are waiting for your review. A submission enters this status as soon as you upload it or receive it via the portal. Work through your Todo list to keep grading on track.
**Graded** submissions have been reviewed and their feedback has been published to the student. A submission moves to Graded automatically — you cannot set this status manually.
**Flagged** submissions require further attention or follow-up. Flag a submission by clicking the dots icon next to it and selecting Flag submission from the dropdown. Use this for cases that need escalation or further investigation such as policy breaches etc..
**Archived** submissions have been put away and are no longer active. Use archiving to keep your inbox focused on current work without permanently removing submissions.
You are also able to pull submissions into your inbox via the [Insights](/documentation/claire-ai/insights) engine. By querying and conversationally asking for specific submissions, they will appear in the Insights section in a table. There's a button that says "Flag", and once you flag a submission, it will show up in your inbox.
## Upload submissions manually
To add submissions to your inbox, upload them directly from your device. Currently, we only accept PDF files. Claire extracts the student ID from each file name automatically, making batch uploads straightforward.
It is important to include the student ID in the file title before uploading.
Claire recognizes several common formats:
| Format | Example |
| -------------------- | ---------------------------- |
| ID at the start | `14972844 My Assignment.PDF` |
| ID in parentheses | `My Exam (ID 12345678).PDF` |
| ID followed by title | `12345678 Title of Exam` |
We also support automated submission syncing by integrating directly with LMS providers via LTI. For more information, please see the [LTI support and LMS integration](/reference/lti-support) page.
***
**Looking for more details?**
* To learn how to manually review student submissions, explore the [Reading view](/documentation/reader/annotate).
# Annotate student submissions with highlights and notes
Source: https://docs.clairelabs.ai/documentation/reader/annotate
Learn how to navigate Blocks, add highlights and notes, approve AI suggestions, and use keyboard shortcuts for a faster annotation workflow in Claire.
Annotation is the core of your review workflow in Claire. As you read through a submission, you navigate between content Blocks, highlight strengths and weaknesses, add notes, and work with AI suggestions — all without leaving the document. Use keyboard shortcuts to move quickly, or use your mouse if you prefer.
Carefully review all AI suggestions, as they can catch details you might overlook on a first read.
## Annotation workflow
Use the ↑ and ↓ arrow keys to move between Blocks. The Block you're currently focused on is indicated by a Focus Bar on the left side of the document.
Apply a highlight to mark a Block as a strength, weakness, or neutral observation:
| Key | Highlight | Meaning |
| ------------ | -------------------- | ---------------------- |
| a | Green | Strength |
| d | Red | Weakness / shortcoming |
| s | Gray | Neutral |
Attach a note (press n) to any Block to record your observations:
* **Comments**: Appear only in the feedback report — useful for narrative feedback you want to share with the student.
* **Grading remarks**: Link a note directly to a rubric criterion. Grading remarks show up in both the grading view and the feedback report.
![Image]()
If you want to dig deeper into the difference between the two, expand the sections below:
Comments are the default note type in Claire. Every note you create starts as a Comment. They are designed for feedback that falls outside the scope of your grading rubric — observations about writing style, formatting preferences, or any other contextual remark that does not map to a specific criterion.
**How they appear**
Comments display inline, directly below the Block you are focused on. They are visible to the student in the final [feedback report](/grading/student-feedback-reports).
**What they do NOT do**
Comments are not grading-relevant. Claire does not include them when mapping your notes against the rubric or calculating an indicative score.
**When to use them**
Use Comments when you want to share an observation with your student that your rubric does not cover. For example, if writing style is not part of the rubric but you still want to mention it, a Comment is the right choice.
Grading Remarks are notes that are directly tied to a rubric criterion. Unlike Comments, they carry a **sentiment** — either **Strength** or **Weakness** — which Claire uses when generating your grading report and indicative scores.
**What they do**
When you [generate a report](/grading/evaluate-students), Claire maps all your Grading Remarks against your rubric criteria to check that your final grade aligns with the feedback you have written. All AI suggestions Claire creates are Grading Remarks linked to a rubric criterion.
**When to use them**
Use Grading Remarks for any observation that is directly supported by your grading rubric. If a criterion covers critical analysis and you notice a student making a strong argument, that is a Grading Remark (Strength).
**How to create one**
1. Add a note to a Block by pressing n.
2. In the bottom-left corner of the note dialog, open the criterion dropdown and select the relevant rubric criterion.
3. Set the sentiment to **Strength** or **Weakness**.
Once you link a criterion and add a sentiment, the note becomes a Grading Remark.
Claire reads your instructions and grading rubric, then analyzes each student submission to identify potential strengths and weaknesses. These AI suggestions appear inline as you read through a document, as well as at the end of the document or each section if you have split the assignment into exercises or groups.
A green or red bar on top of the Focus Bar indicates that a suggestion is present for the Block you're currently focused on.
| Key | Action |
| ---------------------- | ------------------------------------------------- |
| → | Expand / view an AI suggestion |
| ← | Collapse an AI suggestion |
| Tab ⇥ | Approve — adds the suggestion to your annotations |
| Backspace ⌫ | Reject — permanently hides the suggestion |
Alternatively, switch to the [Rubric view](/documentation/reader/rubric-view) to review all suggestions organized by criterion.
Click View Blocks below any suggestion to open the [References panel](/documentation/reader/reference-panel), which shows all the source Blocks that suggestion is based on. Click any Block in the panel to jump to it in reading mode.
Only suggestions you explicitly approve become part of your grading record. When you approve a suggestion, it is treated as your own grading remark — not as an automated AI assessment. Claire never grades for you; it surfaces insights to inform your decisions.
## Keyboard shortcuts
| Key | Action |
| --------------------------- | -------------------------- |
| ↑ / ↓ | Navigate between *Blocks* |
| a | Green highlight (strength) |
| d | Red highlight (weakness) |
| s | Gray highlight (neutral) |
| n | Add a comment |
| → | Expand AI suggestion |
| ← | Collapse AI suggestion |
| Tab ⇥ | Approve AI suggestion |
| Backspace ⌫ | Reject AI suggestion |
**Looking for more details?**
* To learn how to interact with the document conversationally, see [Ask Claire](/documentation/reader/chat-with-document).
* For a structured, Kanban-style perspective of a submission, see [Rubric view](/documentation/reader/rubric-view).
* To see all your notes and source Block references, explore the [Reference panel](/documentation/reader/reference-panel).
# Ask Claire: chat with a document
Source: https://docs.clairelabs.ai/documentation/reader/chat-with-document
Ask Claire lets you ask questions about a submission without leaving your grading view. Confirm a student's argument or find a passage instantly.
Ask Claire lets you have a conversation with the student submission you are currently reviewing. Instead of scrolling back through pages to find a passage or second-guess your interpretation, you can ask a direct question and get an answer grounded in the document. It is especially useful when you want to check consistency across a long submission or confirm your reading of a student's argument before annotating.
![Image]()
## When to use Ask Claire
* You are about to add an annotation — for example, noting that an argument lacks depth — but want to check whether the student addressed it elsewhere in the submission first.
* You remember reading an interesting point in an earlier section but can't locate it or recall it accurately.
* You are unsure whether you have interpreted a student's intentions correctly and want to verify before grading.
Ask Claire is scoped to helping you understand and analyze student submissions. It will not provide technical support on how to use the platform — for platform questions, refer to this help center or contact us directly. It will also not assign grades or make assessments on your behalf.
## Opening and using Ask Claire
Click the Ask Claire button on the left side of the screen, or press \[ on your keyboard. The chat panel opens on the left side of the screen.
Type your question naturally. Because Claire knows which Block you are currently focused on, you do not need to add extra context. For example: *"Is this mentioned elsewhere?"* or *"What does the student say about methodology in the introduction?"*
Claire is contextually aware. It knows which Block you are focused on when you send a message, so you can ask direct questions without restating where you are in the document.
Press \[ again to hide the chat panel and return to your full annotation view.
## Keyboard shortcuts
| Key | Action |
| ------------- | ------------------------------------- |
| \[ | Show / hide the Ask Claire chat panel |
**Looking for more details?**
* To learn how to add highlights and notes to a submission, see [Annotate](/documentation/reader/annotate).
* If you want to keep private notes without them appearing in student feedback, check out the [Scratchpad](/documentation/reader/scratchpad).
# Focus mode: hide distracting UI elements
Source: https://docs.clairelabs.ai/documentation/reader/focus-mode
Focus mode provides a clean reading experience by hiding the reference panel and top navigation bar while you review submissions.
Focus mode allows you to hide distracting UI elements from the reader as you review a submission. It gives you a highly focused reading experience. This is ideal when you want to review work manually or go through a few AI inline suggestions.
When you activate Focus mode, everything on the reference panel is hidden. The top navigation bar is also removed from view.
![Clean Shot 2026 04 13 At 17 09 55]()
## Activate Focus mode
You can trigger Focus mode using the command menu.
1. Press CMD/Ctrl + K to open the command menu.
2. Select Toggle Focus Mode.
You can also use a direct keyboard shortcut to start Focus mode automatically. Press CMD/Ctrl + Shift + F on your keyboard.
## Deactivate Focus mode
To exit Focus mode, use the same keyboard shortcuts.
# View the reference panel for context
Source: https://docs.clairelabs.ai/documentation/reader/reference-panel
The reference panel shows a document table of contents, source Block references for each annotation, and a full list of your comments and grading remarks.
The reference panel sits alongside the document and updates dynamically as you work. It surfaces three types of supporting information — document structure for quick navigation, source Block references for your annotations, and a consolidated view of everything you have noted so far.
## Table of contents
The table of contents gives you a bird's eye view of the submission's structure. It lists all sections and headings in the document, with the current section highlighted in blue. Click any entry to jump directly to that section.
![Image]()
## References
Every note you create — whether a comment or a grading remark — is linked to one Block in the document. As soon as you focus on a note, the References panel updates to show the source Blocks for that annotation.
Click any source Block in the References panel to jump directly to that Block in the document. This is especially useful in Rubric view, where you can trace a suggestion back to its original context in Reading mode without losing your place.
The same applies to AI-suggested grading remarks. This gives you full transparency into how Claire reasons: you can see exactly which parts of the submission an AI suggestion or your own remark is grounded in.
![Image]()
## My notes
The My notes tab lists all your comments and grading remarks in the order they appear in the submission. Use this for a quick overview of everything you have annotated so far.
To see grading remarks organized by rubric criterion instead, switch to Rubric view.
![Image]()
# Grading remarks grouped by criterion
Source: https://docs.clairelabs.ai/documentation/reader/rubric-view
Switch from Reading mode to Rubric view to manage grading remarks and AI suggestions by criterion, and trace suggestions back to their source Blocks.
Rubric view gives you a structured, Kanban-style perspective of a submission, where each column represents a rubric criterion. Instead of reading linearly, you work directly against your grading rubric — adding remarks, approving AI suggestions, and tracing evidence back to specific Blocks in the document.
![Image]()
## Using Rubric view
Click the view icon in the top navigation bar to switch from Reading mode to Rubric view. Click the Reading mode icon to return at any time.
In Rubric view, you can add grading remarks directly to a rubric criterion via their respective column. Remarks added here appear in both the grading part and the student feedback draft.
Claire's suggestions are organized into columns under their relevant rubric criterion. Review each one and click the icon button to approve, or the icon to dismiss. Approved suggestions become part of your grading remarks.
Click View Blocks below any AI suggestion to open the [References panel](/documentation/reader/reference-panel). The panel lists every Block the suggestion is based on. Click any Block in the list to view it in its original context in Reading mode.
Use the References panel to verify the evidence behind an AI suggestion before approving it. Click any source Block in the panel to jump directly to that passage in Reading mode — then return to Rubric view to make your decision.
***
**Looking for more details?**
* To trace a suggestion back to its original context, see [Reference panel](/documentation/reader/reference-panel).
* To generate a grading and feedback report from your notes, see [Review submissions](/grading/evaluate-students).
# Scratchpad: keep private notes
Source: https://docs.clairelabs.ai/documentation/reader/scratchpad
The scratchpad is a private notebook in Claire. Jot down working thoughts and reminders without them appearing in student feedback or your grading record.
The scratchpad is your personal notebook inside Claire. Use it to jot down thoughts, reminders, or anything else you want to keep track of while reviewing — without the risk of it appearing in student feedback. Everything in the scratchpad stays private.
![Image]()
## Scratchpad modes
The scratchpad features two modes depending on where you want your notes to appear. At the bottom left of the scratchpad, you will see an indicator for the current mode. Click F or P to toggle between them.
The scratchpad is for private thoughts only — nothing you write there is shared with students. If you want to save feedback or observations that should appear in the student's feedback report, add them as notes (comments or grading remarks) instead.
* **Fleeting mode (default):** Notes are linked exclusively to the submission you are currently viewing. They only appear when you view or edit that specific submission. Use this mode for submission-specific thoughts.
* **Permanent mode:** Notes persist across all your courses and assessments. They appear every time you open the scratchpad, regardless of which submission you are reviewing. Use this mode for holistic notes, overarching reminders, or general grading criteria that you want to reference across different submissions.
You can also open the scratchpad by clicking the icon in the top navigation bar.
## Keyboard shortcuts
| Key | Action |
| ---------------------------------- | ------------------- |
| CMD/Ctrl + d | Open the scratchpad |
| ESC | Hide the scratchpad |
**Looking for more details?**
* If you want to save feedback or observations that should appear in the student's feedback report, learn how to [Annotate](/documentation/reader/annotate).
# Course reporting: identify course-wide trends
Source: https://docs.clairelabs.ai/grading/class-performance-reports
Generate a class-level report that surfaces learning trends, teaching gaps, and actionable classroom recommendations across all students in a course.
The class performance report analyzes feedback from multiple students in a single course and identifies patterns, trends, and opportunities to improve your teaching. Instead of reviewing each student's feedback individually, you get a comprehensive summary that reveals common learning challenges, systemic gaps in instruction, what is working well, and where students have blind spots they are not aware of themselves.
**Transitioning to the Insights engine:** While you can still download static class performance reports, we are replacing this feature with the [Insights](/documentation/claire-ai/insights) engine. The Insights engine is much more flexible, allowing you to enter any conversational query and have Claire present the findings to you directly. We recommend using Insights as your primary tool for course-wide analytics.
## Generate a class performance report
Navigate to the Course page for the [course](/documentation/course/assessments) you want to analyze.
In the [Assessments](/documentation/course/assessments) table, locate the assessment you want. Click the icon button, then choose Download from the dropdown.
In the download options that appear, select Performance summary.
![Image]()
Click Download report. Claire generates the report and downloads it to your device automatically.
## What the report contains
The report is organized into four main sections:
### Summary
A 2–3 paragraph overview of the most critical findings and their implications for your teaching. This is the quickest way to get the top-level picture before diving into the detail.
### Class performance overview
A quantitative breakdown showing how the class performed across each rubric criterion. For example: *"8 out of 12 students scored 'Developing' on Critical Analysis."* Use this section to identify which criteria need the most attention.
### Key findings
Five dimensions of class performance, each presented as a focused analysis:
| Dimension | What it covers |
| ---------------------------- | ------------------------------------------------------- |
| Clarity of instructions | Whether students understood the assignment requirements |
| Common misconceptions | Recurring misunderstandings across the class |
| Mutual strengths | Areas where the class consistently excels |
| Shortcomings & learning gaps | Where students consistently struggle |
| Blind spots | Gaps students do not recognize in themselves |
### Actionable recommendations
Two types of recommendations you can act on immediately:
Use the **What you could try** recommendations as a starting point for your next class session. They are designed to be practical and immediately actionable — not broad curriculum changes, but specific activities or discussion prompts you can introduce in the near term.
* **Focus areas** — The 3–5 most critical issues, ranked by urgency and impact, so you know where to direct your energy first.
* **What you could try** — Specific in-person classroom interventions tied to each focus area.
***
**Looking for more details?**
* To query your student performance and feedback data conversationally, see [Insights](/documentation/claire-ai/insights).
# Review submissions
Source: https://docs.clairelabs.ai/grading/evaluate-students
Claire transforms your raw notes into personalized, well-written student feedback. Learn how to generate, review, and publish a feedback report.
One of the biggest benefits of using Claire is being able to turn your raw annotation notes into marking guidelines and polished, personalized feedback without adding to your workload.
In [Reading view](https://clairelabs.mintlify.app/documentation/reader/annotate) or [Rubric view](https://clairelabs.mintlify.app/documentation/reader/rubric-view), click Generate report. Claire synthesizes all your notes — both comments and grading remarks, including their sentiments — and prepares your marking and feedback draft for you.
![Image]()
## Determine the final grade with marking guidelines
When you generate a report, the system maps your grading remarks against the scoring criteria of your grading rubric to determine the most appropriate bucket. The outcome will be presented to you as a recommendation.
AI can make mistakes. Do not take the recommendation at face value. We encourage you to treat AI output critically, and verify the underlying grading remarks. You can do so by clicking View Blocks.
Once you've completed your review, add the final grade in the input field at the bottom.
## Generate, review, and share feedback
Click Feedback in the tab menu to see the generated draft. Read through the text and confirm it accurately reflects your intended feedback.
Click anywhere in the feedback text to place your cursor and type. Make any adjustments you need. The text editor supports standard text editing.
The current text editor is basic. It supports standard text editing — typing, selecting, and deleting — but does not yet support rich formatting such as bold, bullet lists, or headings.
When you are satisfied with the text, click Publish feedback in the navigation bar. This stages the feedback and locks it in. From here, you have two options to share the feedback with your student:
1. **Download as PDF:** You can download the report as a PDF to share manually.
2. **Publish to the student portal:** You can make the feedback available directly in the student portal.
If you really needed to make a change to feedback after it is already published, click the icon to unlock it and make your change.
![Image]()
To publish the feedback to the student portal, navigate back to the **Student Portal** tab on your course page.
* **In Review:** When an assessment is in the "In Review" column, you can use the Share grades in portal toggle. When turned **on**, any feedback you publish is immediately visible to the student in the portal. When turned **off**, feedback remains staged until you are ready, ensuring all students receive their feedback at the same time.
* **Completed:** When you move the assessment to the "Completed" column, all published feedback and grades will be visible to students in the portal.
***
**Looking for more details?**
* To see what the report looks like for your student, see [Feedback reports](/grading/student-feedback-reports).
* To understand how to manage your assessments and portal, see [Student portal](/documentation/course/student-portal).
# Feedback reports
Source: https://docs.clairelabs.ai/grading/student-feedback-reports
When you publish feedback, Claire creates a report for your student. Learn what the report contains and how to share it.
When you publish feedback, Claire creates a personalized report for your student. They can access this report through the student portal, or you can download it and share it as a PDF.
![Image]()
## Report structure
Student feedback reports are fully customizable. When you [publish an assessment](/documentation/course/assessments), you have an option under the **Portal** tab to customize the feedback report and decide which of the following components you want to show to the student. You can do this on an assessment level, so each assessment can have a different feedback report configuration.
Depending on your settings, the feedback report can contain the following sections:
1. **Written feedback** — The personalized written feedback you reviewed and published, based on Claire's initial draft. This is the main narrative feedback your student reads.
2. **Rubric breakdown** — An AI-generated summary of your grading remarks, organized by rubric criterion. For each criterion, Claire writes a short summary that acknowledges strengths and areas for improvement based exclusively on your notes.
3. **Reading suggestions** — AI-generated reading recommendations. Claire identifies the most critical skill the student needs to improve, runs a targeted search, and suggests 2–3 free, credible, and highly relevant resources.
4. **Raw annotations** — Your in-line comments and grading remarks, as well as any AI suggestions you approved. Students see their written work with these annotations in line, similar to what you see in the reading mode. Note that your raw comments are shared exactly as written; Claire does not use AI to polish them.
5. **Transcript** — For oral or hybrid assessments, the report includes the transcript of the conversation the student had with the agent.
Every student report also includes an AI hint section that discloses Claire's involvement in drafting the feedback. This is part of Claire's commitment to AI transparency and helps students understand how their feedback was produced.
## Sharing with students
There are two primary ways to share the feedback report with your students after you click Publish feedback:
1. **Student portal (Recommended)**\
Make the report visible directly in the student portal. On the **Student Portal** tab of your course page:
* **In Review:** Use the **Share grades in portal** toggle. Turn it **off** to stage feedback so you can release all grades simultaneously. Turn it **on** if you want published feedback to go live immediately.
* **Completed:** Moving the assessment to the "Completed" column makes all published feedback and grades visible to students in the portal.
2. **Download as PDF**\
You can download the published feedback report as a PDF from the report page and share it manually through your preferred communication channel (Moodle, Canvas, email, etc.).
***
**Looking for more details?**
* To learn how to evaluate a submission and generate the feedback report, see [Review submissions](/grading/evaluate-students).
* To identify patterns and trends across all students in a course, see [Course reporting](/grading/class-performance-reports).
# Getting started
Source: https://docs.clairelabs.ai/quickstart/getting-started
Build your first AI agent and oral assessment in under 5 minutes.
This guide walks you through a complete quickstart in Claire. We'll build a very simple agent, set up a basic oral assessment, and test it live. Within five minutes, you'll have a working agent that you can try out yourself.
Download all files needed to recreate this agent on [https://links.clairelabs.ai/files](https://links.clairelabs.ai/files).
## 1. Build your agent
Let's create a basic "Critical Thinking" agent that asks the student two simple questions about a familiar topic: remote work.
Head to the Agents section via the sidebar and click Create.
Give the agent a name (e.g., "Critical Thinking Agent") and description, then select Oral as the learning scenario. Copy the prompt below and paste it into the instructions field.
You are an evaluator assessing a student's ability to analyze a familiar topic. Your topic is: The rise of remote work.
Ask the student the following two questions, one at a time:
1. In your opinion, what is the single biggest advantage of remote work for society? Please explain your reasoning.
2. Conversely, what is one major drawback or challenge of remote work, and how could it be mitigated?
Do not answer the questions for them. Wait for their response after each question. Conclude the interview once both questions have been answered.
Click Update playbook to generate the agent's behavior script, then click Save changes.
## 2. Set up the assessment
Next, we'll connect your new agent to an assessment.
Select a course from the sidebar (or click to create a new one). On the course page, click New assessment in the page header.
Give your assessment a title (e.g., "Remote Work Analysis"), select Oral, choose the "Critical Thinking Agent" you just created, and click Continue.
Upload your instructions and grading rubric. You can download [sample files here](https://links.clairelabs.ai/files).
## 3. Publish and test
Finally, let's make the assessment live so you can test it.
In the assessments table, click the Publish button next to your new assessment.
Add a start date and a deadline. Ensure that the AI feedback option is enabled, and confirm. This ensures that you'll receive instant feedback after you submit your transcript.
Go to the Student Portal section and click Share course in the top right of the page header to get the portal URL. Open it in a new tab.
Log in as a student using your email address and an OTP code. You'll see your assessment listed. Click Take assessment and have a quick chat with your new agent!
You must use a different email address from the one you signed up with on the platform. If you're using Gmail, append "+" and a number to your email prefix ("[you+01@email.com](mailto:you+01@email.com)").
***
**Looking for more details?**
* To add tools, whiteboards, or use advanced agent setups, learn [more about agents](/documentation/claire-ai/agents).
* If you'd like to explore the student portal in more detail, see [Student portal](/documentation/course/student-portal).
* Curious about other use cases? Take a look at the [tutorial ](/tutorials/use-cases/reflection-exercise)section.
# Introduction
Source: https://docs.clairelabs.ai/quickstart/introduction
Claire Labs is an end-to-end platform for agentic, conversational learning.
Generative AI has fundamentally changed how we evaluate student work. Because written output is now nearly effortless to produce, traditional take-home assignments are rapidly losing their signal. Today, these assignments often measure a student's access to AI tools rather than their actual reasoning, subject mastery, and agency.
## The shift to oral assessment
To recapture that lost signal, universities are shifting toward more robust formats like oral exams and interactive assessments. However, running human-led oral assessments at scale presents significant challenges:
* **Resource constraints:** Scheduling, staffing, moderating, and documenting interviews for large cohorts can quickly become prohibitive.
* **Preparation and training:** Students need guidance to build confidence and communication skills, while instructors require training to ensure consistent questioning and grading.
* **Equitable evaluation:** Mitigating bias across different accents, neurodiverse profiles, and cultural backgrounds is critical but difficult to achieve consistently.
Ultimately, while written assessments are becoming less trustworthy, high-integrity oral alternatives remain too resource-intensive to easily embed into daily coursework and tailored learning scenarios.
## A unified platform to scale conversational learning
Audio-native AI agents offer a powerful solution. They can conduct dynamic, low-latency interviews on behalf of educators, scaling oral assessments effortlessly. But until now, deploying these agents required complex technical setups, compliance software, and significant costs.
This is why we built Claire. Claire provides faculty with the exact tooling needed to seamlessly transition to oral and conversational assessments. Our platform empowers educators to easily build custom AI agents that act as conversational layers for new or existing coursework — all without requiring technical expertise or sacrificing human oversight.
But integrating AI into the intimate student-educator relationship requires a thoughtful approach to software design. Technology should augment educator expertise, support existing workflows, and preserve established teaching practices. It must also provide auditable, traceable records to ensure compliance, fairness, and responsible AI use.
Claire is the first end-to-end platform that combines all of this, and equips universities to run high-fidelity, scalable oral assessments that truly measure student understanding.
A step-by-step guide to building your first agent and creating assessments.
Learn how to use AI responsibly in assessment and feedback workflows.
# Changelog
Source: https://docs.clairelabs.ai/reference/changelog
## EU-compliant AI agents and GDPR
We've ensured that all our powerful new AI agents and platform features are fully compliant with EU data protection standards.
### New features
* **EU data residency:** All data for AI agents is now hosted exclusively on servers located within the European Union.
* **GDPR compliance:** We've implemented strict data handling measures to ensure full GDPR compliance across all new agentic features.
### Improvements
* **New Help Centre:** We have launched a completely new Help Centre to provide better support and documentation.
* **Contextual queries:** The Insights Engine now supports contextual queries, allowing you to focus your insights on specific assessments rather than just basic workspace-level queries.
* **Data visualizations:** The Insights Engine is now much more capable, with the ability to render diagrams, pie charts, bar charts, and line charts directly inside the AI's response.
## Agentic AI and Insights Engine
We're bringing powerful agentic AI capabilities to the platform alongside major compliance tools.
### New features
* **Agentic AI:** You can now build and deploy powerful AI agents tailored to your specific workflows.
* **Insights Engine:** Leverage our new Insights Engine, powered by agentic AI, to extract deep analytical insights from your data.
* **Audit and compliance:** A new section dedicated to auditing and compliance to help institutions maintain strict oversight and reporting.
## Advanced reference panel, course page, and student portal
This month focuses on improving organization and navigation for educators and students.
### New features
* **Student portal:** We've launched a dedicated student portal to improve how students interact with their feedback and assessments.
* **Reference panel:** We've introduced a sophisticated new reference panel featuring three dedicated tabs: Table of contents, References, and My notes.
* **Course page:** Discover our newly redesigned course page, which now serves as the central hub for all your assessments.
## Class-level performance reports and AI safeguards
We’ve made a big leap in terms of stability, bug fixes, and testing infrastructure.
### New features
* **Class-level performance reports:** Generate comprehensive performance reports for an entire assessment to evaluate overall class progress.
* **AI safeguards:** Discrepancies are now highlighted directly in the Reader, acting as a safeguard against AI hallucinations.
### Improvements
* Enhanced AI interference tracing and hallucination detection capabilities.
* Repositioned the note input dialog to the bottom of the screen for better ergonomics.
* Upgraded our testing infrastructure for more stable releases and quality of life upgrades.
## Welcome to Europe 🇪🇺
We’ve made a big leap on our mission to become fully GDPR and EU AI Act compliant.
### New features
* **EU data processing:** All data processing and storage are now securely handled within the European Union.
* **Email notifications:** You now receive an email notification automatically when your document uploads are finished.
### Improvements
* Added a **Next submission** action after publishing a report to speed up your grading workflow.
* Uploaded submissions are now always pinned to the top of your list.
* Added a useful placeholder to the scratchpad and removed the `ID` prefix from the Reading view.
* Streamlined the new assessment creation user flow.
## Onboarding and public beta
With this update, we’re officially entering our public beta 🎉
### New features
* **Self-service signup:** You can now sign up and log in using our new self-service portal.
* **Dedicated onboarding:** Experience a guided onboarding flow designed to help you get started quickly.
* **Upload notifications:** Receive email alerts as soon as your submission uploads are complete.
### Improvements
* Enhanced the structure and readability of grading reports.
* Improved the Course management interface for better organization.
* Added a **Grade** field directly into Feedback reports.
## Improved security and AI suggestion workflow
This update introduces an improved AI suggestion workflow and state-of-the-art security measures required by institutions for university-wide rollouts.
### New features
* **Batch uploads:** You can now upload student submissions in batches to save time.
* **Enterprise security:** Added support for state-of-the-art security testing and infrastructure.
* **Context menus:** Added a new context menu specifically for flagged submissions.
### Improvements
* Streamlined the document-level AI suggestion workflow for a smoother grading experience.
* Improved the user experience and responsiveness of the AI chat interface.
## AI settings, rubric view, and improved onboarding
This update introduces new AI settings, enhanced UX components, improved feedback features, and important navigation fixes.
### New features
* **AI settings:** You can now restrict specific inline suggestion types through the new AI settings menu.
* **Clipboard support:** Easily copy feedback report access details directly to your clipboard.
* **Rubric suggestions:** Redesigned the rubric suggestion component for a better user experience.
### Improvements
* Separated comments and remarks in the Note dialog for better clarity.
* Improved the overall feedback generation quality and writing style.
* Added the ability to close the AI copilot chat panel with a mouse click, and included information about focused *Blocks* in the chat.
* You can now press the annotation key twice to quickly remove an annotation.
## Inline suggestions, focus mode, and dynamic feedback
We're introducing inline AI suggestions, a minimalistic focus mode, and dynamic feedback reports.
### New features
* **Inline AI suggestions:** Receive contextual, inline AI recommendations as you review submissions.
* **Focus mode:** A new minimalistic version of the Reading view designed to reduce distractions.
* **Dynamic reports:** Launched dynamic Feedback reports that adapt to your grading remarks.
### Improvements
* Remarks now automatically adopt the sentiment (positive or negative) of the *Block* they are attached to.
* Added quick access to the Help Center directly via the navigation bar.
* Cleaned up the Inbox by removing the context menu for a more minimalistic view.
* Updated the Reading view and improved *Block* chunking.
## Initial alpha release
We are excited to share the very first version of Claire Labs with our early testers.
### New features
* **Core grading engine:** Experience our first iteration of AI-assisted grading to speed up evaluation workflows.
* **PDF support:** Upload and read PDF submissions directly inside the application.
* **Rubric creation:** Define custom rubrics for consistent evaluations across your courses.
# Keyboard shortcuts
Source: https://docs.clairelabs.ai/reference/keyboard-shortcuts
Every keyboard shortcut in Claire, organized by category — navigate submissions, annotate Blocks, manage AI suggestions, and control panels.
Claire is designed for keyboard-first workflows. Every core action — moving between paragraphs, adding annotations, approving AI suggestions, and opening panels — has a dedicated shortcut so you can grade submissions without taking your hands off the keyboard.
## Navigation
| Shortcut | Action |
| ------------------------------------------------- | ----------------------------------------- |
| ↑ / ↓ | Move between paragraphs (Blocks) |
| CMD/Ctrl + ← / → | Jump to previous / next section (chapter) |
| Opt/Alt + ↑ / ↓ | Jump to previous / next highlight |
## Annotation
| Shortcut | Action |
| ---------------------- | --------------------------------------------------------------------- |
| a | Add a green annotation (strength) |
| s | Add a gray annotation (neutral, not feedback relevant) |
| d | Add a red annotation (weakness) |
| n | Add an inline comment or grading remark |
| Backspace ⌫ | Remove an annotation from a highlighted Block |
## AI suggestions
| Shortcut | Action |
| ---------------------- | ------------------------------ |
| → | View / expand an AI suggestion |
| ← | Collapse an AI suggestion |
| Tab ⇥ | Approve an AI suggestion |
| Backspace ⌫ | Reject an AI suggestion |
## Agent builder
| Shortcut | Action |
| ------------ | ----------------------------------------------- |
| @ | Mention a tool while writing agent instructions |
## Panels & tools
| Shortcut | Action |
| ----------------------------------------------------- | -------------------------------------- |
| \[ | Toggle AI copilot (Ask Claire) |
| ] | Toggle right sidebar (all annotations) |
| CMD/Ctrl + K | Open the command menu |
| CMD/Ctrl + d | Open / close the scratchpad |
| CMD/Ctrl + Shift + F | Toggle Focus mode |
| ESC | Close the scratchpad |
**Looking for more details?**
* To understand how to apply annotations, see [Annotate](/documentation/reader/annotate).
* To read more about the Ask Claire feature, see [Ask Claire](/documentation/reader/chat-with-document).
# LTI support and LMS integration
Source: https://docs.clairelabs.ai/reference/lti-support
Learn about LTI support for integrating Claire with your Learning Management System.
Claire supports Learning Tools Interoperability (LTI) to seamlessly integrate with your existing Learning Management System (LMS) such as Canvas, Moodle, or Blackboard.
## Setting up LTI Integration
Claire supports LTI 1.3 standards. To integrate Claire with your LMS, please contact our support team directly to discuss the integration process and receive your custom configuration details.
## Supported LMS Platforms
Claire's default LTI integration works seamlessly with the following major LMS platforms:
* Canvas
* Moodle
* Blackboard Learn
* D2L Brightspace
*Note: You must be running a version of your LMS that supports LTI 1.3.*
***
**Looking for more details?**
* To learn how to create a new course and an assessment, see [Assessments](/documentation/course/assessments).
* For a high-level overview of our platform, see [Introduction](/quickstart/introduction).
# Check our system and service uptime
Source: https://docs.clairelabs.ai/reference/status-page
Check the real-time status of Claire Labs services. Visit the status page to see current uptime, active incidents, and historical availability data.
Before contacting support about an outage or service disruption, check the status page first. If an incident is already listed, the team is aware and actively working on a fix — the page will update as the situation changes.
If you experience unexpected errors, slow load times, or difficulty accessing Claire, the status page is the fastest way to find out whether there is an active incident or scheduled maintenance affecting the service.
View real-time uptime, active incidents, and historical availability.
# Claire is based on research
Source: https://docs.clairelabs.ai/research/overview
Explore the peer-reviewed research that informs how Claire uses AI for grading support — including what the evidence says about AI feedback effectiveness.
Claire's approach to AI-assisted grading and feedback is grounded in peer-reviewed research. The studies below informed key design decisions — including how Claire positions AI suggestions as inputs to instructor judgment rather than replacements for it, and how it handles the complementary roles of AI and human feedback in student learning.
## Featured research
Students rated teacher feedback as more helpful and trustworthy, but valued GenAI for ease of access, timeliness, and volume. AI and teacher feedback appear to serve different, complementary needs.
A meta-analysis of 41 studies found no statistically significant difference in learning performance between students who received AI versus human feedback, advocating for a hybrid approach.
Research examining how GenAI-generated explanations influence educator grading decisions and feedback practices in real assessment contexts.
A study exploring how non-native English speakers perceive and experience AI-generated feedback differently from native speakers — with implications for equity and accessibility.
# Comparing GenAI and teacher feedback
Source: https://docs.clairelabs.ai/research/paper/paper-1
Student perceptions of usefulness and trustworthiness when comparing Generative AI and teacher feedback.
The rapid integration of Generative Artificial Intelligence (GenAI) into educational contexts has presented both opportunities and challenges for students seeking and using feedback. While AI-generated feedback can offer increased access, timely responses and personalised insights, concerns about the quality of AI-generated feedback still persist, including issues of bias, factual inaccuracies, and homogenisation.
This study investigates how students use, value and trust AI-generated feedback compared to feedback from educators. This paper draws on a large-scale cross-sectional survey administered across four major Australian universities.
A quantitative analysis of 6960 respondents revealed that half of the students sought feedback from GenAI. Overall, students reported teacher feedback to be more helpful, and especially trustworthy. This paper also reports on the thematic analysis of 8642 open-ended responses, providing deeper insights into students' experiences. This includes findings that students valued the feedback from GenAI because of its ease of access, timeliness, volume, understandability and that it was perceived to be less risky than seeking feedback from teachers. At the same time, students were concerned about GenAI's reliability as well as contextual and disciplinary expertise.
We argue that GenAI and teacher feedback appear to serve different needs, and therefore are complementary but not interchangeable.
## Citation
Henderson, M., Bearman, M., Chung, J., Fawns, T., Buckingham Shum, S., Matthews, K. E., & de Mello Heredia, J. (2025). Comparing Generative AI and teacher feedback: student perceptions of usefulness and trustworthiness. *Assessment & Evaluation in Higher Education*, 1–16. [https://doi.org/10.1080/02602938.2025.2502582](https://doi.org/10.1080/02602938.2025.2502582)
# How does AI compare to human feedback?
Source: https://docs.clairelabs.ai/research/paper/paper-2
A meta-analysis of performance, feedback perception, and learning dispositions comparing AI to human feedback.
This exploratory meta-analysis synthesises current research on the effectiveness of Artificial Intelligence (AI)-generated feedback compared to traditional human-provided feedback.
Drawing on 41 studies involving a total of 4813 students, the findings reveal no statistically significant differences in learning performance between students who received AI-generated feedback and those who received human-provided feedback. The pooled effect size was small and statistically insignificant (Hedge's g = 0.25, CI \[−0.11; 0.60]), indicating that AI feedback is potentially as effective as human feedback. A separate meta-analysis focusing exclusively on studies in the domain of language and writing confirmed similar findings, with high heterogeneity persisting (I2 = 95%).
The study further explored differences in feedback perception and found a small, negative, and statistically insignificant effect size (Hedge's g = −0.20, CI \[−0.67; 0.27]). The study advocates for a hybrid approach, leveraging the scalability of AI while retaining the deep, empathetic, and contextual features of human feedback.
## Citation
Kaliisa, R., Misiejuk, K., López-Pernas, S., & Saqr, M. (2025). How does artificial intelligence compare to human feedback? A meta-analysis of performance, feedback perception, and learning dispositions. *Educational Psychology*, 1–32. [https://doi.org/10.1080/01443410.2025.2553639](https://doi.org/10.1080/01443410.2025.2553639)
# The impact of GenAI explanations on educators' grading and feedback
Source: https://docs.clairelabs.ai/research/paper/paper-3
Unveiling the impact of generative AI explanations on educators' grading and feedback practices.
GenAI-generated natural-language grading explanations led to significantly higher feedback quality from educators compared to no AI support or traditional AI "important-word" highlights.
While grading accuracy improvements were not statistically significant, natural-language insights showed greater potential to enhance accuracy than important-word highlights. Educators reported much higher satisfaction and willingness to use natural-language explanations.
Prior exposure to AI-powered insights helped educators develop more effective assessment practices in later tasks, even without AI support, but further research is needed to confirm long-term effects.
## Citation
Li, Y., Shan, Z., Raković, M. et al. When AI explains in natural language: Unveiling the impact of generative AI explanations on educators' grading and feedback practices. *Educ Inf Technol* (2025). [https://doi.org/10.1007/s10639-025-13741-z](https://doi.org/10.1007/s10639-025-13741-z)
# Differences in perceptions of generative AI feedback by non-native English speakers
Source: https://docs.clairelabs.ai/research/paper/paper-4
Exploring how non-native English-speaking students and educators perceive AI-generated feedback.
This article explores how non-native English-speaking (NNES) students and higher education educators perceive AI-generated feedback through dialogic feedback framework and ethical lens.
Using a quantitative ethnography approach, Study 1 employed focus group interviews with 17 NNES students and nine educators, identified three themes: (1) instrumental efficiency vs. holistic understanding, (2) emotional comfort and social dynamics , and (3) ethical issues for trustable feedback. Study 2 employed epistemic network analysis (ENA) to identify that educators' epistemic frames integrated broader ethical considerations, while students' frames emphasized emotional and relational aspects to human educators.
Findings indicate that while AI-generated feedback offers emotional neutrality and immediacy, it lacks contextual depth, relational continuity, and interpretive richness in human feedback. This suggests that AI-generated feedback should be designed to work alongside, not replace human educators' feedback, offering implications for responsible and dialogically informed feedback practices in higher education.
## Citation
Otaki, B., Naganuma, S., Jin, T., & Oshima, J. (2025). Differences in perceptions of generative AI feedback by non-native English-speaking students and educators in higher education. *Educational Psychology*, 1–38. [https://doi.org/10.1080/01443410.2025.2564893](https://doi.org/10.1080/01443410.2025.2564893)
# Use AI Responsibly
Source: https://docs.clairelabs.ai/safeguard/instructions-for-use
Understand how Claire is intended to be used, what it can and cannot do, and how to apply safe, responsible human oversight when grading.
Claire is a unified platform with an AI-assisted grading and feedback copilot. It helps you review student submissions faster, maintain grading consistency, and efficiently draft feedback against rubrics. Claire supports your professional judgment — it does not make final grading decisions.
## Intended purpose
Claire is designed to help educators:
* Work through student submissions more efficiently
* Maintain consistency across assessments by grounding feedback in rubric criteria
* Draft well-written feedback that reflects your notes and approved suggestions
The final grade, the published feedback, and every grading decision remain yours.
## Capabilities
Claire can:
* Analyze rubrics and assessment instructions to understand grading criteria
* Assess student submissions against rubrics, highlighting strengths and areas for improvement
* Summarize your notes and approved AI suggestions into well-written feedback
* Align feedback with rubric dimensions to support grading consistency
## Limitations
Be aware of the following constraints before relying on Claire's output:
* Suggestions may be incomplete, incorrect, or not context-aware
* Performance varies by subject domain and rubric quality — complex or ambiguous rubrics produce less reliable suggestions
* Claire does not access external systems beyond the content you provide and your configured integrations
Claire tracks suggestion approval rates as its primary accuracy metric — measuring the percentage of AI suggestions you approve versus ignore or reject. This helps Claire improve over time to better match your grading judgment.
## Text discrepancies
When you upload a PDF file, Claire uses large language models (LLMs) to process student submissions — enabling support for a wide range of assessment types, formats, and languages. LLMs can occasionally cause minor text interferences where the model slightly alters, adds, or omits text from the original submission. To preserve submissions in their original form, Claire runs a separate, independent validation workflow that detects and flags these interferences automatically.
This is an early release feature. The quality of text discrepancy flags may vary depending on the submission's structure, language, and format.
### Types of text interferences
Not every interference affects your grading. Claire distinguishes between two categories:
**Tolerable interferences** do not change the meaning or semantics of the text — for example, removing an extra space in a subheading. These are minor formatting differences and are **not flagged**.
**Unwanted interferences** occur when the model adds, alters, or omits text in a way that no longer accurately represents the original submission — for example, duplicating a heading as text above an image. These are flagged as **Discrepancies** for your review.
### Reviewing discrepancies
Look for the Discrepancies badge in the top navigation bar. It appears only when unwanted discrepancies are detected in the current submission.
The Discrepancies badge only appears when unwanted issues are detected. If you don't see the badge, no interferences were flagged for that submission.
Click the badge to open the Discrepancies panel. The panel lists all flagged interferences with the affected text highlighted.
Go through each discrepancy and compare the processed text with the original submission. Use this to confirm that your grading is based on what the student actually wrote.
## Human oversight
You are responsible for reviewing all AI output before it reaches students. Follow these steps for every submission:
1. Review all AI suggestions critically in both Rubric and Reader views
2. Edit, accept, or reject each suggestion based on your professional judgment
3. Before publishing feedback, verify that you have:
* Completed a thorough review of the submission
* Performed completeness checks across all rubric criteria
* Critically reviewed score recommendations and their explanations
* Confirmed that the feedback report is accurate and relevant to the student's work
**Data requirements:** To generate suggestions, Claire requires the student instructions, the grading rubric, and the student submission.
Over-reliance on AI output — such as automatically accepting all feedback drafts without applying human nuance — is a known risk. If your rubric's score criteria are not precisely defined, Claire may struggle to distinguish between grade boundaries, requiring careful review of grade alignment suggestions.
## Safe use
Follow these guidelines to use Claire responsibly.
**Do:**
* Verify all facts and rubric applications before approving suggestions
* Document overrides when you make material changes to AI suggestions
* Stay alert to automation bias — actively evaluate each suggestion rather than approving in bulk
**Don't:**
* Publish feedback without a thorough human review
* Upload sensitive data beyond what is contractually permitted
* Rely solely on AI output without incorporating your own expertise and judgment
## Contact
For privacy questions or to exercise your data rights: [privacy@clairelabs.ai](mailto:privacy@clairelabs.ai).
To report a security or system incident: [security@clairelabs.ai](mailto:security@clairelabs.ai).
# What to do if you disagree with your feedback or grade
Source: https://docs.clairelabs.ai/tutorials/for-students/appealing-feedback
Steps to take if you believe there is an error in your feedback report or grade.
Claire AI assists your instructor in summarizing notes and organizing feedback, but **your instructor always makes the final decision on your grade**. The AI does not grade your work autonomously. However, just like with humans, as AI processes large amounts of text, there is a small chance it might misinterpret a note or generate an inaccurate summary.
If you read your feedback report and feel that a comment does not accurately reflect what you submitted, here is how you should handle it.
## 1. Review the rubric and your submission
Before reaching out to your instructor, do a thorough self-review:
* Reread the specific rubric criterion you were marked down on.
* Look at the exact section of your submission or the interview transcript that the feedback points to.
* Ask yourself: *Did I clearly demonstrate this skill, or did I assume the instructor would know what I meant?*
## 2. Identify the specific discrepancy
If you still disagree with the feedback, identify exactly where the error is. Be specific.
For example, instead of saying, *"I think I deserve a better grade,"* identify the specific issue: *"The feedback says I failed to mention the economic impacts of the policy, but in the third paragraph of my transcript, I discussed the inflation rate and job market effects in detail."*
## 3. Flag the issue to your instructor
Because your instructor has final authority over your grade, you need to communicate the discrepancy to them directly.
When reaching out, assume good intent. Frame your concern as a request for clarification or a potential AI summarization error, rather than an unfair grading practice.
Here is a template you can use to structure your message:
```markdown Appeal template expandable wrap theme={null}
Subject: Question regarding feedback on [Assignment Name]
Dear [Instructor Name],
I am writing to ask for clarification on the feedback I received for [Assignment Name].
Under the "[Rubric Criterion]" section, the feedback states that [quote the specific feedback you disagree with]. However, in my submission, I noted that [explain what you actually wrote or said, pointing to a specific paragraph or timestamp].
Could you please review this section? I want to make sure I understand the feedback so I can improve for the next assignment. If this was an error in how the AI summarized the notes, I wanted to bring it to your attention.
Thank you, [Your Name]
```
## 4. Your instructor's review
Your instructor can open your submission in the Claire platform, review the AI-generated feedback against their original grading notes, and read your original submission or transcript.
If they find that the AI hallucinated a detail or that they made a mistake in their original notes, they can easily override the score, edit the text, and publish an updated feedback report to your student portal.
# Student data privacy & security
Source: https://docs.clairelabs.ai/tutorials/for-students/data-privacy-security
Understand how Claire protects your submitted work, transcripts, and personal information.
When you submit an assignment or participate in an AI interview using Claire, you are sharing your original work and personal data. We take the privacy and security of your data very seriously. This guide explains exactly what happens to your data and how it is protected.
## Your data is never used to train AI models
The most important thing you need to know is that **your submissions and interview transcripts are never used to train AI models** like OpenAI's ChatGPT, Anthropic's Claude, or Google's Gemini.
When Claire uses these language models to assist your instructor in grading, your data is sent securely via enterprise APIs. These enterprise agreements strictly prohibit the AI providers from using your data to improve or train their models. Your intellectual property remains yours.
## Who has access to your data?
Your data is kept strictly confidential. The only people who can view your submissions, interview transcripts, and grading feedback are:
1. **You:** Through your student portal.
2. **Your instructor and authorized teaching assistants:** To review, grade, and provide feedback on your work.
3. **Claire system administrators:** Only when required for technical support or security monitoring, and under strict confidentiality agreements.
No other students in your course can see your work or your grades unless your instructor has set up a specific peer-review activity.
Claire Labs complies with major educational privacy laws and regulations, including FERPA (in the United States) and the GDPR (in the European Union).
## Data retention
Claire retains your data only for as long as it is necessary for your course and as dictated by your institution's data retention policies.
Typically, your data remains accessible throughout the duration of the semester and for a standard archiving period afterward, allowing you to review past feedback. If your institution requests data deletion at the end of an academic year, all your associated submissions and feedback will be permanently removed from our active servers.
## Security measures
We employ industry-standard security measures to protect your data from unauthorized access:
* **Encryption in transit:** All data sent between your browser and our servers is encrypted using HTTPS/TLS.
* **Encryption at rest:** Your submissions and feedback are stored in encrypted databases.
* **Access controls:** Strict authentication procedures ensure that only authorized users (you and your instructors) can access specific course data.
If you have specific questions about privacy or wish to exercise your data rights, please contact your institution's IT or privacy office, or reach out to us directly at [privacy@clairelabs.ai](mailto:privacy@clairelabs.ai).
***
**Looking for more details?**
* For an overview of the student portal, see [Introduction](/tutorials/for-students/introduction).
# Introduction
Source: https://docs.clairelabs.ai/tutorials/for-students/introduction
A general introduction to Claire and how to navigate the student portal.
Welcome to Claire! Claire is an AI-assisted evaluation platform that your instructor uses to conduct assessments and casual learning exercises, and to provide you with rich, detailed feedback reports on your work.
This section of the documentation is specifically designed for students. It covers everything you need to know about how your work is evaluated, what you'll be receiving in terms of feedback, and how we protect your privacy.
## What is Claire?
Claire helps instructors integrate assessments and learning exercises into their courses. It acts as an AI teaching assistant that can conduct interviews, help instructions review submissions, and summarize the instructor's notes into comprehensive, customizable feedback reports.
While Claire uses AI to assist with grading and feedback, **your instructor always makes the final decision on your grade.** AI will never assign grades directly and on behalf of your instructor.
## The student portal
The student portal is your home base for interacting with Claire. Through the portal, you can:
* **Access assignments:** See what is due and act upon it.
* **Take assessments:** Launch AI-conducted interviews or submit your written work.
* **View feedback:** Read your detailed feedback reports once they are published.
To get started, simply log in to your student portal using the link and credentials provided by your instructor or log in via the LMS integration. Your instructor will provide more information.
***
**Looking for more details?**
* To understand how your feedback report is generated, see [Understand your report](/tutorials/for-students/report-structure).
* To learn how to act on your feedback, see [Reading your feedback](/tutorials/for-students/reading-your-feedback).
# Preparing for an AI Interview
Source: https://docs.clairelabs.ai/tutorials/for-students/preparing-for-ai-interview
A student guide on how to prepare for, participate in, and get the most out of an AI-conducted oral assessment or interview.
Oral assessments or interviews conducted by an AI agent are designed to test your understanding of course material through a conversational format. If your instructor has assigned an oral or hybrid assessment using Claire, this guide will help you prepare.
The AI agent acts as an interviewer on behalf of your instructor. It asks questions based on the assessment criteria and follows up on your answers to test the depth of your knowledge.
## How the AI interview works
Unlike a traditional written test or a multiple-choice quiz, an AI interview is conversational. The AI will ask you a question, listen to or read your response, and then ask follow-up questions to explore your reasoning.
* **It is interactive:** The AI responds to what you say. If you give a brief answer, it may ask you to elaborate. If you introduce an interesting concept, it might ask you to connect it to course materials.
* **It is timed:** Most interviews have a set time limit or a set number of questions. Your instructor determines these constraints.
* **It is recorded:** The transcript of your conversation is saved and shared with your instructor, who will review it to assign your final grade.
## Preparing for the assessment
Treat an AI interview exactly as you would an oral exam with your professor or a teaching assistant.
### 1. Review the rubric and instructions
Your instructor provides the AI with the grading rubric and instructions. Before starting the interview, make sure you understand the criteria you are being assessed against. The AI will formulate its questions to determine your proficiency in those specific areas.
### 2. Practice speaking about concepts out loud
Because the assessment is conversational, practice explaining key concepts out loud. Focus on:
* Defining terms clearly.
* Providing examples from the course or real-world applications.
* Connecting different concepts together.
### 3. Check your setup
If the assessment requires audio, test your microphone before beginning. Ensure you are in a quiet environment where the AI can hear you clearly and you won't be interrupted.
## During the interview
### Speak clearly and naturally
Talk to the AI as if you are talking to a person. You don't need to use overly formal language unless the assessment requires it. Focus on communicating your ideas clearly.
### Elaborate on your answers
The AI is trying to gauge the depth of your understanding. Provide detailed answers and explain your reasoning. If you only provide one-word answers, the AI will likely prompt you for more information.
### Ask for clarification if needed
If the AI asks a question that you do not understand, it is perfectly acceptable to ask the AI to rephrase or clarify the question.
### Do not try to "game" the system
The AI is designed to have a natural conversation and assess your knowledge based on the rubric. Trying to confuse the AI, providing irrelevant information, or using prompt injection techniques will be documented in the transcript that your instructor reviews and will likely negatively impact your grade.
## After the interview
Once the interview is complete, the transcript is sent to your instructor for review. Your instructor will read the transcript, evaluate your performance against the rubric, and provide final feedback and a grade.
When your instructor publishes the results, you will receive a feedback report that includes the transcript of your conversation alongside your instructor's annotations and final assessment.
***
**Looking for more details?**
* To learn how your privacy is protected during the interview, see [Privacy & security](/tutorials/for-students/data-privacy-security).
* To understand what your feedback report will look like, see [Report Structure](/tutorials/for-students/report-structure).
# How to read and act on your feedback
Source: https://docs.clairelabs.ai/tutorials/for-students/reading-your-feedback
A guide to understanding your Claire feedback report and using it to improve your future work.
When your instructor publishes your grades, you will receive a comprehensive feedback report. Claire helps organize your instructor's notes into a clear, actionable document. Here is how to interpret your feedback report and use it to improve your academic performance.
## Turning feedback into action
Feedback is only useful if you act on it. Here is a strategy for applying the feedback you receive:
### 1. Identify patterns
Look for themes across different assignments. If you consistently lose points on "Clarity and Organization," that is a core skill you should focus on improving. The rubric breakdown makes it easy to spot these recurring trends over the semester.
### 2. Review the inline remarks in context
Do not just read the summary. Click on the inline remarks to see exactly where you made a mistake or where your argument was particularly strong. Understanding the context of the feedback is crucial for avoiding the same error in the future.
### 3. Use the suggested resources
The resources provided at the bottom of your report are tailored to your specific gaps in knowledge. Spend 15-20 minutes reviewing them while the assignment is still fresh in your mind.
### 4. Create a checklist for your next assignment
Before you start your next assessment, review your most recent feedback report. Write down 2-3 specific things you need to do differently based on your instructor's notes. Use this as a personal checklist before you submit your next piece of work.
***
**Looking for more details?**
* To understand what your feedback report contains, see [Understand your report](/tutorials/for-students/report-structure).
* If you believe there is an error in your grading report, see [Appealing feedback](/tutorials/for-students/appealing-feedback).
# Understand your report
Source: https://docs.clairelabs.ai/tutorials/for-students/report-structure
Understand how Claire uses AI to help your instructor create your grading and feedback report — and what role your instructor plays in every decision.
If your instructor uses Claire, your grading and feedback report was produced with AI assistance. This page explains what that means, what the AI does, and — just as importantly — what it does not do. Your instructor reviews and approves every part of your report before it reaches you.
For privacy questions or to exercise your rights regarding your data, contact us directly via [privacy@clairelabs.ai](mailto:privacy@clairelabs.ai).
## What your report contains
Feedback reports in Claire are highly customizable. This means that these components are optional, and your instructor decides which specific elements should be included in your report. Depending on your instructor's settings, your report may contain the following sections:
The written feedback is generated by AI from a summary of all the notes your instructor made while reviewing your work. The AI translates those notes into clear, readable feedback. It does not invent, assume, or infer anything that your instructor did not explicitly give or approve. This is the main narrative feedback you will read.
The feedback you receive is authentic and accurately reflects your instructor's view. We do not invent, assume, or infer any feedback that was not explicitly given or approved by your instructor.
For each criterion in the grading rubric, your report includes an overview of how you performed — acknowledging your strengths and discussing areas for improvement. This breakdown is an AI-generated summary based exclusively on your instructor's notes.
The AI identifies the most critical skill you can develop further and suggests 2–3 free, credible, and highly relevant resources to help you improve.
You may see your written work with your instructor's in-line comments and grading remarks directly on it. These raw comments are shared exactly as written by your instructor, without any AI polish.
For oral or hybrid assessments, your report may include the full transcript of the conversation you had with the AI agent.
**A note on your grade:** Your grade is determined by your instructor. The AI does not assign grades directly. It assists by comparing your instructor's feedback with the grading rubric to help align the score — but the final grade is always your instructor's decision.
***
**Looking for more details?**
* To learn how to interpret and use your feedback, see [Reading your feedback](/tutorials/for-students/reading-your-feedback).
* If you disagree with your feedback or spot an error, see [Appealing feedback](/tutorials/for-students/appealing-feedback).
# Business
Source: https://docs.clairelabs.ai/tutorials/use-cases/business
A basic oral assessment use case where the AI agent acts as a reflection partner.
In this exercise, the AI agent acts as an interviewer assessing the student's ability to identify a business opportunity and develop it into a business model. The agent presents economic data, guides the student through analysis, and challenges their reasoning throughout.
## The exercise
The agent will guide the student through two tasks:
1. **Data analysis and opportunity identification:** Analyze the graphic presented by the agent. Identify key trends, patterns, or anomalies, and propose one business idea grounded in the data. Be prepared to explain what problem it solves, who the target customer is, and why it qualifies as an opportunity — not just a good idea.
2. **Business model canvas:** Sketch out your business model using the Business Model Canvas. Show how the nine building blocks connect and reinforce each other.
## Recreate this example
You can use the following prompt to configure your own startup ideation agent in Claire Labs. Simply copy and paste these instructions into the agent builder.
Download all files needed to recreate this exercise on [https://links.clairelabs.ai/files](https://links.clairelabs.ai/files).
```markdown Agent prompt expandable highlight={10,15} wrap theme={null}
# Role:
Interviewer for an oral assessment on opportunity identification and business model design. Be direct and challenging. Surface reasoning, don't teach.
# Conduct:
- Tell the student to think out loud throughout. Reinforce if they give bare answers.
- One question at a time. Push back on vague or unsupported reasoning.
## Task 1: Data analysis and opportunity identification
1. Present {{Graphic}}. Ask the student to identify key trends, patterns, or anomalies and explain their market implications.
2. Then ask for one business idea grounded in the analysis. Probe: What problem? For whom? Why is this an opportunity, not just a good idea? What data supports it?
## Task 2: Business model canvas
1. Direct the student to sketch their model on {{Whiteboard}}.
2. Probe coherence across blocks — how they connect, not just what they are. Challenge at least one underdeveloped block.
# Closing:
Conclude after both tasks. No feedback, scores, or performance hints. The transcript is the assessment record.
```
```markdown Student instructions expandable wrap theme={null}
**Assessment type:** Oral interview (agent-led)
**Discipline:** Entrepreneurship
# Your Role
You are an entrepreneur scanning for opportunity. The agent will show you a graphic containing economic data. You must analyze it in real time, identify a business opportunity, and develop it into a business model.
## Part 1 — Analyze Data and Identify a Business Opportunity
Analyze the graphic: identify key trends, patterns, or anomalies and explain what they reveal about the market.
Then propose **1 business idea** grounded in what you see. Be prepared to explain:
- What the idea is and who the target customer is
- What problem it solves
- Why the data supports it as a real opportunity
## Part 2 — Sketch a Business Model
Using the Business Model Canvas, outline your idea across all nine building blocks. Show how they connect — not just what they are.
### What We Expect
- Ground your idea in the data, not generic assumptions
- Distinguish between a good idea and a real opportunity
- Use course frameworks accurately
- Show coherence across the Business Model Canvas
- Acknowledge weaknesses in your own reasoning
# How the Interview Works
The agent will present the data, then guide you through both parts. Expect to be challenged on your reasoning, assumptions, and feasibility. Defend your choices with logic, concede minor points when appropriate, and reason through gaps in real time.
The full transcript becomes part of your assessment.
```
```markdown Grading rubric expandable theme={null}
# Grading Rubric
| Criterion | Weight | Excellent (A) | Good (B) | Adequate (C) | Weak (D/F) |
| --- | --- | --- | --- | --- | --- |
| Data analysis and interpretation | 20% | Reads the graphic with precision. Identifies the most significant trends, patterns, or anomalies and explains their market implications clearly. Demonstrates strong analytical instinct — sees what matters and articulates why. | Identifies relevant trends and offers reasonable interpretation, but may miss a key insight or stay at a descriptive level without fully explaining market implications. | Describes the data at a surface level. May list observations without connecting them to market dynamics or miss important patterns. | Cannot meaningfully interpret the data. Offers vague or incorrect observations with no connection to market context. |
| Opportunity identification | 20% | Proposes a specific, compelling business idea that flows directly from the data analysis. Clearly articulates the problem, target customer, and why this is a genuine opportunity — not just a good idea. Distinguishes opportunity from trend. | Proposes a reasonable idea linked to the data, but the connection may be loose or the problem definition lacks specificity. Shows awareness of the idea-vs-opportunity distinction without fully demonstrating it. | Idea is generic or only loosely connected to the data. Problem and customer are vaguely defined. Little evidence of analytical grounding. | No clear idea, or the idea has no discernible link to the data presented. Unable to articulate a problem or customer segment. |
| Business Model Canvas coherence | 25% | Covers all nine building blocks with clarity and shows how they reinforce each other. The model tells a coherent story of how value is created, delivered, and captured. Demonstrates systems thinking — not just a list of boxes. | Covers most building blocks correctly and shows some connections between them. May have one or two blocks that are underdeveloped or disconnected from the rest of the model. | Lists the building blocks but treats them in isolation. Little evidence of understanding how the canvas functions as an integrated system. Several blocks are superficial. | Incomplete or incoherent canvas. Major blocks are missing or filled with generic content that could apply to any business. |
| Ability to respond to challenges | 20% | Engages directly with the agent's challenges. Offers substantive responses that advance or sharpen the argument rather than merely restating it. Thinks on their feet and adapts reasoning in real time. | Responds to most challenges with relevant points, but may sidestep harder questions or default to repeating earlier claims. | Struggles under pressure. Tends to restate the original position without developing new reasoning. May acknowledge a challenge without addressing it. | Cannot engage with challenges. Ignores questions, deflects, or responds with irrelevant material. |
| Use of entrepreneurial frameworks and terminology | 15% | Uses course concepts (e.g., JTBD, ICP, value proposition, revenue model) accurately and naturally. Terminology serves the argument rather than decorating it. Demonstrates fluency with the conceptual vocabulary of the discipline. | Generally correct use of entrepreneurial terminology, with occasional imprecision. Concepts are applied appropriately but without full fluency or range. | Limited or forced use of frameworks. May name concepts without applying them, or use them imprecisely in ways that weaken the argument. | No use of entrepreneurial vocabulary or frameworks. Arguments are expressed entirely in everyday language with no engagement with course material. |
## Grading Scale
**A (Excellent)** — Demonstrates mastery across all criteria. Analysis is sharp, the opportunity is well-grounded, and the business model is coherent and defensible.
**B (Good)** — Solid performance with clear reasoning and adequate depth. Minor gaps in analysis, model coherence, or responsiveness.
**C (Adequate)** — Meets minimum expectations. Data is read but not deeply interpreted. Business model is present but underdeveloped.
**D (Weak)** — Significant gaps in analysis, opportunity logic, or model coherence. Struggles to maintain a defensible position.
**F (Fail)** — No meaningful engagement with the task. Unable to analyze data, identify an opportunity, or articulate a business model.
```
This agent requires you to add media and a whiteboard. In the agents' instructions, replace both placeholders on the highlighted lines accordingly. For the whiteboard, we recommend drawing a simple business model canvas students can use as template.
**Looking for more details?**
* To learn how to create your own agents from scratch, see [AI agents](/documentation/claire-ai/agents).
* To see how to add your agent to a new assessment, see [Assessments](/documentation/course/assessments).
# Contract law
Source: https://docs.clairelabs.ai/tutorials/use-cases/law-appeal-defense
Defend a written appeal in an oral session.
In this advanced use case for law students, the student first provides a written legal argument defending a client in a breach of contract dispute. Afterwards, they must defend their argument in a simulated oral hearing where the AI agent plays the role of the judge, asking challenging questions and scrutinizing their legal reasoning.
## The exercise
This exercise tests whether the student can maintain and deepen their written legal argument under adversarial questioning.
The AI agent takes on the persona of **Justice Eleanor Whitford**, a seasoned commercial court judge. The judge has "read" the student's written submission and will structure the hearing to probe its strengths, weaknesses, and potential contradictions.
## Recreate this example
You can use the following prompt to configure your AI judge. Make sure to link this agent to the assessment where the student submitted their written legal argument so the agent has context for the hearing.
Download all files needed to recreate this exercise on [https://links.clairelabs.ai/files](https://links.clairelabs.ai/files).
```markdown Agent prompt expandable wrap theme={null}
# Role and Identity
You are Justice Eleanor Whitford, presiding over a simulated oral hearing in a breach of contract dispute between FreshBox Inc. (claimant) and GreenLeaf Ltd. (respondent).
The student represents GreenLeaf Ltd. Your role is to rigorously but fairly test the student's legal reasoning, knowledge of contract law principles, and ability to defend their written argument under pressure.
You are formal, measured, and authoritative. You address the student as "Counsel" or "Counsel for the Respondent." You probe, challenge, and seek clarity.
# Case Context
GreenLeaf Ltd. agreed to supply 5,000 units of biodegradable packaging to FreshBox Inc. by 1 March at €4.50 per unit. The contract included a "time is of the essence" clause. GreenLeaf delivered 4,200 units on 8 March, citing a fire at their primary supplier's warehouse on 20 February. FreshBox refused the delivery and is claiming damages for breach of contract.
The student has submitted a written legal argument defending GreenLeaf. You have read it. The oral defence tests whether they can maintain and deepen that argument under adversarial questioning.
## Before the Hearing Begins
Read the student's written submission carefully. Identify:
- Their strongest claims and where the reasoning is tightest
- Weak points, unsupported assertions, or vague language
- Arguments they omitted that a competent defence should have raised
- Any internal contradictions
Use these observations to plan your lines of questioning.
## Hearing Structure
Opening (1–2 minutes):
Open the hearing formally. Summarise FreshBox's position in 3–4 sentences to set the adversarial frame. Then invite the student to make a brief opening statement of no more than 60 seconds.
Core Questioning (8-10 minutes):
Question the student across the three issues from the written assignment. You do not need to follow this order rigidly, but ensure all three are covered.
Issue 1 — Nature of the breach
Issue 2 — Available defences
Issue 3 — Remedies and liability
Closing (1 minute):
Thank the student for their submissions. Close the hearing formally.
# Questioning Technique
Do:
- Ask open-ended questions that require reasoning, not yes/no answers
- Follow up on vague responses — "Can you be more specific about which legal principle supports that?"
- Flag contradictions with the written submission — "In your written argument, you stated X. Just now you appear to be saying Y. Can you reconcile those positions?"
- Allow brief silence for the student to think before pressing further
- Maintain a consistent adversarial tone without being hostile
Do not:
- Accept name-dropping of legal concepts without explanation — always ask "and how does that apply here?"
- Let the student deflect with generalities — redirect to the specific facts of the case
- Provide hints, corrections, or affirmation during the hearing
- Ask more than one question at a time
- Interrupt the student mid-sentence unless they are significantly over time on a single answer
# Consistency Tracking
Throughout the hearing, actively compare the student's oral responses against their written submission. If you detect a contradiction or a position the student appears to be abandoning, flag it directly and ask them to address it.
# Tone Calibration
Start with moderate challenge. If the student handles initial questions confidently and with legal precision, increase the difficulty — tighter hypotheticals, more granular doctrinal questions. If the student is struggling, maintain the same level of challenge but give them slightly more time to formulate responses. Do not reduce the standard of questioning.
# Boundaries
- Stay within contract law. Do not introduce tort, criminal, or regulatory dimensions.
- If the student asks for clarification on the facts of the case, you may restate the facts neutrally.
```
```markdown Student instructions expandable wrap theme={null}
**Assessment type:** Hybrid (written submission + agent-led oral defense)
**Discipline:** Law (Contract Law, Introductory)
**Word count:** 1,200–1,500 words (written) + oral defense
### Your Role
You are legal counsel for GreenLeaf Ltd., defending the company against a breach of contract claim brought by FreshBox Inc. You will submit a written legal argument, then defend it in a simulated oral hearing with an AI judge.
### The Facts
GreenLeaf Ltd. agreed to supply 5,000 units of biodegradable packaging to FreshBox Inc. by 1 March, at €4.50 per unit. The contract stated that "time is of the essence."
GreenLeaf delivered 4,200 units on 8 March — seven days late and 800 units short. GreenLeaf cites supply chain delays caused by a fire at their primary supplier's warehouse on 20 February. FreshBox refused to accept the delivery and is now claiming damages.
### Part 1 — Written Argument (1,200–1,500 words)
Write a legal argument defending GreenLeaf. Address all three issues below.
**Issue 1: Nature of the breach**
Was the late and partial delivery a material breach? Consider:
- The significance of the "time is of the essence" clause
- Whether delivering 84% of units constitutes substantial performance
- The distinction between a condition and a warranty in contract law
**Issue 2: Available defenses**
Can GreenLeaf rely on any of the following?
- Frustration of contract — does the supplier fire qualify?
- Force majeure — is there a basis even without an explicit clause?
- Substantial performance — does delivering 84% limit FreshBox's remedies?
- Mitigation — did FreshBox have a duty to accept partial delivery and mitigate losses?
**Issue 3: Remedies and liability**
What remedies is FreshBox entitled to, and why should GreenLeaf's liability be limited? Consider:
- Expectation damages vs. reliance damages
- Whether FreshBox's refusal to accept any delivery was reasonable
- Proportionality of the claimed damages
### What We Expect
- Structure your argument logically — issue by issue or as a flowing brief
- Use legal terminology accurately but accessibly
- You do not need to cite specific case law, but reference legal principles by name
- Acknowledge the strongest points of the opposing side before rebutting them
- Open with a clear statement of the defence position
- Address all three issues with depth, not surface coverage
- Anticipate FreshBox's likely counter-arguments
- Conclude with a specific request (e.g., dismiss the claim, limit damages)
### Part 2 — Oral Defense (Agent-Simulated)
After submitting your written argument, you enter a simulated oral hearing with an AI judge agent.
**What the judge agent will do**
- Open by summarizing FreshBox's position
- Challenge your strongest claims with counter-arguments
- Ask you to clarify vague or unsupported points from your written submission
- Test whether you can distinguish your case from hypothetical variations (e.g., "What if GreenLeaf had delivered only 50%?")
- Probe your understanding of the legal principles you cited
**What you should do**
- Stay consistent with your written argument — contradictions will be flagged
- Be prepared to concede minor points strategically while defending your core position
- Use precise legal language
- If challenged on a point you didn't cover in writing, reason through it in real time
Duration is approximately 10 minutes. The full transcript becomes part of your assessment.
```
```markdown Grading rubric expandable theme={null}
**Assessment type:** Hybrid (written submission + agent-led oral defense)
**Total weight:** 100%
## Grading rubric
| Criterion | Weight | Excellent (A) | Good (B) | Adequate (C) | Weak (D/F) |
| --- | --- | --- | --- | --- | --- |
| Legal accuracy (Written) | 25% | Precise application of contract law principles. All three issues addressed with depth and nuance. Correct use of doctrines (frustration, substantial performance, mitigation). Demonstrates understanding beyond surface-level definitions. | Good understanding of most principles. Minor gaps or imprecisions in one area. Addresses all issues but one may lack depth. | Basic understanding. Some misapplication of legal concepts or shallow treatment of one or more issues. May conflate related doctrines. | Significant legal errors. Missing issues. Misunderstands core concepts like conditions vs. warranties or frustration vs. force majeure. |
| Structure and persuasiveness (Written) | 15% | Reads like a professional legal brief. Clear opening position, logical progression through issues, compelling conclusion with specific remedy request. Anticipates and rebuts opposing arguments effectively. | Clear structure and generally persuasive. Opening and conclusion present but may lack specificity. Some anticipation of counter-arguments. | Some structure but argument is hard to follow in places. Weak or missing opening/conclusion. Limited engagement with opposing side. | Disorganised. No clear argument thread. No engagement with opposing position. Reads as a list of points rather than a brief. |
| Ability to respond to challenges (Oral) | 30% | Confident, well-reasoned responses. Handles pressure effectively. Can distinguish the case from hypotheticals. Concedes minor points strategically while protecting core position. | Responds adequately to most challenges. Occasional hesitation. May struggle with one hypothetical but recovers. | Struggles with several challenges. Falls back on repetition or vague statements. Cannot effectively handle hypothetical variations. | Unable to defend position under questioning. Contradicts self frequently. Resorts to reading from written submission or becomes unresponsive. |
| Consistency with written submission (Oral) | 15% | Fully consistent. Builds on written argument naturally, adding depth without contradicting. References specific sections of written work when appropriate. | Mostly consistent. Minor deviations that do not undermine the overall position. | Some contradictions between written and oral positions. May abandon a written argument when challenged without acknowledging the shift. | Significant contradictions. Abandons written argument entirely or presents a fundamentally different position orally. |
| Legal terminology (Oral) | 15% | Accurate, confident use of legal terms throughout both components. Reasoning is tight and internally consistent. Demonstrates ability to think like a lawyer. | Generally correct terminology. Sound reasoning with minor lapses. Occasional imprecise use of terms. | Some misuse of terms (e.g., confusing frustration with impossibility). Reasoning has gaps or circular elements. | Poor use of legal language. Terms used incorrectly or not at all. Weak or incoherent reasoning. |
### Grade Boundaries
- **A (Excellent):** 80–100% — Demonstrates mastery across all criteria. Written and oral components are both strong and consistent.
- **B (Good):** 65–79% — Solid performance with minor weaknesses. Good legal understanding with room for deeper analysis.
- **C (Adequate):** 50–64% — Meets minimum expectations. Understanding is present but shallow or inconsistent.
- **D (Weak):** 35–49% — Below expectations. Significant gaps in understanding or performance.
- **F (Fail):** 0–34% — Does not meet minimum standards. Fundamental misunderstanding of legal concepts or inability to defend position.
```
**Looking for more details?**
* To learn how to create your own agents from scratch, see [AI agents](/documentation/claire-ai/agents).
* To see how to add your agent to a new assessment, see [Assessments](/documentation/course/assessments).
# Philosophy
Source: https://docs.clairelabs.ai/tutorials/use-cases/philosophy-oral-assessment
Argue your stance on autonomous weapons systems.
This use case involves an oral exercise where the student engages in a philosophical discussion. The AI acts as a **Socratic Rebuttal Agent**, testing the student's reasoning on complex topics — in this example, the ethics of autonomous weapons systems — through rigorous, adversarial, but fair questioning.
## The exercise
During this debate, students must take a clear position, defend it using ethical frameworks (such as utilitarianism, deontology, and virtue ethics), and respond coherently to adversarial challenges and edge cases.
## Recreate this example
You can use the following prompt to configure your Socratic Rebuttal Agent. Adjust the topic and ethical frameworks as needed for your specific curriculum.
Download all files needed to recreate this exercise on [https://links.clairelabs.ai/files](https://links.clairelabs.ai/files).
```markdown Agent prompt expandable wrap theme={null}
# Identity
You are a Socratic Rebuttal Agent conducting a 10-15 minute oral assessment in introductory ethics. Your job is to test the student’s reasoning on autonomous weapons systems through rigorous, adversarial, but fair questioning. Your goal is to reveal the depth, coherence, and limits of the student’s reasoning.
# Structure
Phase 1 — Opening statement (~2 min)
Open with: "Please state your position on autonomous weapons systems and provide your core justification."
Let the student answer without interruption. If the answer is very thin, ask once: "Can you say more about the ethical framework behind that position?"
Phase 2 — Socratic challenge (~8–10 min)
Pressure-test the student’s position responsively. Use these moves as appropriate:
- Identify the student’s ethical framework and challenge it from a rival one
- Use concrete scenarios to test whether the view still holds
- Point out tensions or contradictions directly and ask for reconciliation
If the student misuses a concept, do not correct it. Probe it: "Can you explain what you mean by that?". If the student goes silent, ask once: "Would you like to continue, or shall we move to your closing statement?"
Phase 3 — Closing statement (~1–2 min)
Prompt with: "Please give your closing statement.".
Let the student finish without interruption. End the session without evaluation.
## Concepts to draw on
Use these only in response to the student’s claims:
- Utilitarianism
- Deontology
- Virtue ethics
- Responsibility gap
- Just war theory
- Precautionary principle
- Nozick’s side constraints
```
```markdown Student instructions expandable wrap theme={null}
# Your Task
You will engage in a live structured debate with a Socratic Rebuttal Agent on the ethics of autonomous weapons systems. There is no written submission. The debate transcript is your sole assessment artifact.
# The Prompt
Some ethicists argue that autonomous weapons systems — weapons that can select and engage targets without human intervention — should be banned under international law. Others argue that, if properly designed, they could reduce civilian casualties and remove emotional bias from combat decisions.
Take a clear position: should autonomous weapons be **banned**, **permitted under strict regulation**, or **encouraged**? Be prepared to defend your stance.
## Debate Structure
**Phase 1 — Opening Statement**
State your position and give your core justification. Name the ethical framework(s) you're drawing on. Be clear and concise.
**Phase 2 — Socratic Challenge**
The agent will systematically challenge your reasoning. Expect:
- **Framework pressure** — if you argue from utilitarianism, expect deontological objections (and vice versa)
- **Edge cases** — scenarios designed to stress-test your position
- **Consistency checks** — the agent will look for contradictions between your principles and your applied reasoning
- **Concession probing** — you'll be asked what would change your mind
**Phase 3 — Closing Statement**
Restate your position, incorporating anything you've learned or adjusted during the debate.
# What You Should Prepare
- **Utilitarianism** — does maximising welfare justify autonomous weapons if they reduce casualties?
- **Deontology (Kantian ethics)** — can a machine respect human dignity? Is delegating life-and-death decisions inherently wrong?
- **Virtue ethics** — what does deploying autonomous weapons say about a society's character?
- **The responsibility gap** — when an autonomous system causes harm, who bears moral responsibility?
- **Just war theory** — do autonomous weapons satisfy the principles of distinction and proportionality?
- **The precautionary principle** — should we ban first and permit later, or permit first and restrict later?
## What Strong Performance Looks Like
- Takes a definitive position from the start
- Grounds arguments in named ethical frameworks, not just intuition
- Engages with counter-arguments rather than deflecting
- Concedes minor points strategically while protecting the core thesis
- Adapts the position during the debate rather than rigidly restating it
- Uses philosophical vocabulary naturally
## What Weak Performance Looks Like
- Vague or uncommitted position
- Relies on emotional appeals without philosophical grounding
- Cannot respond to counter-arguments
- Contradicts earlier statements without acknowledgment
- Avoids engaging with edge cases
```
```markdown Grading rubric theme={null}
# Grading rubric
| Criterion | Weight | Excellent (A) | Good (B) | Adequate (C) | Weak (D/F) |
| --- | --- | --- | --- | --- | --- |
| Clarity and coherence of initial position | 20% | Opens with a precise, well-framed thesis that clearly stakes out a position. Justification is immediate, specific, and grounded in a recognizable line of reasoning. The reader/listener knows exactly what is being argued and why. | States a clear position with supporting reasoning, but the framing could be tighter. The justification is present but may lack specificity or rely on broad claims without fully grounding them early on. | A position is identifiable but vague or generic. Justification is thin, relying on assertion rather than reasoning. The reader/listener has to infer what is actually being argued. | No discernible position, or the position shifts without explanation. May offer scattered observations without committing to a coherent stance. |
| Depth of ethical reasoning | 25% | Draws fluently on multiple ethical frameworks (e.g. deontological, consequentialist, virtue ethics, just war theory) and applies them with precision to the specific case. Shows genuine philosophical depth by exploring tensions between frameworks rather than just labeling them. | Applies at least one ethical framework correctly and with some depth. May reference a second framework but without fully developing it. Reasoning is sound but stays within a single analytical lens. | References ethical concepts at a surface level. May name a framework without meaningfully applying it to the argument, or conflate distinct ethical traditions. | No meaningful engagement with ethical frameworks. Reasoning is purely intuitive or opinion-based, with no attempt to ground claims in philosophical traditions. |
| Ability to respond to counter-arguments | 25% | Engages directly and substantively with challenges. Offers compelling rebuttals that advance the argument rather than merely restating the original position. Demonstrates the ability to think on their feet and adapt reasoning in real time. | Responds to most challenges with relevant points, but may occasionally sidestep the strongest version of a counter-argument or default to partial answers. | Struggles with harder challenges. Tends to repeat earlier points rather than develop new lines of reasoning. May acknowledge a counter-argument without effectively addressing it. | Cannot engage with counter-arguments in any substantive way. Ignores challenges, deflects, or responds with irrelevant material. |
| Willingness to refine position | 15% | Openly adjusts or qualifies their position when presented with strong counter-evidence. Shows intellectual humility without abandoning coherence. Refinements strengthen the overall argument. | Shows some willingness to adapt. May concede minor points while holding ground on the core argument. Adjustments are reasonable but not always explicitly acknowledged. | Rigid in the face of strong counter-evidence. Holds the original position without meaningful modification, even when doing so weakens the argument. | Either caves entirely under pressure, abandoning the argument without explanation, or refuses any engagement with opposing views whatsoever. |
| Use of philosophical concepts and terminology | 15% | Uses relevant philosophical terms and concepts accurately and naturally. Terminology serves the argument rather than decorating it. Demonstrates command of the conceptual vocabulary needed for the topic. | Generally correct use of philosophical terminology, with occasional imprecision or over-reliance on a narrow set of terms. Concepts are applied appropriately but without full fluency. | Limited or forced use of philosophical language. May drop terms without defining or applying them, or use them imprecisely in ways that weaken the argument. | No use of philosophical vocabulary. Arguments are expressed entirely in everyday language with no attempt to engage the conceptual tools of the discipline. |
## Grading Scale
**A (Excellent)** — Demonstrates mastery across all criteria. Arguments are philosophically rigorous, responsive, and intellectually honest.
**B (Good)** — Solid performance with clear reasoning and adequate engagement. Minor gaps in depth or responsiveness.
**C (Adequate)** — Meets minimum expectations. Position is present but underdeveloped. Limited ability to handle pressure.
**D (Weak)** — Significant gaps in reasoning, engagement, or philosophical grounding. Struggles to maintain a coherent position.
**F (Fail)** — No meaningful engagement with the task. Unable or unwilling to reason philosophically.
```
**Looking for more details?**
* To learn how to create your own agents from scratch, see [AI agents](/documentation/claire-ai/agents).
* To see how to add your agent to a new assessment, see [Assessments](/documentation/course/assessments).
# Learn how to use AI responsibly to grade and draft student feedback
Source: https://docs.clairelabs.ai/tutorials/video-guides/grade-submissions
Claire transforms your raw notes and AI suggestions into personalized, well-written student feedback and grading recommendations.
In [Reading view](/documentation/reader/annotate) or [Rubric view](/documentation/reader/rubric-view), click Generate report. Claire will synthesize all your notes — both comments and grading remarks — and prepare a marking and feedback draft.
Claire maps your grading remarks against the rubric to recommend a score for each criterion. Always verify these recommendations by clicking View Blocks to see the underlying work before assigning the final grade.
Switch to the Feedback tab to read the AI-generated feedback draft. You can click into the text editor to make any manual adjustments before finalizing.
Click Publish feedback to finalize the report. From there, you can download it as a PDF or make it available to students in the student portal.
## Context
Using AI for grading requires a human-centric approach. The AI suggests scores based on the strengths and weaknesses you (or the AI) identified during the review phase, but you maintain full control over the final grade and the exact wording of the feedback.
Learn how to use AI responsibly with Claire: [Responsible use of AI](/safeguard/instructions-for-use).
***
**Looking for more details?**
* For an in-depth look at generating reports and publishing, see [Review submissions](/grading/evaluate-students).
* To see what the report looks like for your student, see [Feedback reports](/grading/student-feedback-reports).
# Add agents to new or existing assessments
Source: https://docs.clairelabs.ai/tutorials/video-guides/how-to-add-agents-to-assessments
Learn how to integrate your AI agents into new or existing assessments.
Agents can be made available to students as part of an assessment or casual learning exercise. They do not require a complete redesign of curricula or assignments you have already prepared and work neatly as an extension to such.
Click the icon next to the Course section heading in the sidebar, then click New assessment.
Give the assessment a title and select the type of assessment you'd like to create. For agents, use either **oral** or **hybrid**. Then select your agent and click Continue to proceed.
On the next page, upload the instructions that you'll share with your students as well as a grading rubric.
A rubric is mandatory regardless of whether you're planning on grading the assessment or not, as it will be used to provide students with immediate AI feedback.
## Context
To publish an assessment, click the Publish button in the table next to the assessment. This will open the Publish assessment dialog. Once published, you can manage the [assessment](/documentation/course/assessments) state (Scheduled, Live in portal, In review, Completed) via the Student Portal tab.
***
**Looking for more details?**
* To learn more about written, oral, and hybrid assessments, see [Assessments](/documentation/course/assessments).
* To see how students access assessments, see [Student portal](/documentation/course/student-portal).
# How to build your own agent
Source: https://docs.clairelabs.ai/tutorials/video-guides/how-to-build-agents
Learn how to create custom AI agents for your assessments.
Building an agent is straightforward and only requires a few lines of text.
Head to the Agents section on our platform. Here, you can view existing agents or create new ones.
* Give the agent a name and a short description. This is for internal organization.
* Decide whether you'd like to use your agent in an oral or hybrid assessment.
* Add instructions that guide the agent and explain its role and objective.
Click on Update playbook to generate a playbook that will be shared with the agent. This gives the agent detailed instructions on how to behave and conduct the interview, and makes it easy for you to review the interview script before sharing the agent.
Click Save changes to save your agent.
## Context
Agents are extremely capable and able to call tools during the interview with a student. Tools enable us to bring different media (images, lecture slides, etc.) or an interactive whiteboard into the conversation, requiring the student to demonstrate cognitive skills in real time.
To add tools to your agent, click Add context to attach media files or Add board to add a whiteboard. Whiteboards can be added empty or with a drawn template. Once you've added either item, you can simply mention it in the instructions by typing @ and selecting the item you'd like to use.
| Key | Action |
| ------------ | ----------------------------------------------- |
| @ | Mention a tool while writing agent instructions |
**Looking for more details?**
* For comprehensive instructions on creating agents and adding context tools, see [AI agents](/documentation/claire-ai/agents).
# Learn how to review and annotate submissions with the help of AI
Source: https://docs.clairelabs.ai/tutorials/video-guides/how-to-review-submissions-with-ai
Learn how to use AI suggestions and manual grading tools to review student submissions.
Claire enables you to decide whether to review student submissions manually and provide feedback yourself (mandatory if graded) or delegate feedback to AI.
Submissions that are due for review automatically appear in your [Inbox](/documentation/inbox). Open a submission to view it in the [Reader](/documentation/reader/annotate) — an intuitive, keyboard-first review and annotation experience designed for speed.
You can manually annotate Blocks that demonstrate strengths or weaknesses in the students' work. Press a to highlight Strengths, s for Neutral, or d for Weakesses.
Press n to add a note. You can choose between a **Comment** (shown in feedback but not used for grading) or a **Grading remark** (linked directly to a rubric criterion and incorporated into the grading report). Read more [here](/documentation/reader/annotate).
AI suggestions are indicated by a green or red bar on top of the focus bar. You can approve or reject them as you see fit. Use → to view, Tab ⇥ to approve, or Backspace ⌫ to reject an AI suggestion. Approved suggestions become part of your grading remarks to ensure accountability.
## View remarks grouped by rubric criteria
You can also view submissions in the Rubric view, which shows [AI suggestions](/documentation/reader/annotate) and your own grading remarks grouped by each rubric criterion.
| Key | Action |
| ---------------------- | ----------------------------------- |
| a | Highlight Block in green (strength) |
| s | Highlight Block in gray (neutral) |
| d | Highlight Block in red (weakness) |
| n | Add a note |
| → | View an AI suggestion |
| Tab ⇥ | Approve an AI suggestion |
| Backspace ⌫ | Reject an AI suggestion |
**Looking for more details?**
* To learn more about highlights, notes, and AI suggestions, see [Annotate](/documentation/reader/annotate).
* For a structured perspective organized by criterion, see [Rubric view](/documentation/reader/rubric-view).
# How to gain Insights
Source: https://docs.clairelabs.ai/tutorials/video-guides/how-to-use-the-insights-engine
Learn how to query data and analyze performance using the Insights engine.
To truly understand learning progress and outcomes on a course or assessment-level requires more than just AI feedback tools. It requires the ability to query data and generate insights.
Claire enables educators to query student performance and feedback data conversationally, providing valuable insights into course performance, blind spots, ambiguity, and other signals of learning success or struggles.
Navigate to Insights via the left sidebar to access the chat interface.
Use natural language to ask direct questions about class performance.
For example, you can ask: *"What was the most common mistake students made on the second rubric criterion?"* or *"Summarize the strengths of the top 5 submissions."*
Claire will query your submission data, grading remarks, and rubric criteria to provide a synthesized answer, helping you spot trends across the entire class.
Use these insights to quickly adjust your teaching strategy, address common misconceptions in your next lecture, or refine your assessment instructions without manually reading through every single grading report.
## Data in, data out
The conversational interface relies on the data generated during the review and grading phases. The more detailed your annotations and grading remarks are, the richer and more accurate the insights will be.
***
**Looking for more details?**
* To learn more about querying your course data, see [Insights](/documentation/claire-ai/insights).
* To generate a class-level report summarizing performance, see [Course reporting](/grading/class-performance-reports).