Applies ToMicrosoft Teams Microsoft Teams for Education

What is AI Feedback Suggestion?

AI Feedback suggestions is a new feature that makes it easier for educators when they want to provide feedback to students as part of their Assignments flow using Microsoft 365 Copilot  In Assignments, educators can choose to have Copilot review their summarized feedback and enhance it, educators that are using Rubric can make have Copilot review their Rubric selection and summarize it.

Copilot always use Educator input as the basis of the suggested summarized feedback, and the educator will always need to confirm the suggested update and add potential edits, before returning to the student.  

An educator should not assign grades based on AI generated feedback.

What can AI Feedback Suggestion do?

Copilot can do the following:

  • Enhance summarized feedback – Copilot can suggest updated summarized feedback, when using Assignment Speedgrader. For enhance summarized feedback the educator must first add a minimum of 50 characters  before copilot can expand on the feedback. Once generated the educator can choose to have the feedback written in different tones. Basic, Instructional or Coaching tone. The educator can choose which tone to use.

    • Basic Tone: Clear and concise. Provides straightforward feedback without extra detail, ideal for quick corrections or clarifications.

    • Instructional Tone: Actionable and detailed. Offers step-by-step guidance to help students understand how to improve and build specific skills.

    • Coaching Tone: Supportive and motivational. Encourages self-reflection, fosters growth, and inspires students to take ownership of their learning journey.

  • Rubric Selection Summary – Copilot can suggest a summary of your rubric selection when using an assignment with a rubric. For rubric selection summary, the educator first selects the rubric values that reflect the student's performance. Once generated, the educator can choose to have the feedback written in one of three tones: Basic, Instructional, or Coaching. This allows the educator to tailor the tone to the student's needs. If the educator has also included rubric feedback and summarized feedback, Copilot will also include this as input to the suggested summarized feedback.

In addition to the educator initial feedback, and rubric selection, Copilot will also consider the  Rubric, Grade, Language and Feedback tone when Copilot generates the suggested feedback  

What is AI Suggested Feedback intended use(s)?

AI Suggest Feedback is intended for educators in the Speedgrader. This feature is only available to educators and educators must review and approve the generated content before it is shown to students. The features always use educator input for any suggested feedback it generated.  

An educator should not assign grades based on AI generated feedback.  

What are the limitations of AI Suggested Feedback? How can users minimize the impact of limitations when using the system?  

AI-suggested feedback is not grounded in factual knowledge and while it writes in a fluent, grammatically correct way, the content it generates can be inaccurate or inappropriate. It can’t understand meaning or evaluate accuracy, so educators need to review what it writes, and use their best judgment. Educators must validate and review all generated feedback to ensure accuracy and relevance. 

To improve the quality of the output, educators should provide clear and detailed input, such as well-defined assignment instructions and rubrics. The feature is optimized for text-based content, such as essays or written assignments, and may not perform as effectively with numeric or science-focused content. For the best results, use this feature with assignments that require detailed written feedback. 

An educator should not assign grades based on AI generated feedback.  

What operational factors and settings allow for effective and responsible use of AI suggested Feedback?

Effective and responsible use of AI-suggested feedback requires thoughtful implementation of operational factors and careful configuration of settings. Key considerations include: 

  1. Clear and Detailed Input 

    • Provide precise assignment instructions and well-constructed rubrics to guide the AI in generating contextually relevant feedback.

    • Ensure rubrics are aligned with the desired learning outcomes and contain clear performance criteria.

  2. Educator Review and Validation 

    • Always review and validate AI-generated feedback to confirm accuracy, relevance, and alignment with the specific needs of the student.

    • Avoid over-reliance on AI; use it as a complement to human judgment, not a replacement.

  3. Content Type Alignment 

    • Recognize the system's limitations with numeric or highly technical/science-focused content.

  4. Tone Selection

    • Choose an appropriate tone (Basic, Instructional, or Coaching) that matches the student's needs and developmental goals.

    • Experiment with tones to balance clarity, motivation, and actionable guidance.

  5. Educator Training 

    • Provide educators with training on how to effectively input data and interpret AI-generated outputs.

    • Share best practices for integrating AI feedback into teaching strategies.

Supported Languages

In general, AI Suggested Feedback] supports the following languages output: EN-US. 

For other locals in english are used as input, e.g EN-UK, or EN-AU, the system will output in EN-US  

The system hasn't been specifically evaluated for the wide variety of dialects and sociolects represented in English.  

Generative AI models are trained using vast amounts of data, and that data is most often in English. This can sometimes result in better performance in English as compared to non-English languages. As with any deployment of generative AI models, we encourage users to be mindful of the limitations of these systems for their specific use case, and cultural and linguistic contexts.    

Microsoft is planning to add more supported languages and locals to AI Suggested Feedback. This article will be updated as additional languages are supported.  

System Evaluation

AI suggested Feedback was evaluated through extensive manual and automated testing. More evaluation was performed over custom datasets for offensive and malicious prompts (user questions) and responses. In addition, AI suggested Feedback is continuously evaluated with user feedback. 

The system hasn't been specifically evaluated for the wide variety of dialects and sociolects represented in English.  

What should I do if I see unexpected or offensive output when using AI Suggested Feedback?

If you see offensive outputs, please submit feedback in Assignments using the thumbs up/thumbs down feedback in the Grading UI so that we can improve our safeguards. Microsoft takes this challenge very seriously, and we are committed to addressing it. 

How can AI Suggested feedback be disabled?

AI-Suggested Feedback is available only to educators with a Microsoft 365 Copilot license. The feature is automatically limited to these users and is not accessible to others without the required license. To learn more how to acquire a Microsoft 365 Copilot license visit Microsoft 365 AI Education Tools | Microsoft Education.

Learn more

AI feedback for educators

Need more help?

Want more options?

Explore subscription benefits, browse training courses, learn how to secure your device, and more.