Responsible use of artificial intelligence in developing teaching and education

Principles of feedback guide all our work with student feedback. The aim of the principles is to support an interactive, impactful and streamlined feedback culture at Aalto. Student feedback is a valuable tool in which we want to invest. AI can be used as an aid in processing feedback, but it does not replace human processing of feedback. For example, AI can pick out topics, classify responses or look for connections, especially when handling a large amount of open responses. However, AI outputs must always be reviewed by a human.

Aalto AI Assistant is the university's AI tool that can be used as part of the analysis of student feedback. These guidelines provide instructions for the safe and responsible use of AI in feedback analysis. The following guidelines aim to assist you in the processing procedure.

Before collecting feedback 

1. Consider responsibilities

To ensure ethical conduct, please familiarise yourself beforehand with the guidelines on student feedback and generative AI if you are not already aware of them. Responsible data use is part ofAalto’s ethical principles. You are responsible for following the rules on storing, processing and sharing of the data you collect and analyse. Transparency in important: when you are planning to use AI as a part of your analysis processes, make preparations for sharing information with your respondents: you can do this during lectures, in the communications about the survey, or in the survey itself.  

2. Consider what you ask when formulating the survey

When planning to collect feedback, format the questions in a way that they do not lead respondents to disclose their identities or, for example, their health information. Do not ask anything unnecessary, that is, questions where the answers received will not be utilised. From the perspective of feedback impact, the purpose of collecting it must always be considered in advance. 

3. Instruct respondents in the survey communications

Students should be instructed to respond without providing idenfiable information or health information. You can do this in the lectures, any communication materials or in the survey. Make sure to inform respondents also about the purpose and methods of feedback analyses.

Sometimes the feedback form includes a reminder, for example, the course feedback form might read ‘Please do not include confidential or secret information, such as your own or others’ personal or health information.’ The guidelines on data classification provide more information on what kinds of data can be processed and stored in different systems and services. Ensure that your respondents are aware of how their answers will be used in the development of teaching and learning. Conducting research on student feedback always proceeds through the research permissions process before the research begins.

Handling feedback before AI-assisted analysis

4. Select the portion of the data for AI processing

Examine the data you have received. Do the respondents seem to have understood the questions as intended? Has anyone included information in their responses that could identify an individual or individuals? When using AI tools for processing student feedback, the responsibility for data security is emphasised. Decide which parts of the data analysis you need AI assistance with, and separate it from all other data. It is particularly important to separate all data that could identify the respondent from the data to be analysed. Lists of respondents and names occurring in responses are concrete examples of this. Also, remember to consider indirect identifiers, which when combined could identify the respondent. For example, in very small courses, a combination of major, year of starting studies, and accumulated study credits, if asked in the feedback, may be enough to identify the respondent.

Using AI in the feedback processing procedure does not replace humans, but serves as an aid, and it is important to recognise its limitations. Small amounts of feedback often do not require AI assistance.

Handling feedback with the help of AI

5. Process the data using Aalto AI Assistant

Once you have extracted the appropriate data from your original dataset for processing by AI, you can transfer it to Aalto AI Assistant. When using generative AI to analyse student feedback, always use Aalto AI Assistant, as it is the only AI tool permitted for handling student feedback at Aalto University. The use of Aalto AI Assistant ensures that the data or prompts fed into it are not used for training the AI model. You can use Aalto AI Assistant, for example, to theme responses, summarise them, or outline how topics that interest you appear in the responses. Always critically evaluate the AI's results before using them as part of your development work. Once your analysis is complete, always remember to delete the data and discussions you used from Aalto AI Assistant. You can find detailed instructions on this in Aalto AI Assistant's guidelines.

AI can be more beneficial when you want to search for and process feedback on a specific recurring issue. It might not be able to identify all necessary aspects when it comes to recognising less frequent accessibility problems or student concerns.

Examples of Aalto AI Assistant prompts to aid in student feedback analysis

• "You are developing a master's level course at the university. Extract the most frequently occurring suggestions for course improvements and positive feedback from the student feedback and create concise bullet point lists."
• "You are developing basic course content. Separate the topics in the student feedback that students found unclear. Assess which topics are generally mentioned together, and how often different topics are mentioned."
• "You are a teaching assistant working on educational development. Select relevant feedback on electronic study systems and analyse the themes related to them. Arrange them from the most common to the least common."
• "You are assisting a teacher in planning course arrangements for next year. Select feedback related to course arrangements, categorise them according to whether they concern physical spaces, lecture schedules or practical sessions, and describe the students' experiences."
• "Identify the key themes from the student feedback, such as the quality of teaching, student services, learning environments, and well-being. Also note new or unexpected perspectives. Create recommendation suggestions from the observations that help university learning services improve the student experience and better meet needs."

It might also be useful to try out the Aalto AI Assistant's Prompt book. It is worth experimenting with different prompts to find those that best suit your needs.

6. Process and share the results

Remember that using AI to handle feedback does not yet mean that feedback has been processed. The outputs of AI can help in reflecting on the feedback, but they must be reviewed, and further development actions need to be decided and implemented by a human. Peer discussions about feedback and developmental actions have proven to be very useful. It is important that the impact of feedback on teaching and education is also visible to the students. We therefore encourage you to share the results of your feedback analysis openly with your stakeholders, students and colleagues. However, when sharing your findings and results, ensure that your audience has the right to see the level of information you intend to share.

Read also: The new PowerBI report views for course feedback now published | Aalto University

Example cases

CASE 1. Tom Teacher 

• What kind of feedback data do I want to analyse? Tom teaches a large course with nearly 200 students and collects feedback from them at the end of the course. The feedback automatically includes Aalto and school-related questions, but Tom wants to add a few specific questions related to the course. When drafting these questions, he pays special attention to ensuring that they provide useful information for course development and do not encourage students to provide identifiable information about themselves. Tom has allocated ten minutes in the last session of the course for giving feedback and reminds students beforehand about the importance of giving feedback and guides them to answer without identifiable information. He also talks to his students about feedback analyses and the plan to test AI as a part of the processing.
• What is the purpose of my analysis? After Tom has collected the feedback, his aim with the analysis is to get an overall picture of the hundreds of open responses by asking the AI to pick out the most frequently recurring feedback topics and classify them under relevant main headings. For this purpose, Tom inputs the open question data from the feedback into the AI and instructs it to identify and classify recurring themes.
• Can I input the feedback data as it is into the AI? Before inputting the feedback data into the AI, Tom checks that no one has written names or other information that could allow individuals to be identified. He knows definitely not to provide the AI with the list of respondents.
• What tool can I use? Tom uses Aalto University's own secure Aalto AI Assistant to process the feedback, which is the only AI tool approved at Aalto for handling feedback.
• How do I process the results of the analysis? When Tom gets the AI-assisted analysis ready, he reviews it critically against the raw data and notes that the AI has oversimplified a few points. He corrects the results of the analysis accordingly and adds his own observations about the course and feedback that the AI did not notice. Now, Tom has a good overall picture of the feedback and it is easy for him to identify prioritised development areas and refine them into concrete actions. The remaining task is to put them into practice in the next course implementation.
• How do I communicate the results? Tom discusses the feedback with his fellow teachers. Additionally, he usually utilises the feedback response feature of the MyCourses course feedback tool, where he thanks students for their feedback, highlights a few key points, and openly tells students about the development areas and actions for the next implementation. This way, students know their feedback has an impact. Likewise, Tom summarises the main points of the feedback and informs new students at the beginning of the next course about the development areas and actions selected based on the feedback.

CASE 2. Leena LES expert  

• What kind of feedback data do I want to analyse? Leena wants to examine the national feedback from graduates to see what kind of feedback students have given about the student support services provided by the study services, comparing whether the services added this year have changed students' experiences compared to previous years. She checks the privacy statements of the feedback, ensuring her planned analysis is in line with them. She selects open response data from graduates' feedback from different years, choosing open-ended questions that encourage students to give feedback to the university about the services. She ensures that there is no identifiable information in the response fields.
• What is the purpose of my analysis? Leena's goal is to compare qualitative feedback across different graduation years and identify themes related to the services. She also wants to examine whether different themes have different tones of open feedback in various years to identify services students find beneficial and look for potential areas for improvement.
• Can I input the feedback data as it is into the AI? Before inputting the feedback data into the AI, Leena checks that no one has written names or other personal information in the feedback. She does not combine her selected open response fields with any data fields that could identify respondents.
• What tool can I use? Leena uses Aalto University's own secure Aalto AI Assistant because it is the only approved tool for handling feedback.
• How do I process the results of the analysis? When Leena allows the AI to process the feedback, she directs it to handle the feedback in the desired way: she sets precise limits for the AI and directs it to seek the desired information. She processes each year's feedback year by year and refines the results by asking further questions and clarifications. Once she has a good understanding of each year's feedback, she asks the AI to compare feedback from different years and then compares the AI's results with her own interpretation. She selects suitable themes for further processing and consults her colleagues about the results. In collaboration with other experts, the results are selected and a draft action plan is formed.
• How do I communicate the results? Leena and other experts share their results and discuss the drafted action plan more broadly within the community. With the help of community comments, further actions are selected and cooperation plans are formed. Based on new questions raised, the data is re-examined from emerging perspectives and refinements are made to the follow-up plans.