In collaboration with
As part of a NOLAI co-creation project, Eindhoven University of Technology, Het Stedelijk Enschede, and TA-Help.me have joined forces to develop a responsible AI assistant that helps teachers clarify students’ questions. With the help of our AI assistant, students think more deeply about their questions and arrive at insights, solutions, or a more focused question. The AI will never try to answer the question, which allows the teacher to provide more targeted and efficient guidance.
When a lesson relies heavily on independent work and question-driven instruction, the number of questions from students often increases rapidly. This can mean that the teacher does not always have time to answer every question. Next to this, the teacher spends a lot of time trying to identify the core of the problem, leaving little time to actually help the student.
When a student asks a question, the AI assistant tries to help the student clarify it without answering it. The AI assistant evaluates the question based on five different categories that define a good question based on research from the University of Eindhoven. If the question falls short in one of the categories, the AI asks a follow-up question to clarify that category. Once the questions scores well on all the categories, it is sent to the teacher.
| Category | Score |
|---|---|
| Content | |
| Expected outcome | |
| Steps taken | |
| Problem analysis | |
| Terminology |
The AI is developed in a co-creation project with researchers from the University of Eindhoven and an experienced teacher from Stedelijk Enschede. By combining research with first-hand insights and needs from the education sector we were able to develop an AI tool that actually matches the needs and wants from the educational sector. This makes for a tool that is both pedagogically sound and tested in a real educational setting. Since the tool is designed to support the student in getting to the core of their questions, control always remains at the user. If a student disagrees with the AI and believes the question is good enough, they can always bypass the AI and ask the teacher the question directly. Likewise the teacher can reject the question clarified by the AI.
The prototype was extensively tested and validated in a real classroom with students. Over a period of several weeks, students used the AI prototype during class. All the questions they asked had to be refined first with the AI assistant before they were forwarded to the teacher.
The research show that the questions became better and more complete through the use of the AI prototype. This suggests that the system serves as a tool for improving questions. The instructor received 52.9% more complete questions.
More about the project
The prototype is fully integrated into TA-Help.me. The refined questions are placed directly in the queue, so the teacher can clearly see who is next in line and what that student’s question is. This allows the teacher to be well-prepared to answer the question or to see that other students have the same questions. After class, the teacher can gain insight into which topics generated the most questions.