Simulator for oral exams: AI-supported avatars offer protected examination situations [DiL]

Oral examinations of all kinds present students with particular challenges. The combination of subject-specific knowledge, subject-specific skills, situational reactions, appropriate communication and a rarely practised stressful situation requires intensive preparation. Students often have limited opportunities to practise realistic exam situations and rarely receive detailed feedback on their performance.
To address these challenges, we are developing a novel simulator for oral exams that is subsequently able to provide extensive feedback to examinees. This combines advanced technologies: 

1. Digital avatar: An interactive, digital avatar simulates a realistic examiner.
   
2. AI-supported exam model: A large language model, fed with an extensive knowledge database and exam questions, enables authentic and adaptive exam discussions.
   
3. Rhetoric analysis: The ‘NOVA’ software is able to analyse the speaking behaviour and body movements of examinees (eye contact, hand movements, manner of speaking, pitch, volume, etc.).
   
4. AI-supported content analysis: Transcription and content analysis of the answers analysed and feedback by AI tools 
   
5. Data-based feedback: Detailed, objective feedback on technical and communicative aspects of the exam performance.
   

The simulator is available during ongoing operation and is constantly being further developed. 
This seminar offers the opportunity to work at the interface of university didactics and artificial intelligence and to actively shape the future of exam preparation.
Bayreuth lecturers can now use the exam simulator with their students free of charge.

Learning objectives of the seminar:
After completing the seminar, participants will be able to 

1. Discuss and critically reflect on the potentials and limitations of AI-supported exam simulations.
   
2. Evaluate the use of behaviour analysis tools in examination situations from a didactic and ethical perspective.
   
3. Develop concepts for integrating the simulator into existing courses and exam preparation programmes.
   
4. Identify and discuss possibilities for further development and adaptation of the system for different subject areas.
   

Software:

• ChatGPT 4o via API: https://openai.com/api/
  
• Interactive avatar: https://labs.heygen.com/interactive-avatar
  
• NOVA: https://github.com/hcmlab/nova
  

Your speaker:
Paul Dölle is a research associate at the Centre for Higher Education Teaching (ZHL) at the University of Bayreuth. His expertise in the field of digital teaching innovations, which he has developed since 2014, is reflected in his management of the university didactics seminar programme. He developed the ExamSim project, an AI-based simulator for oral examination situations that enables realistic examination scenarios using a digital avatar. He was also significantly involved in the application process for the QUADIS (funding volume 4.989 million euros) and XRCampus projects.
In his seminars, he teaches the basics of university didactics, modern teaching methods such as flipped classrooms, AI-supported tutoring systems, interactivity in large groups and the creation and optimisation of teaching and learning videos. He attaches particular importance to a goal- and practice-orientated way of working in an interactive and positive atmosphere. Another focus is on the qualification of tutors and e-tutors.
More about Paul Dölle: https://www.zhl.uni-bayreuth.de/de/ueberuns/paul-doelle/index.php