This thesis builds on recent research of AI literacy. While AI is integrated in many forms and shapes into every day life, many people are not aware of its basic limitations and functioning. This master theses aims at educating people in a playful way about AI's data dependency by letting them experience concepts such as data bias and short cut learning in an image recognition task. 

As part of your thesis, you can turn AI bias into a playful learning experience. The idea is to use data of a citizen science project iNaturalist. There you can get nature images of suitable licences to use for model training. The focus on nature is set intentionally due to multiple reasons: People are on average not very knowledgable about identifying plants, but with a bit of help they may develop a heuristic quickly, and thus can be educated to identify AI errors. Further, whether this interaction can foster nature relatedness can be evaluated additionally.

The game flow could, for example, be as follows:

1. Train AI on summer-only data of oak and beech tree images with some oak images from autum. The AI works… at first.
2. Winter images confuse it, as it seemed assume that brown leaves can only be of an oak tree, and reveals the bias (e.g., color shortcut). 
3. Let user collect new winter photos in the game (Drag&Drop).
4. Let them retrain the model with new data and explore performance changes.
5. Reflect: What happened? What did I learn? How is AI different from me?

One inpiring game example is https://research.google/ai-quests. The focus of this work, however, should lie specifically on data influence on the AI prediction. The game needs to ensure that users understand what is happening and why.

If you choose to work on this thesis, your objective would be to conceptualize, develop and evaluate such a gamified learning experience. You do not have to have extensive prior knowledge in machine learning / computer vision, as simple models are sufficient to showcase this effect (see google's teachablemachine format).

Procedure

To reach this goal, the following steps can be taken:

1. Get familiar with the idea,
2. Read literature to understand AI literacy and data bias in AI, and gamification,
3. Implement Prototype (back-end, front-end),
4. Conduct interviews with at least 5 individuals to evaluate the updated prototype (let them play, ask questions afterwards),
5. Analyse the interviews using a content analysis with MAXQDA,

As this is an explorative thesis, your own ideas are welcome!

References

• Pinski, M., & Benlian, A. (2024). AI literacy for users–A comprehensive review and future research directions of learning methods, components, and effects. Computers in Human Behavior: Artificial Humans, 2(1), 100062.
• Tu, J., Smith, G. M., Bassanelli, S., Marconi, A., & Nacke, L. E. (2026, June). Conceptualizing How to Design for AI Literacy through Game Artifacts. In Proceedings of the 2026 Designing Interactive Systems Conference (pp. 1056-1080).

• Liu, H., Lai, V., & Tan, C. (2021). Understanding the effect of out-of-distribution examples and interactive explanations on human-ai decision making. _Proceedings of the ACM on Human-Computer Interaction_, _5_(CSCW2), 1-45.
• Rosella Gennari, Alessandra Melonio, Maria Angela Pellegrino, and Mauro D'Angelo. 2023. How to Playfully Teach AI to Young Learners: a Systematic Literature Review. In Proceedings of the 15th Biannual Conference of the Italian SIGCHI Chapter (CHItaly '23). Association for Computing Machinery, New York, NY, USA, Article 1, 1–9. https://doi.org/10.1145/3605390.3605393