Social and Environmental Impact of Machine Learning Competitions

Machine learning competitions are used to develop and compare algorithms, typically with the aim of achieving optimal performance in a specific task. However, questions also arise about how these competitions are organised, what motivates people to participate, and what happens beyond the final ranking.

In this project, students can explore various aspects of machine learning competitions according to their interests and backgrounds. Possible directions include:

• Environmental impact: How many submissions are made during a competition, and what can we learn about the computational resources involved? Students could analyse publicly available competition data to explore how competitions’ environmental impact might be assessed.
• Competition design: How do competitions differ in terms of their rules, evaluation metrics, leaderboards and incentives? Students could compare existing competitions and examine how different design choices might affect participation and collaboration.
• User experience: What motivates people to take part in ML competitions, and what might discourage them? Students could investigate how competitions are perceived by students or newcomers, what they expect from them, and whether they would consider participating.

Depending on the chosen topic, the project could involve analysing existing competition data, comparing competition formats, reviewing literature, or conducting a small user study.

The specific research question and approach will be developed together with the supervisor.

Groups of 2+ students (from any study program, mixed groups are welcome).

Supervisor: Sophie (daily supervisor), Veronika (co-supervisor/co-examiner)