Mathematical Biology Seminar - Alex Stewart

Alex Stewart; Luddy School of Informatics, Computing, and Engineering; Indiana University Bloomington

Can evolutionary models help us build better social algorithms?

Evolutionary models have a long history as mathematically and computationally sophisticated tools for understanding how populations change over time. Their insights have not been limited to biological evolution and genetics. Indeed, the concept of a "meme" – a transmissible idea subject to the pressures of natural selection – reflects the impact of evolutionary ideas on our understanding of how culture develops. It is therefore natural that evolutionary models have been widely applied in recent years to pressing social questions, including the spread of misinformation on social media, the risks associated with AI and the origins of political polarization. Unfortunately, empirical tests of such models are often difficult, since they require access to a ground truth that is hard to measure. I will first discuss efforts to connect evolutionary models of social media to data. I will then discuss the potential of such models to act as tools for "stress testing" social algorithms, such as crowdsourced community moderation (e.g., X's Community Notes) and the output of LLMs.