Bernard Koch explores the mechanisms underlying cultural diversification and collapse in fields like AI, Music, and Science. His work blends computational methods with Bayesian statistics and qualitative interviews to develop evolutionary theories and rich historical narratives.

Current research focuses on how different evaluation systems (e.g., peer review, benchmarking) guide scientific fields along different trajectories, with significant ethical, epistemic, and cultural consequences. Through historical case studies like AI's convergence on deep learning and social psychology's long entanglement with controversial racial hereditarian research, he illuminates the strengths and weaknesses of different evaluation systems. 

Beyond evaluation, he is generally interested in mechanisms that drive the evolution of cultural ideas. Ongoing work develops theories and Bayesian models to explain such dynamics across spaces like music genres and news cycles. This framework has been applied to explain how heavy metal evolved through the birth and death of more than 30,000 bands and the subgenres they explore over fifty years. He has published in Sociology (Sociological Methods and Research, Sociological Methodology), Computer Science (NeurIPS, WWW), Bioethics (Hastings Center Review), and general interest venues (Science). His work has also been featured by TIME, Venture Beat, and Mozilla.