Bernard Koch explores the mechanisms underlying cultural diversification and collapse in fields like AI, Music, and Science. His work blends AI with Bayesian statistics and 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, Koch 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 dynamics in spaces like Music and news cycles. Ongoing work deploys this approach to explain the birth and death of tens of thousands of metal bands and subgenres over more than fifty years.

Bernard has published in diverse venues across Sociology (Sociological Methods and Research, Sociological Methodology), Computer Science (NeurIPS, WWW), Bioethics (Hastings Center Review), and general interest (Science). His work has also been featured by TIME Magazine, Venture Beat, and Mozilla.