The Question
The rapid advancement of artificial intelligence is reshaping nearly every aspect of modern life, from scientific research to governance. However, beneath this explosive growth lies a troubling trend: AI research may be converging toward a “scientific monoculture.” The field is increasingly dominated by a few key technologies like transformers and large language models, while abandoning a greater diversity of conceptual and methodological approaches. This potential loss carries serious risks — reduced innovation, diminished resilience, and vulnerability to technological dead-ends. Yet, without quantitative tools to track how AI’s intellectual landscape has evolved over its 75-year history, we lack the empirical evidence needed to determine if this narrowing is real or perceived. This gap in understanding is critical: As AI becomes deeply embedded in society, with profound implications for its societal impact, we urgently need rigorous, data-driven insights to guide future research and ensure the field remains innovative and epistemically resilient.
The Approach
This project — Mapping the Evolution of AI: A Data-Driven Exploration of Epistemic Diversity and Future Frontiers — aims to create an unprecedented, data-driven “map” of AI’s 75-year history through a comprehensive analysis of publications, conference papers, citations, patents, funding records and other documents from the largely informal literature at its beginnings. The project is constructing a longitudinal dataset using sources like OpenAlex and Dimensions; developing computational methods to visualize the field’s evolving structure; tracking how AI subfields emerge, branch and converge; and defining quantitative metrics for epistemic diversity — the variety of research approaches within AI.
The methodology combines cutting-edge techniques: Natural language processing models will generate embeddings capturing conceptual similarities across decades of research, and complex network analysis will map relationships between publications, authors and institutions. Conceptually, the research team treats the history of AI as a living epistemic ecosystem. Phylogenetic networks will reveal how different concepts pollinated across subfields and how shifts in meaning, unlikely combinations and “roads not taken” have continually expanded AI to where it is today.
This quantitative history will reveal whether AI research is indeed narrowing into a dangerous monoculture. By illuminating hidden biases and identifying neglected but promising subfields, the history of AI becomes a living resource rather than a closed chapter. Finally, the research team will produce interactive visualizations to engage researchers, policymakers and the public. These tools will foster strategic dialogue, framing AI not as a story of inevitable progress, but as a complex living web of research threads that must be understood to shape the future.
The Mapping the Evolution of AI project is being led by Stefan Thurner and Helga Nowotny at the Complexity Science Hub Vienna, Austria, in collaboration with Vittorio Loreto and Luc Steels at the Sony Computer Science Laboratories in Rome, Italy, and Paris, France.
