VUB researchers call for greater epistemic diversity in artificial intelligence

Brussels, 16 January 2026 - Artificial intelligence is often presented as a rapid succession of technological breakthroughs. VUB professors and researchers Luc Steels, Ann Dooms, and Remi van Trijp fundamentally challenge this dominant narrative. In their book History of Ideas in the Science of AI, they portray AI not as a linear progression, but as a dynamic and living network of ideas that continuously evolves through cross-fertilization, shifts in meaning, and unexpected combinations. Central to their argument is the concept of epistemic diversity: the importance of allowing different research traditions, theories, and methods to coexist in order to foster innovation and resilience in AI research. According to the authors, this diversity is currently at risk due to the strong focus on dominant paradigms such as generative AI.

AI as an ecosystem of ideas

Drawing on insights from evolutionary biology and complexity theory, Luc Steels develops an innovative analytical framework for understanding the development of scientific knowledge. He describes AI as an epistemic ecosystem in which ideas emerge, mutate, disappear, and sometimes regain relevance decades later. This historical layering shows that even seemingly abandoned lines of thought may contain valuable building blocks for future research.

Mathematics, language, and human insight

In her contribution, Ann Dooms explores the interaction between mathematics and computational systems. She first looks back at the historical origins of the computer in research on the foundations of mathematics. Today, in turn, computational models assist mathematicians in identifying connections within the vast web of proven results. Remi van Trijp approaches language as a complex adaptive system and examines how meaning emerges, changes, and is shared-an approach that contrasts sharply with dominant statistical language models, which primarily capture the current state of affairs.

An invitation to rethink

Despite their different perspectives, the authors share a core insight: the greatest challenge for AI does not lie in replicating existing human knowledge, but in understanding how humans develop, adapt, and transmit new insights.

History of Ideas in the Science of AI is deliberately not an encyclopedic overview, but an invitation to reflection and to rethinking the history and future of AI. The book is aimed at researchers, students, and policymakers who wish to understand how AI has become what it is today, and which lines of thought from the past may offer new perspectives for the future of the field.

Info

The book was developed within the framework of the Willy Calewaert Chair of deMens.nu, awarded to VUB emeritus professor and AI pioneer Luc Steels at the Vrije Universiteit Brussel. It is published by VUBPress and is available in print from the authors and digitally on Zenodo and Apple Books.

Contact:

• [email protected]
• 0479/56 35 30

Read more