Computational Meta-Theory in Cognitive Science
: A Theoretical Computer Science Framework

  • Federico Adolfi

Student thesis: Doctoral ThesisDoctor of Philosophy (PhD)

Abstract

Cognitive science is a multidisciplinary effort to create knowledge about cognitive systems. As such, it is scoped and limited (and potentially compromised) by assumptions held by cognitive scientists about knowledge-creating processes. In particular, assumptions about the computational properties of the problems researchers tackle are hugely consequential. What complicates answering the scientific questions we care about? Which questions should we tackle first? Are we asking feasible questions? Intuitive reasoning about general and specific answers to these issues is severely limited and often misleading. The lack of conceptual scaffolding and embedded tools to reason in mathematically and computationally rigorous ways means that these (often implicit) meta-theoretical beliefs are seldom examined formally. Without these safeguards in place, there is a greater likelihood of being misled, getting stuck, wasting time and resources, and missing opportunities to improve our knowledge of cognitive systems. To address this, the present work articulates a computational meta-theoretic framework and illustrates its application. This framework for inquiry builds on an existing dialogue between the theoretical computer sciences and the cognitive sciences and involves a currently developing methodology that has so far eluded explicit formulation. To begin, the principled need for computational meta-theory is motivated by drawing attention to a recurring revival of problematic yet unexamined assumptions about how cognitive explanations can or should be discovered. Practical necessity is argued via a case study of how empirical research benefits from the availability of such a higher-level formal framework. Then, the methodology is articulated in terms of its conceptual and formal foundations; these are grounded in cognitive science meta-theory and theoretical computer science, respectively. The resulting computational meta-theoretic machinery is applied to long-standing questions and illustrated through case studies spanning various domains and disciplines. Finally, a synthesis of this work and other recent efforts deploying the methodology is discussed together with their findings, implications, and the prospects for a computational meta-theoretic framework for cognitive science.
Date of Award23 Jan 2024
Original languageEnglish
Awarding Institution
  • University of Bristol
SponsorsErnst Strüngmann Institute for Neuroscience in Coop. with Max Planck Society
SupervisorDavid Poeppel (Supervisor) & Jeffrey S Bowers (Supervisor)

Keywords

  • Cognitive Science
  • Theoretical Computer Science
  • Computational Complexity
  • Computational Modeling
  • Meta-theory
  • Formal theory
  • Artificial Intelligence
  • Computer Science
  • Philosophy
  • Psychology
  • Neuroscience

Cite this

'