Connecting levels of analysis in the computational era

TL;DR

This paper discusses the evolution of levels of analysis in neuroscience and AI, emphasizing the shift from three to four levels due to advances in simulation, data, and algorithms, and highlights their interdisciplinary interactions.

Contribution

It introduces a four-level framework for analyzing neural and AI systems, expanding on Marr's original three-level approach to incorporate modern computational methods.

Findings

Four levels of analysis: observations, models, algorithms, functions

Bidirectional interactions influence interdisciplinary research

Convergence of data and simulation advances the framework

Abstract

Neuroscience and artificial intelligence are closely intertwined, but so are the physics of dynamical system, philosophy and psychology. Each of these fields try in their own way to relate observations at the level of molecules, synapses, neurons or behavior, to a function. An influential conceptual approach to this end was popularized by David Marr, which focused on the interaction between three theoretical 'levels of analysis'. With the convergence of simulation-based approaches, algorithm-oriented Neuro-AI and high-throughput data, we currently see much research organized around four levels of analysis: observations, models, algorithms and functions. Bidirectional interaction between these levels influences how we undertake interdisciplinary science.