Source-linked AI summary
Quantifying the role of neurons for behavior is a mediation question
Ilenna Simone Jones, Konrad Paul Kording
TL;DR
The paper asks how neurons mediate the influence of stimuli on behavior and why typical recordings make that question difficult to answer. It examines how mediation assumptions shape neuroscience language and argues for stronger experiments and testable theory.
Problem
The paper asks whether recorded neural activity mediates stimulus effects on behavior, but typical experiments record only a tiny proportion of neurons, making mediation analysis vulnerable to confounding.
Source: pp. 1–2Method
The paper contrasts instrumental-variable reasoning for recording-only experiments with experiments testing correlation, necessity, sufficiency, and exclusion, and examines how language and theory shape mediation claims.
Source: pp. 1–3Results
Typical neuroscience experiments do not establish information flow or causal chains, while terms such as representation and encoding implicitly suggest mediation.
Source: p. 2Takeaways & Limitations
The field should pursue experiments closer to meaningful mediation claims and develop theories that make testable predictions.
Source: pp. 2–3Instrumental-variable reasoning requires that there be no unknown causal paths among the variables being analyzed.
Source: p. 2
Abstract
from arXiv · showhide
Many systems neuroscientists want to understand neurons in terms of mediation; we want to understand how neurons are involved in the causal chain from stimulus to behavior. Unfortunately, most tools are inappropriate for that while our language takes mediation for granted. Here we discuss the contrast between our conceptual drive towards mediation and the difficulty of obtaining meaningful evidence.