Dr. jonny phd on Nostr: npub1skvad…laky3 my nonexpert take: I tend to think terms have a plurality of ...

npub1skvad2l2wrxgdmt6yxk9kt2rjhw5tucjzhf54pktfq2gg0qhgwyqdlaky3 (npub1skv…aky3) my nonexpert take: I tend to think terms have a plurality of meaning, and definitions only take hold in some context as boundaries against neighboring concepts. So mechanism has many meanings that are true across scales and modalities. When thinking about neural computation in a systems/circuits context, I would tend to call the mechanism the particular configuration of intra and intercellular dynamics in a circuit that give rise to an algorithm - how a circuit is doing something vs. What it is doing. I dont think causality has a unidirectional arrow there, which is often where I feel like hunts for causality get bogged down. The circuit dynamics are the way they are so they can do the algorithm, and the algorithm is the way it is because of the circuit dynamics: the causal hierarchy is tangled.

Its useful to be clear about what a model can and cant do, but i dont think its all that useful to spend time to try and divide models along categorical lines as "mechanistic" or "non-mechanistic" because a) its only meaningful in context, and b) it abstracts away from questions that directly characterize the model - does this model give additional predictions about network behavior on perturbation? Under which simplifying conditions does it hold? Etc. In other words, I usually think definition wars are a sign we've lost track of what we actually want to know.