And the more I talk to you, and think through your work, the more i find myself pulled toward a story that's far more complicated than most sketches of computational explanations!
-
-
Replying to @NeuroYogacara @evantthompson and
Hmm... Playing devil’s advocate: I doubt that we can ever achieve a complete and accurate mapping as
@evantthompson suggest, and, if we do, I am not so sure that it would constitute a “true” explanation, perhaps because I ignore what the latter means. >2 replies 0 retweets 3 likes -
Replying to @twitemp1 @PsychScientists and
Fair points. But I didn't mean "true explanation;" I meant something that's truly an explanation. I don't think the mapping needs to be complete, but it does need to specify what the neural code actually is; otherwise it's "as-if" talk that hasn't been sufficiently grounded
1 reply 0 retweets 4 likes -
Replying to @evantthompson @PsychScientists and
Well, I disagree. I'm happy assuming different explanatory levels. To use a silly simile, we do not need to appeal to subatomic particles to account for the dynamics of a pendulum.
1 reply 0 retweets 3 likes -
Replying to @twitemp1 @PsychScientists and
I'm not arguing against distinct explanatory levels. I'm saying computation is a heuristic without a specification of how its implemented.
1 reply 0 retweets 4 likes -
Replying to @evantthompson @PsychScientists and
Would you consider that a model such as Rescorla-Wagner requires specifying neural implementation?
1 reply 0 retweets 3 likes -
Replying to @twitemp1 @PsychScientists and
That depends on what question we're asking, but my understanding is that there are data relating neural firing patterns to the R-W model, in which case the model has an explanatory grip.
1 reply 0 retweets 4 likes -
Replying to @evantthompson @PsychScientists and
There is neural data relating to TD (a real-time instantiation of RW), but the explanatory scope of the model is, IMO, independent of it. It rests in its predictive power. >
1 reply 0 retweets 2 likes -
Replying to @twitemp1 @evantthompson and
Finding a correlation between a model and brain activity further strengthens it but it does not alter its value as a behavioural model.
2 replies 0 retweets 2 likes -
Replying to @twitemp1 @evantthompson and
But then the question becomes what sort of thing do you want to predict? I could make a experiment proving that mice embody a rule that says “if hungry cat, then run”. Model makes accurate predictions. But would be silly. So what is the grainsize / how deep to go?
1 reply 0 retweets 0 likes
Here's an example to voice my concern: "Bayesian brain" suggests the brain is really a Bayseian machine. That's a stronger claim from saying it can be usefully modeled using Bayesian tools. I've no issue with the latter; the former needs a lot more support.
-
-
Replying to @evantthompson @theblub and
I do agree! I am not saying that a cognitive/behavioural model equals to any of their possible physical instantiations. Not in the least! In fact I am claiming that its value is independent of it.
1 reply 0 retweets 5 likes -
Replying to @twitemp1 @evantthompson and
Yes. This. And what’s super interesting is that the difficulties faced by linking dopamine transients to TD don’t undercut the value of the updating model. Perhaps *both* because R-W is insufficient, and it’s not implemented by tonic+physic spiking in the basal ganglia!
1 reply 0 retweets 1 like - 5 more replies
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.