Thanks Ben!

Ezra Klein show?

Thanks Kameron!

Also thank you, I’m really glad you enjoyed it!

It didn’t make it into the preprint but dopamine neurons respond to D2 stim as a sign-flipped TD error, which fits nicely in that framework

Great question… my current thinking based on plasticity rules is that they learn pessimistically biased sign-flipped value from dopamine dips, as in @sromeropinto.bsky.social’s and Adam Lowet’s papers

Thanks Alex!

Thank you!

Thank you Laura!

Thanks Adrien! Yes, that’s exactly right!

Thanks Eddy!

Thank you Sam! It was super fun to present in your lab!

Thanks so much to all co-authors, especially my mentor Naoshige Uchida @naoshigeuchida.bsky.social! It has been a joy to work on this (ongoing) project!

I’m excited about this work because it shows how the microcircuit configuration of the dopamine system could control the degree of preference for current versus future rewards (can you resist that marshmallow for two marshmallows later?). Future work is headed in that direction!

This suggests the exciting possibility that the time horizon of dopaminergic learning (its temporal discount factor) is set by the balance of excitation and inhibition in this pathway.

We found dopamine neurons are hardwired to automatically take the temporal derivative of their input from striatal neurons—thus accomplishing a key step in TD learning.

Dopamine neurons famously signal temporal difference (TD) errors, a teaching signal for learning to predict rewards, but the mechanisms that produce these signals are unknown. We also don’t know how the parameters governing dopaminergic learning arise from biological components.