Textbook Group meeting for Parr, Pezzulo, Friston 2022 .

all right well thank you for that very interesting prior discussion on on latter texbook group so I hope that we can continue from that first few minutes that we had okay this is the last recorded discussion feel free to come next week for um the other cohorts also unrecorded discussion again we'll kind of like look over what has been written think about what we want to do with textbook group and everything kind of continuing so but here's chapter 5 and or like anything on the first half of the book like anyone could go any part from five or first half or we could look at the questions and look at which ones or spend a minute and upvote questions that we're interested in and then look at which ones have been upvoted a lot but not answered for first time sure question sound good okay in chapter five anyone can just people can be clicking the thumb and then we'll just go in these or raise your hand and like go for anything else but anything in chapter five that anybody wants to go to first um table 5.1 with the neurotransmitters um I think it's page 98 yep that's the one um so is what being stated here pretty much saying that you dopamine is increasing the Precision of you know the the brains uh you know I guess what the bra understands of of policies so it's saying you these policies are more precise if you know the uh a newon or a group of new on is getting more to mean yeah so to kind of like pull back on the table this is going to be linking specific molecules and roles they play in mechanisms in the brain to specific variables in generative models and then the table is like describing well here's some sheer anatomical facts these are empirical measurements and then also describing and citing some studies that might have studied um the relationship between like the activity level or the amount or the signal of some neurotransmitter to some aspect of a generative model and then that could be more correlative evidence like size of a brain region or um changes in the functional fmri um okay or it can be more causitive with like CH giving a manipulation and then showing that the two things are changing together so it doesn't mean it's the only function it does or its whole role or anything but it's just saying these parameters are being being identified are not just like weird statistical artifacts they map on to like natural components of behavior so it's proposing there some connection between your know dopamine and the policies that the brain is building there's to to really look at this in the dopamine case thankfully which is actually cool because it's most most things like the kind of case for pluralism has not always been written out this is okay we'll just go in here but in this this paper they specifically are looking at the meeso cortic olymic dopamine so this is something like what is shown in the um figures in chapter 5 with dopamine controlling the balance between the habit-based policy selection and the expected free energy based policy selection so that's kind of that's the computational function that is being posited in the model that's in the chapter and it's what it's being identified on for um in the table and then this paper's like philosophers biology and saying well people have modeled dopamine in terms of the honia the salience and the reward prediction error basically free energy and they say well there's already evidence to support pluralism for that system but you could make probably analogous Arguments for like every brain region every Gene Etc but that's the whole thing with the maps in the territories it's like okay it's a subway map of this city but it's not an everything map okay okay any other random part of five or one of these questions someone wants to think about sure just to uh to follow up on that last uh question and point really quickly um I think there's a good example this is in an older paper I think it's called um dopamine reward learning and active inference and um it looks at a lot…