Interoception as modeling, allostasis as control

all right welcome to act inf livestream number 50.0 it is october 17th 2022 and we are starting the series on interoception as modeling allostasis as control welcome to the active inference institute we're a participatory online institute that is communicating learning and practicing applied active inference this is recorded in an archive live stream so please provide us with feedback so we can improve our work all backgrounds and perspectives are welcome and we'll be following good video etiquette for live streams today in 50.0 we're starting this series discussing the paper interoception as modeling allostasis as control the authors are eli sinesh jordan thuryalt dana brooks john wilhelm van de mint lisa feldman barrett and karen s quigley it is a paper published in the journal biological psychology in january of 2022 and this is just an introduction and context video as dean here and i work through this paper and discuss some related topics we're going to go over some big questions basic aims and abstract roadmap of the paper then go into the paper touch on a few key points and images and themes and then primarily look forward to discussing it with any authors or anyone else who'd like to join in the coming weeks all right so i'm daniel i'm a researcher in california and the paper was selected by decentralized upvote it was uh fun to read and it brings some new topics into our regime of attention so i'm looking forward to this and i'll pass to dean thanks tally i'm dean i'm here in calgary i think what's interesting about any paper that um pulls two seemingly discrete things and puts them proximal and gets them entangling um and then pulls them apart again and tries to look at sort of what happens when things get combined is always something that i'm always interested in that in fact it was the title of this and the fact that they were bringing allostasis and interoception together and what what does that mean was the reason why i spent a little bit of time on this and really am quite interested in hearing what the authors have to say now that they put this idea to to uh work all right how would you describe the big question well i mean i think the focus here is on on what happens and what we can say about control requirements necessary to carry out physiological tasks varying degrees of complexity um i'm i've never been really sure how the the specificity of different outcomes and behaviors looks in aggregate but i think one of the things that we can say is if there's a predictive component and there's a feedback component and they're focusing on two particular types of that um it's interesting perhaps what kind of generalities we can walk away with um never you know sort of not getting lost in the weeds of well this happened here and then that happened there but just on that general level what what sorts of things when those two things are put together um should we say we say has a certain pattern or commonality nice i agree i would put it in the category of asking how a given system independent or modality independent phenomena like allostasis which is anticipatory changes in physiology connects with a domain or a use case like interoception and it does seem like a lot of the models that we've explored have been visual or conceptual or symbolic and the interoceptive modality has popped up again and again in discussions and so this is a cool opportunity to look at some formalisms for interoception interoception okay here's the paper uh as described earlier just a few of the highlights that were written by the authors on the publication page interoception enables the brain to anticipate the body's upcoming metabolic needs interoception provides performance metrics for viscera motor regulation we connect control theory with decision making and motor control of the body we model allostatic control as optimal control with the time varying reference signal they talk about some limitations of previous models and some unique…