Textbook Group meeting for Parr, Pezzulo, Friston 2022 .

all right thank you everyone today is February 5th 2024 and Oli will be facilitating this discussion on chapter one for cohort 6 so thanks Oli okay uh hello everyone thanks for joining for our first discussion for chapter one uh well um as you might have read in in the content and uh also for the introduction in chapter one uh this is sorry this chapter gives a kind of um overall picture of uh what kind of theory active inference is uh and specifically uh what kind of approach uh the textbook is going to take explaining different aspects of um active inference Theory uh well one of the main themes of uh this chapter and U also the overall organizational scheme of the textbook is uh the difference between two kind of approach to active inference literature namely uh the high road and the low road so basically uh for high road uh the prior assumption is the existence of quote unquote things and then explaining from that um seemingly obvious assumption into what uh what requirements that quote unquote thing uh should have in order for it to have U many various uh aspects such as agency uh as a kind of decision maker and so on so basically uh it's not something specific to cognitive agents but uh the whole framework can also be applied to many other uh uh many other agents uh as long as it can be described in terms of its internal and external Dynamics so uh that's one approach to active inference the other one uh which uh is usually called Low Road in the literature is beginning from uh some of the concepts developed uh from back in the 19th century uh for example by people such as Helm holes uh in which they kind of they it tries to generalize uh the basian reasoning uh into um unto un developing uh the notion of free energy and specifically variation free energy uh for which we can um we can formulate uh the way that that the organisms or agents can perceive the world and also uh the related variable or related so-called parameter uh so for uh for perception we need a kind of lower boundary which is the AAL free energy and uh for um for control or for decision making uh we need uh another uh variable or parameter namely expected free energy so by describing these two parameters as kind of lower bound of certainty we can uh again formulate how uh organisms or cognitive agents can engage in both perception and also action so basically when we say active inference it's uh it implies a kind of integration of these two um these two aspects of perception and action so they're not uh essentially distinct from each other uh both of them uh both of them kind of um follow a similar trajectory or similar mathem iCal uh technology but uh in somewhat mirror or uh reversed way so um I hope I uh I could give a kind of overall meaning of what high road and low road uh mean but obviously we'll uh we'll get to see much more detail of both of these approaches so the uh the chapter two deals with um the introdu uction of low road approach and then beginning from chapter 3 we'll see how high road approach differs from uh the low road one so um if anyone has any thought or comments or question on these two concepts or we can go to the question section and see uh and discuss some of the questions positive there okay so maybe we can open the question section or by the way yeah no on that question um so in the in the chapter it talks about the fact that organisms or just agents can not the motivation is to reduce surprise but they may not able to act on surprise directly but instead on what's called proxies could you and they see that's a variation the free end variation free energy could you clarify what we mean by that or just giv some examples of that sure so yeah the concept of surprise uh actually is a little bit different than uh the psychological conception of surprise uh that we're familiar with in everyday parlament but uh basically what we mean inactive inference by surprise is its uh statistic in its a statistical sense so uh it's surprisal…