Conclusion (Discussion) — Course home

Hello, welcome everybody. It's November 1st, 2023, and we are in the closing discussion section for this first instance of active inference for the social sciences. It's been a journey. I'm sure we'll have many fun things to talk about and explore today. So, Avil, please take it from here and maybe give an overview of what we're going to do and talk about today. Avil, Okay, so this is the conclusion session of our course on active inference and for the social science. We have talked extensively about active inference, maybe less extensively about for the social science parts. So the goal of the session would be to assess what we talked about and how it is relevant or not to the practice of social sciences. Basically, this is a discussion session, so there is no clear plan, but I'd like to address how formally active inference accounts for agency and normativity and identity as a kind of integrated phenomenon, how it can compare or add to our grounds or clash with, I don't know, sociological and political model of agency and of social evolution and of etc. etc. All the topics that we addressed, which are numerous, and discuss about the plausible future for active inference, what it can be used to, maybe what it will be used to, certainly, how it may change and how it's relevant to the discussion. So, yes, that is the crux of it. We have brought in Alexei, who is a philosopher interested in social ontology, specifically the study of social kinds, which is from Kairos Research. And I do not know who Colin is, but he's with us. Hi, Colin. So how should we proceed, Daniel? What do we do? If you'd like, maybe pose a specific question or prompt, which could be preceded by a statement, and then we can all hear everyone's voice on that. Okay, so a good question would be, active inference describes agency and normativity and identity as a kind of integrated phenomenon. I think we should go over how precisely it works. So, yes, let's do that. I will find some, how do you say, illustration that will make it easier. But let's hear people's thoughts on that, please. Can you pose your question just a tad more clearly? How are agency, normativity, and identity addressed separately and together in active inference? Mal, go for it. So, I think we like to look at things in terms of interrelations. In active inference, nothing is defined by itself. It's defined in relation to other things. And this definition in relation to other things also defines the types of constraints that will be pushed on the thing which is separate from other things. Once you have this, you necessitate these interrelationships to be formative, ontological, and fundamentally contextual. So, nothing, no relationship is perfectly reproducible across a different context. It will take on a hermeneutics, a semantics, a different context, given the exact context of the spatial temporality of the entities. Those are some first thoughts. Does anybody have, like, big objections to what I just said? Okay, so let's start. Oh, yeah, please continue. Sorry. Let's start with the concept of agency, for instance. Agency is just action selection, right? So, agents are modeled. They have a generative model, and they use it to predict, like, sensory inputs based on possible actions. They try to select those to minimize the difference between predicted and sensory inputs. So, they're just trying to minimize free energy. And this minimization is fundamentally due to the relationship of constraints in their environment. So, let's imagine that I am an individual who... Oh, sorry. I saw a comment, and I didn't... So, yeah. So, you have an individual that tries to exert control over the environment. Now, if you scale this in nested structures, you can try to think of power structures and how agency within those power structures is modulated by the different kinds of constraints that are both embedded within the environment itself, within the materiality of the environment, and also within the…