Epistemic Communities, Social Scripts, AI

Hello everybody and welcome back to the Active Inference Insights podcast brought to you by the Active Inference Institute. I'm your host, Darius Parvezi-Wayne, and today I have the pleasure of chatting to Mau Alboracen. Mau is a Cognitive Computing PhD candidate at the University of Quebec in Montreal, as well as the Director of Product Innovations and Research at Versys, a company which leads the way in applying the Active Inference Framework to AI. With expertise in computational neuroscience, sexology, consciousness, and much more, Mau's work is broad, innovative, and deeply exciting, which makes me rather excited for this conversation. Mau, welcome to the show. Thank you so much for joining us. It's, as I said, it's an absolute pleasure. Well, thanks for having me. So, yeah, as I sort of mentioned in the introduction, your work covers a very broad range of topics and areas of research, but as you know, this podcast is exploring the themes of Active Inference and the Free Energy Principle. So, I want to mainly focus on that, even though we can see about how these different themes tie in with that overarching framework. If one goes on your Google Scholar and sees your work on Active Inference, again, they'll see it's diverse. Where I want to start is the paper that you wrote on epistemic communities. I guess the first question our audience might have, just as a reminder for those listening, this podcast is very much meant as a gentle introduction to Active Inference, as gentle as an introduction to Active Inference can be. So, just for our audience, my first question is very simple, and it's just, what is an epistemic community? So, it has many definitions. It's a loosely used term in a variety of fields, but the way we've used it is essentially a community of people that share very specific types of beliefs that lead them to sample reality in the same way. So, if you think about, like, your norms, they usually are tied to a specific way of thinking. Like, for instance, the fact of going to work. The fact of going to work nine to five is tied to an ideology that values a certain type of productivity, consumerism, etc., and a certain type of relationships, where we are together in the morning, together in the evening, and we are with different people during the day, etc., etc., etc. So, your life becomes shaped by these ideas that give rise to norms. Cool. And let's sort of tie in the Active Inference Framework then to this idea that we have a community which is tied together through sampling similar norms, similar sensory observations. Perhaps you could start just by giving us a brief or, yeah, relatively accessible notion of what Active Inference is. I mean, Carl did a very good job last week at explicating this, maybe just as a reminder to people. And then how that framework and the truths that the free energy principle discloses can be applied to the formation of an epistemic community. Yeah. So, Active Inference as a principle is rather simple. It's really just the idea that we try to minimize how much we are likely to dissolve. We don't want to die. So, in order to do so, we're going to try to map the world and understand what might be coming such that we can avoid moments where we might be dissolving, or moments that are dangerous for us. More surprising. We try to create evidence of ourselves. We try to predict ourselves existing in the future. And to do this, we make a prediction, act in the world, try to change something, see if it changes the way we thought it was going to change, and then update our model of the world. So you've got this sensory active loop where the agent is embedded in the world rather than just outside of it, mapping the world. So, it's a pretty crucial idea that the agent necessarily has a perspective which is defined by how they are separate and yet embedded in the world. They have a boundary. We generally refer to that as a mark of blanket. This boundary is both defining and defined by the…