Games and will appropriation

I often wonder whether the fact that games were generally thought of as the quintessential analogue of intelligence is just due to their clear links with liberalism; you have an environment and you have agents that try to maximise some notion of reward in that environment. Life seems radically different (no clear rewards, no environmental resets, no fixed set of actions, no clear transition function), but most games provide a good abstraction for a competitive universe. Nobody plays games were your first state is the most rewarding state – rewards accumulate over time. You start as a horrible low life and you keep getting progressively refined as a player/agent to the point where you reach some far away game heaven, the land of high scores and opponent defeats. The traditional control paradigm had an inverse reward structure – you start happy, but your environment tries to make you miserable, so you need to discover the right set of actions to remain stable. Mathematically, you end up with exactly the same solutions, it’s just a matter of perception. More modern work sometimes does away with external rewards completely, and just assumes that an agent maximises some internal intrinsic reward.

In all cases the underlying theme is that the universe/your head won’t just let you be. If you are a hippo laying around in your muddy pool having a good time, at some point you will either be hunted by a lion (i.e. negative external reward) or feel hungry (i.e. negative internal reward) and have to act. The (liberal) moral of the story here is agency – you can change your fate by choosing your actions. In this sense liberalism tries to empower people to allow them to play the game in ways that maximises their future freedom. This has been formalised by various authors, popular texts include “Salge, Christoph, et al. “Empowerment–an introduction.” Guided Self-Organization: Inception. Springer, Berlin, Heidelberg, 2014. 67-114.” and Turner, Alex, et al. “Optimal Policies Tend To Seek Power.” Advances in Neural Information Processing Systems 34 (2021)..

This “flourishing” that liberalism offers is one of its most attractive propositions, but the whole edifice starts to break down when confronted with the fact that one agent’s environment is another agent’s will – i.e. there are other consciousnesses than you in this world and your empowerment might as well mean the literal enslavement of somebody else. Liberalism has trouble accepting this as a proposition (hence the fixation with abstract law and the market, i.e. the rules of the game). Following the breakdown in the alliance between liberals and socialists in the early 80s (see Przeworski, Adam. Capitalism and social democracy. Cambridge University Press, 1986. for a discussion), the games we play tend to be more and more of the enslaving variety.