Inference to the Best Decision

Here is the picture so far.Starting from conceptual analy sis of existence, selfevidencing unfolds in terms of perceptual and active inference, proceeding in a self-supervised manner within a Markov blanket through precision control.Guided by balancing of abductive reasoning pro cesses, self-evidencing works through finessing of an internal generative model and selective sampling of observations.Those observations in turn provide evidence for the beliefs encoded in that model.As the model is progressively finessed, it comes to represent the causal structure of the environment, where individual agents' self-evidencing paths can take dif fer ent trajectories, depending upon their unique experiences and ensuing empirical priors.Now we shift our focus more directly to active inference, decision-making, and action.We ask specifically how self-evidencing relates to rationality.Self-evidencing leads to a conception of rationality that I will label inference to the best decision.I suggest that this is a compelling concept that unifies descriptive and normative accounts of rationality and that makes sense of the way that we muddle through life's decisions.I begin this chapter by linking the imperative for self-supervision to trial and error pro cesses in action, further developing the hill-climber analogy (encountered in chapter 3).I suggest taking an abductive perspective toward decision-making, in terms of inference to the best decision, designed as a parallel to inference to the best explanation.I follow this with a list of the best-makers for this notion.I bring this closer to everyday decision-making by focusing on deliberation and counterfactual reasoning, which I then extend to exploration and exploitation.I end by proposing that rationality just amounts to inference to the best decision.