Human stochastics: Unconscious, ratiomorphic processes as the foundation of perception and learning

Abstract

The purpose of this paper is to demonstrate how unconscious-ratiomorphic processes are fundamental for all perceptions and learnings in general. This is especially relevant due to artificial intelligence (AI) is presenting us with new challenges. So, the “end of all theory“ has even been announced as a provocation to scientific theory — because ample correlations are supposedly sufficient from a pragmatic point of view. Is this really a loss? The answer is: No, because there have never been real certainties in the philosophical sense. However, people generally find uncertainty unpleasant, as psychology and cognitive science prove. There are theoretical and practical reasons: If we dispense with “causality“, methodological abysses open up. Our everyday life also relies on causality, because responsibility seems impossible without it. So, is “technical stochastics“ incomprehensible, useless or even dangerous in the age of AI? No, because we can't model cognitive reality using only conscious-rational processes (as demanded by epistemological rationalists). However, perception and learning are based on unconscious-ratiomorphic processes — what we call “human stochastics“. Humans are not passive beings. Rather, they constantly carry out active inferences based on epistemic actions. The brain uses Bayesian statistics to minimize “prediction errors“. As a “predictive mind“, it uses a hierarchical multi-level model to simultaneously examine invariances at various granularities — and thus improves the prediction of action effects. Perception and learning are understood as stochastic processes. Taken literally, “human stochastics“ is unavoidable, childishly simple and commonplace. Hence, with “human stochastics“ we can model the “actual genesis“ of percepts and learning dynamics in detail.