Probabilities and hallucinations

I’ve written before about how probabilities are used to understand human perception, understanding, and learning.  Joshua Tenenbaum uses probabilistic inferencing to account for how we come to learn concepts, acquire language, and understand the world around us quickly, and with very little information (How to Grow a Mind).   Optical illusions are created by statistical judgments that the brain seems to be carrying out (what we see is the most likely interpretation of sensory data).  In cognitive science, probabilistic programs are used more and more to model cognitive processes.  These programs contain the causal structure of some chosen range of possible events, and use probabilities to address the things that are not known. They run like simulations, providing a way to predict events when run forward, and a way to analyze them when run backward. Many of the models have strong predictive value when applied to real world situations. One such experiment, designed to determine how preschoolers will generalize an event, strongly supports the idea that 15-month old children used an intuitive sense of the effects of sampling to make a generalization in their immediate experience (or to not make one).

Anil Seth, at the University of Sussex, proposes that this nontrivial mathematical idea is responsible, not only for how we perceive and learn about the world around us, but also how we perceive ourselves. His research begins with the observation that the body creates what we see by organizing sensory signals and prior experience with probabilities. For Seth, we don’t so much perceive the world as generate it which, he notes, is consistent with a Helmholtzian view of things (see last month’s post).  And this generation of a world that we experience is not just the organization of signal coming from the outside. The brain is also addressing signal from the inside – like blood pressure or how the heart and internal organs are doing – as much, if not more, than the outside. This side of the brain’s attention concerns control and regulation. Seth argues that the brain is consistently engaged in error-reducing predictive processing, to keep us alive.

With the continuous flow of signal from outside the body and within the body, the brain makes its best guess about what’s happening.  Seth calls our shared experience a controlled hallucination. What we typically call a hallucination occurs when, for whatever reason, the brain pays less attention to signals coming from outside the body and runs more purely on expectations.

For Seth, conscious experience, and our multilayered experiences of selfhood, are also constructed from the organization of sensory data and internal states, and characterized by the most likely meaning of that data. Even our sense of owning our own bodies is a consequence of the brain’s predictions about the data it receives. Experimental evidence supports this claim. The rubber hand experiment, for example, is one in which an individual is asked to focus on a fake hand while their real hand is kept out of sight. The fake hand will begin to feel like part of the participant’s body when tactile stimuli are arranged to suggest this. In another experiment, the electronic image of a hand is perceived as belonging to the body when it is programmed to flash gently red in time with that individual’s heartbeat.

The brain’s probabilistic judgments are not easily unraveled. They are based on a complex set of physiological processes that include the body’s sense of its own position and its own motion, (produced by stimuli that arise from within the body), the body’s sense of its own internal state, as well as what it receives from the five senses that we recognize. What we perceive is some reconciliation of integrated signals and expectation. And when expectations are given priority over incoming data, one will perceive what is not there.

In this paradigm, mathematical processes are found living. They are a feature of our biology. But another important implication of this work is that, with all of these error-reducing guesses that the brain is making (using probabilities and past experience), we could be mistaken about anything – about what’s out there as well as about ourselves. In the history of science, mathematics has consistently corrected mistakes in our perception of reality – from the earth’s position in the solar system, to the calculus of classical mechanics, to the probability-driven theories of quantum mechanics. So it seems to be addressing our view of things from both ends.  As pure structure, related to fundamental life processes, mathematics may also be uniquely capable of clarifying what we can understand about ourselves or, even more importantly, what we can understand about our relationship to what we believe is our reality. Insights gained from a cognitive science perspective must, inevitably, connect to cosmological questions like Wheeler’s participatory universe, or the Qbism view of the wave function described here.  The promise of a genuinely fresh approach to longstanding philosophical issues, like the viability of an objective point of view, intrigues me in these discussions, and mathematics plays a consistently significant role in substantial paradigm shifts like this one.