Becoming as inference
philosophy, dynamical systems, process philosophy, pragmatism, probabilistic inference, physics, statistical physics
historical note
I originally connected these ideas on and decided to continue this line of thinking under the title of this page Becoming as inference on .
brief description
If we just try to describe what scientists do without examining the internal experience they have, then we think of them as constructing models and performing Bayesian model selection and parametric inference. In addition to this explicit application of probabilistic inference, we also have a model of the implicit functioning of the brain that considers it as a device that performs probabilistic inference, which has been associated with several different names including the predictive coding model, Bayesian brain hypothesis, prediction error minimization framework, or the free energy principle. If we consider the functioning of simple organisms in this same context, we can construct an analogy to the Bayesian brain hypothesis that results in various additional Bayesian `X` hypotheses such as the Bayesian biology hypothesis, the Bayesian chemistry hypothesis, Quantum Bayesianism, and so on.
On one hand this suggests that developing a sufficiently flexible framework for describing the processes within which probabilistic inference occurs is potentially fundamental. On another, it may be that at this level of abstraction doing so would eventually appear vacuous.
pan-inferentialism, the notion that probabilistic inference is embodied by dynamical systems and unifies the curiosity of humans as manifest in the experience of scientists with the models they create of the functioning of physical and biological systems
key references
philosophy
stochastic processes and statistical physics
probabilistic inference
- A tutorial on the free-energy framework for modeling perception and learning by Rafal Bogacz in 2017
Bayesian brain hypothesis
systems biology
outline
I. The fundamental role of dynamics in our understanding of the world
- pragmatic process philosophy derived from Nicholas Rescher
II. Modeling phenomena within the framework of probabilistic inference
III. The scientist’s experience of inference and the scientist’s experience as inference
- Martin Biehl’s active inference synthesizes the predictive coding model of The Helmholtz Machine by Peter Dayan, Geoffrey Hinton, Radford Neal, and Richard Zemel in 1995, Karl Friston’s free energy principle, Jakob Hohwy’s prediction error minimization framework all of which have previously been associated with the Bayesian brain hypothesis.