Evolving evolutionary bene ficial innate expected outcomes in active inference agents
Keywords
No Thumbnail Available
Authors
Issue Date
2021-06-25
Language
en
Document type
Journal Title
Journal ISSN
Volume Title
Publisher
Title
ISSN
Volume
Issue
Startpage
Endpage
DOI
Abstract
Active inference rests on a system of predictive processing where the
brain acts as a prediction engine called the generative model. In active inference,
the goal of the generative model is to predict what will be observed
as accurately as possible. An active inference agent can decrease prediction
error in two ways: adjusting the generative model to update predictions or
changing the world via actions to fit predictions. Critics point out that this
description lacks goal-directed behaviour. In response to this, proponents of
active inference often appeal to evolution. Over generations, humans would
evolve an innate expectation for certain observations that would result in
goal-directed evolutionary benefi cial behaviour. In this paper, we construct
a toy model in which active inference agents will evolve evolutionary benefi cial expected outcomes that result in goal-directed behaviour.
Description
Citation
Supervisor
Faculty
Faculteit der Sociale Wetenschappen