Home > Programs > Zenith Grants > Zenith Grant Awardee

Zenith Grant Awardee

Samuel Craig Fletcher

Regents of the University of Minnesota

Project Title

Agential Abstraction/Representation Theory

Project Summary

What's the difference between a rock and a laptop? One is intuitively a computer, and the other not. Is there a principled difference between objects that compute, and those that don't? Does the universe compute, in some sense? Philosophers and cognitive scientists have developed different ways of understanding how the brain might be a computer, but such accounts of physical computation–what it means for a physical system to compute–are increasingly relevant in the natural sciences. Not only have biologists and computer scientists looked to the natural world to find systems that might compute with hardware very different from electronic laptops, but physicists have sought to understand the working of quantum mechanics and the universe itself in computational terms. This project develops Agential Abstraction/Representation Theory (AART), a new account of what it means for a physical object to compute that centers intelligent agents. According to AART, computation occurs whenever an intelligent agent actually and essentially uses an object to compute. The project will then explore the implications of this theory for understanding physical and biological systems in computational terms and for the energy costs of computing.

Technical Abstract

Scientists and philosophers interested in whether the mind can be understood as a computer, or whether nature is fundamentally computational, have sought to find a principled and objective way of distinguishing between objects that compute and those that don't. Building on work by Horsman et al., I have recently outlined Agential Abstraction/Representation Theory (AART) as an account of physical computation that addresses this need by centering the concepts of intelligent agency and epistemic community: what counts as a computation is relative to the cognitive faculties of an agent performing the computation and to the community of knowers to which that agent belongs. In this project, I extend AART in two ways. The first is to explain why many natural processes can be fruitfully modeled as if they were computational, even though according to AART no computation occurs because there are no intelligent agents representing abstract functions in those processes. The second is to apply AART to a debate concerning the justification and validity of Landauer's principle, which connects physical computations with thermodynamic costs of entropy and energy. I will thus test AART's potential to render computation objective yet grounded in the abilities of communities of intelligent agents.

Skip to content