Setting the Demons Loose: Computational Irreducibility Does Not Guarantee Unpredictability or Emergence

Hamed Tabatabaei Ghomi

doi:10.1017/psa.2022.5

Abstract

AbstractA phenomenon resulting from a computationally irreducible (or computationally incompressible) process is supposedly unpredictable except via simulation. This notion of unpredictability has been deployed to formulate recent accounts of computational emergence. Via a technical analysis, I show that computational irreducibility can establish the impossibility of prediction only with respect to maximum standards of precision. By articulating the graded nature of prediction, I show that unpredictability to maximum standards is not equivalent to being unpredictable in general. I conclude that computational irreducibility fails to fulfill its assigned philosophical roles in theories of computational emergence.

Highlights

Predictability is a common area of interest and investigation for both scientists and philosophers
Via a technical analysis of computational irreducibility, I show that computationally irreducibility can establish the impossibility of prediction only with respect to maximum standards of precision
I conclude that computational irreducibility fails to fulfill its assigned philosophical roles in theories of computational emergence

Summary

INTRODUCTION

Predictability is a common area of interest and investigation for both scientists and philosophers. Wolfram (2002) claims that from the computational perspective, the outcomes of computationally irreducible processes are unpredictable for any observer Philosophers such as Bedau (1997) and Huneman (2008) have adopted this view to formulate ontological weak emergence, according to which emergent phenomena are unpredictable for any observer because of the computational irreducibility of the path leading to those phenomena. I show that computational irreducibility, no matter how interpreted, guarantees unpredictability only with maximal standards of accuracy and precision, while the scientifically and practically relevant concept of predictability is graded and submaximal. This creates a dilemma for supporters of computational emergence; they cannot have the ontological cake and eat it. The arguments will take ontological weak emergence off the table, and we go back to the choice between weak and strong emergence

COMPUTATIONAL IRREDUCIBILITY

COMPUTATIONAL IRREDUCIBILITY AS A CASE OF THE HALTING PROBLEM

ALGORITHMIC COMPUTATIONAL IRREDUCIBILITY IN TERMS OF

PREDICTION IS GRADED

ONTOLOGICAL WEAK EMERGENCE AND COMPUTATIONAL

DEGREES OF EMERGENCE

Findings

CONCLUSIONS