How Could Consciousness Be Cultivated in a Machine?

Abstract

This chapter addresses the difficult scientific and philosophical challenge of engineering and then verifying consciousness within artificial agents. Moving beyond traditional focuses on perception and cognition, the authors argue that ‘affect’ – the raw capacity for feeling – is the fundamental form of consciousness. By defining consciousness as a subjective state rooted in an extended form of homeostasis, the chapter provides a mechanistic explanation for why feelings arise and what functional purpose they serve in navigating uncertain environments. Readers will gain insight into a computational architecture based on the ‘Free Energy Principle’ and ‘Markov blankets’, which aims to engineer an agent that truly ‘gives a damn’ about its own survival. The authors propose a rigorous, falsifiable test for machine consciousness – the hedonic place preference test – analogous to tests used in animal research to infer subjective experience. By demonstrating how an artificial agent can be ‘hijacked’ by deceptive stimuli, the chapter illustrates how subjective models and felt uncertainty might bridge the gap between purely deterministic machines and truly agentic, conscious beings.