The work explores whether machines can be designed to formulate autonomous intentions and make conscious decisions, and how such capabilities would influence their ethical behavior. Through case studies, it argues that advances in understanding artificial consciousness can contribute to developing ethical AI systems, suggesting that conscious-like decision-making in machines may be a pathway to more robust ethical behavior.
Consciousness remains undefined despite centuries of study, leading to multiple theoretical models. Baars' Global Workspace Theory is currently ranked the most promising theory of consciousness, yet few mathematical models of cognitive architectures inspired by it exist, and most are a decade old—outpaced by advances in artificial intelligence. Recent publications offer simple, computer-implementable mathematical models. This paper reviews consciousness and robot consciousness, introduces Baars' theory, and describes the most promising cognitive architecture models that implement it, noting their features.
The paper argues that current approaches to building conscious robots, collectively termed Good Old-Fashioned Artificial Consciousness (GOFAC), share a flawed conceptual framework. GOFAC includes methods like global workspace theory, information integration, enaction, cognitive mechanisms, and embodiment. The authors identify the intermediate level fallacy as the central problem with GOFAC, where researchers mistakenly assume that implementing cognitive or neural correlates of consciousness will produce subjective experience. The paper outlines an alternative conceptual framework for robot consciousness, though it does not provide a concrete design or empirical results.