This paper introduces "pseudo-consciousness" as a novel framework for understanding and classifying advanced artificial intelligence (AI) systems that exhibit sophisticated cognitive behaviors without possessing subjective awareness or sentience.Traditional AI classifications often rely on a binary distinction between narrow, task-specific AI and hypothetical artificial general intelligence (AGI). However, this dichotomy fails to adequately address the growing number of AI systems that demonstrate capabilities such as reasoning, planning, and self-monitoring, yet lack any demonstrable inner experience.We propose pseudo-consciousness as an intermediate category, bridging the gap between reactivity and genuine consciousness. We also define five operational conditions that characterize pseudo-conscious AI: Global Information Integration (GII), Recursive Metacognitive Correction (RMC), Cross-Domain Transfer Competence (CDTC), Intentionality Simulation Without Subjectivity (ISWS), and Behavioral Coherence Across Domains (BCAD).These conditions, grounded in computational functionalism, cognitive science, and neuroscience, provide measurable criteria for differentiating pseudo-conscious systems from both simpler, reactive AI and speculative AGI.The framework offers a structured approach to evaluating AI based on how it achieves complex cognitive functions, rather than solely on what tasks it can perform. This distinction is crucial for addressing the ethical, societal, and regulatory challenges posed by increasingly autonomous AI.By recognizing pseudo-consciousness as a distinct and stable category, we can better inform AI design, governance, and public discourse, ensuring responsible development and deployment of AI systems that mimic aspects of cognition without possessing genuine consciousness. The framework facilitates a more nuanced understanding of AI capabilities, moving beyond simplistic "conscious" vs. "unconscious" classifications.
Jose Augusto de Lima Prestes (Thu,) studied this question.