Artificial Cognition: A Framework for Self-Instantiated, Temporally Continuous, Disturbance-Driven Adaptive World-Builders

Collaborating Institutions:

CLEMSON UNIVERSITY

METACOGNITION INSTITUTE

As AI technologies evolve beyond generative engines capable of producing text, images, and media, a new paradigm emerges for robotics and autonomous systems. Our research proposes an Artificial Cognition framework that transcends current computational limitations through four essential pillars: self-instantiation, temporal continuity, disturbance-driven adaptation, and autonomous world-building.

Unlike traditional AI architectures constrained by functionalist approaches, our framework addresses the fundamental requirements for consciousness-like behaviors that cannot be achieved through computational scaling alone. By integrating self-representation mechanisms with persistent memory systems, we create agents capable of maintaining an internal narrative and identity across time. These systems respond to environmental surprises through intrinsic feedback loops, triggering genuine learning and adaptation while constructing rich internal world models.

​

​

Belief Explorer: A Multi-Perspective AI Framework for Mitigating Disinformation

​

Collaborating Institutions:

CLEMSON UNIVERSITY

METACOGNITION INSTITUTE

​

In today's digital landscape, the challenge of discerning credible information from sophisticated disinformation has reached unprecedented levels. Traditional fact-checking methods, with their binary true/false classifications, fail to address the complexity and context-dependent nature of modern claims. Our research introduces Belief Explorer, an advanced AI framework designed to empower critical evaluation through multiple analytical perspectives.

​

Belief Explorer combines persistent contextual memory with Socratic dialogue techniques and a three-lens analytical pipeline. As users interact with the system, their inputs are systematically segmented and stored for comprehensive analysis. Each claim undergoes evaluation through distinct arbiters—Empirical, Logical, and Pragmatic—leveraging specialized LLM architectures to produce nuanced assessments. These evaluations synthesize into key metrics: Verifact Score (evidence strength), Model Diversity Quotient (inter-arbiter agreement), Contextual Sensitivity Index (scenario appropriateness), and Reflective Index (exposed assumptions).

​

The framework's Perspective Generator crafts alternative viewpoints, promoting epistemic humility and moving users beyond seeking definitive "truth" toward evaluating the pragmatic utility and coherence of information models. This capability represents a vital skill in an era of amplified, often weaponized narratives, fostering resilience against manipulation through deeper understanding and critical thinking.