Human-Centered Artificial Intelligence: Integrating Human Cognition, Psychological Assessment and Responsible Digital Behavior for Real World Well-Being

Abstract:

This article enunciates a construct-centered synthesis of artificial intelligence as a psycho-technical infrastructure that increasingly mediates cognition, affect, motivation, identity, and social coordination across clinical, educational, organizational, and civic ecologies. It argues that psychologically consequential AI must be governed first as a construct validity problem and only second as a predictive optimization problem, because statistically impressive models can remain ontologically mis-specified when they operationalize socially contingent proxies, ignore contextual meaning, or induce feedback-driven performativity. The article builds an epistemic bridge between psychological explanation and algorithmic prediction by translating nomological networks, reliability, measurement invariance, and counterfactual reasoning into design-grade requirements for AI systems that assess, profile, recommend, or intervene. It then reframes AI as an implicit theory-engine of mind and behavior, integrating dual-process cognition, signal detection, evidence accumulation, reinforcement dynamics, and extended cognition to explain predictable human-AI failure modes such as automation bias, miscalibrated trust, attentional capture, and dependency formation. The article further specifies how psychometrics, affective inference, and trait profiling can fail under proxy confounding, dataset shift, concept drift, and explanation illusions, and it delineates mechanism-fidelity constraints for AI-mediated intervention grounded in process-based change, autonomy support, and iatrogenic risk minimization. Finally, it advances a global governance lens that treats psychological integrity, equity, contestability, and lifecycle auditability as enforceable socio-technical obligations, offering a cross-sector roadmap for academically rigorous, policy-relevant, and implementation-ready psychologically safe AI.