Can Artificial Intelligence Make Us Happier? Dynamic Capability-Dependency Theory and a Two-Horizon Framework for AI-Mediated Well-Being
Artificial intelligence increasingly mediates work, learning, emotional support, decision making and social interaction, yet the central welfare question remains under-theorized: when does AI make human beings happier, and when does an apparent gain in the present become a loss in the future? This paper develops Dynamic Capability-Dependency Theory (DCDT), an integrative framework that treats AI-mediated happiness as a two-horizon process. The first horizon concerns acute affective relief, convenience and enjoyment. The second concerns stocks of competence, autonomy, human relatedness, meaning and dependency that accumulate through repeated use. Evidence from randomized trials, longitudinal studies and workplace deployments indicates genuine near-term benefits, including productivity gains, symptom reduction and temporary reductions in loneliness, but also shows that outcomes vary with design, usage intensity, autonomy, relational context and time horizon. DCDT formalizes these mechanisms as a dynamic state model and introduces the Temporal Well-Being Reversal condition, in which initially positive AI effects become negative after capability erosion, relational substitution or dependency accumulates. The paper derives testable propositions, specifies an empirical research program and introduces an AI-Happiness Impact Assessment that evaluates both immediate and durable effects. The central claim is not that AI is intrinsically happiness-enhancing or happiness-reducing. Rather, AI changes the production function of happiness by redistributing effort, agency, attention and relationships across time. High-value AI therefore should be judged not only by how well it satisfies a user now, but by whether repeated use leaves that user more capable, more autonomous, more connected to other humans and better able to pursue a meaningful life.