Associative Integration and Information Selection in Intelligence: A Diagnostic Framework and Research Program

Intelligent systems must combine available information and use the resulting structures appropriately. This article develops a diagnostic framework and research program around two organizing hypotheses: associative integration provides the foundation of intelligent activity and information-guided selection adapts the resulting structures to the world; randomness can expand the effectively reached repertoire of associations, expressions, and exploratory actions under finite resources. The framework distinguishes component availability, integration, accessibility, and appropriate use. Its broad foundation thesis is a programmatic commitment. A separate system-specific mechanism hypothesis tests prespecified integration routes, rather than functional necessity already implied by a task definition. The proposed contribution is a coordinated comparison of relational interventions in humans and learning machines. Eight local predictions span learning, memory, generalization, embodiment, reasoning, search, and collective inquiry. A ninth, exploratory prediction tests cross-capacity coherence through shared-participant measurement and shared-route interventions. Priority designs separate composition benefits from misleading scaffolds and test competitor-identity errors across relation confusability, with error-conditional checks for changes in overall error rate. A further factorial test compares stochastic proposals with deterministic diversification under informative or uninformative selection. Trial-level planning simulations illustrate sensitivity to participant and item variation; they supply no evidence for the theoretical hypotheses. The program's value depends on comparative prediction and coordinated intervention effects beyond established mechanisms.