Functional Intelligence Framework

The Functional Intelligence Framework (FIF) addresses a fundamental problem in intelligence research: observed performance is routinely used to infer intelligence, although performance depends on multiple interacting systems that are not themselves intelligence. FIF proposes a functional architecture explaining how intelligence is transformed into observable performance. It defines intelligence as adaptive predictive modeling capacity and distinguishes this capacity from the physiological and psychological state of the person, the availability of their cognitive capacities, the demands of the task, environmental conditions, and the mechanisms through which intelligence is expressed. The framework specifies eleven interacting functional components, two global modulators—State and Access—and normative-interpretive frameworks that shape cognition and evaluation without constituting intelligence itself. It explains how individuals with comparable underlying intelligence may perform differently, how similar performance can arise from different functional mechanisms, and why impaired performance does not necessarily indicate diminished intelligence. Beyond its theoretical contribution, FIF establishes a research programme for empirical evaluation and measurement. Six accompanying appendices examine existing evidence, operationalize functional components using available datasets, specify separate assessment profiles for Person, Task, and Environment, compare FIF with established theories, define falsifiable predictions, and introduce the Open Intelligence Measurement Instruments (OIMI) initiative. FIF therefore proposes a shift from treating performance as a direct proxy for intelligence toward investigating the functional architecture through which intelligence becomes observable, constrained, and measurable. Manuscript version: 1.9Research initiative: How Intelligence WorksPublication status: Preprint under journal peer review. The research was first submitted for journal consideration on 6 July 2026. An earlier version was registered on OSF on 1 August 2026. This deposit makes manuscript version 1.9 publicly available.