The REAL framework: evaluating the cognitive and formative impact of generative AI

Purpose This paper aims to address a gap in contemporary scholarship on generative artificial intelligence (AI) by shifting attention from the functional evaluation of AI systems to the formative impact of repeated AI use on the human user. It introduces formative evaluation as a complementary approach concerned with how sustained AI use may shape cognitive skills, habits, reasoning, judgment and agency over time. To address this concern, the paper proposes the REAL Framework (Retained, Eroded, Atrophied, Leveraged) as a practical means of evaluating and intentionally shaping patterns of AI use. Design/methodology/approach This conceptual paper develops the REAL Framework through interdisciplinary synthesis and theoretical analysis. It integrates scholarship on human agency and principal-agent relationships with research on cognitive offloading, cognitive reserve, desirable difficulty, automation bias and human–AI collaboration. Emerging empirical research on the cognitive effects of generative AI is used to contextualize and support the framework. The resulting four-dimensional framework is then illustrated through its application to professional practice and considered in relation to potential objections, limitations and directions for future empirical research. Findings The analysis identifies formative evaluation as a distinct and underdeveloped dimension of AI evaluation. While emerging research increasingly documents potential cognitive effects of generative AI use, practical means of evaluating and responding to those effects remain limited. The REAL Framework addresses this gap through four dimensions: Retained, Eroded, Atrophied and Leveraged. Together, these dimensions distinguish between AI use that preserves human agency and cognitive engagement, use that risks over-delegation and capability loss, and use that intentionally leverages AI to enhance human thinking and judgment. Research limitations/implications The REAL Framework is a conceptual and reflective framework that has not yet been empirically validated or operationalized as a measurement instrument. Longitudinal research is needed to examine the formative effects of sustained generative AI use and determine whether structured application of the framework influences patterns of AI delegation or preserves cognitive engagement and capability. Future research should develop and test measures associated with the framework’s four dimensions and examine their applicability across educational, professional and organizational contexts. Practical implications The REAL Framework provides individuals and institutions with a practical means of evaluating how generative AI is being incorporated into learning and professional practice. Its four reflective dimensions can help users identify inappropriate over-delegation, preserve important cognitive and professional capabilities, and redesign AI use to support rather than replace human reasoning and judgment. The framework may also inform institutional guidance, professional development and AI literacy initiatives by shifting attention from whether AI should be used toward how it can be used while preserving meaningful human agency and engagement. Social implications As generative AI becomes increasingly embedded in education, work and everyday decision-making, its societal effects will depend not only on the capabilities of AI systems but also on how their use shapes human capabilities over time. The REAL Framework highlights the importance of preserving human agency, reasoning, judgment and cognitive engagement while benefiting from technological advancement. It encourages a broader societal conversation about which human capacities should remain meaningfully exercised and which may appropriately be delegated to increasingly capable AI systems. Originality/value This paper introduces formative evaluation as a distinct conceptual approach to understanding generative AI use, complementing prevailing functional approaches concerned primarily with AI systems, governance and responsible use. It advances the REAL Framework as a structured means of translating emerging concerns about AI’s cognitive and formative effects into reflective practice. By connecting human agency and delegation with cognitive development, the framework shifts attention from evaluating what AI can do to evaluating what repeated AI use may be doing to the human user.