Reconciling Art, Science, and Prediction in Real Estate Valuation: Toward a Hybrid Epistemology for the Ai Era

Abstract

Real estate valuation has long been characterised by tensions between subjective professional judgement and objective scientific methodology, resulting in a persistent art-science dichotomy within the discipline. Traditional valuation practice has relied heavily on interpretive expertise, contextual reasoning, and experiential judgement, while subsequent efforts to professionalise the field have emphasised standardisation, empirical validation, and methodological consistency. More recently, advances in artificial intelligence (AI), automated valuation models (AVMs), and machine learning have introduced a third paradigm, prediction, shifting valuation practice toward data-driven forecasting and algorithmic intelligence. These developments raise fundamental questions regarding the nature of valuation knowledge and the epistemological foundations that underpin professional valuation practice in the AI era. This conceptual paper examines the epistemological evolution of real estate valuation across art, science, and prediction paradigms and argues for the need to move beyond binary debates that position these approaches as competing or mutually exclusive. Adopting a conceptual analytical approach, the study synthesises peer-reviewed valuation literature, professional valuation standards, and contemporary research on AI-enabled valuation. Through comparative epistemological analysis and the application of a Pragmatist philosophical lens, the study develops the Valuation Epistemological Continuum (VEC), a unifying framework that conceptualises valuation knowledge as progressing from interpretive understanding to methodological validation and predictive optimisation. The findings demonstrate that art, science, and prediction are underpinned by distinct yet complementary epistemic logics, each contributing essential forms of knowledge to valuation practice. Rather than displacing professional judgement, predictive intelligence can be understood as extending the epistemological repertoire of valuation when appropriately integrated. This paper contributes theoretical clarity to emerging debates on AI-enabled valuation and provides a coherent epistemological foundation for future empirical, technical, and interdisciplinary research, particularly studies examining structured human-AI collaboration in real estate valuation. As a conceptual and theoretical study, this paper does not involve empirical model testing or performance evaluation.