AI Recommendation Systems and Consumer Trust: A Conceptual Model of Personalisation, Explainability, and Privacy Risk

Abstract

Artificial intelligence (AI) based recommendation systems now mediate most consumer discovery journeys in e-commerce, yet the literature linking recommendation quality, personalisation, and explainability to consumer trust and purchase intention remains scattered across disconnected empirical streams. This conceptual paper synthesises theory from recommendation-agent research, trust theory, the Stimulus-Organism-Response paradigm, and explainable-AI scholarship to build a single, internally consistent, testable model, integrating propositions from 92 peer-reviewed sources indexed in Scopus and/or Web of Science, with a deliberate concentration on 2020-2026 literature spanning e-commerce trust research, algorithm aversion/appreciation psychology, hybrid Structural Equation Modelling-Artificial Neural Network (SEM-ANN) methodology, and privacy-calculus research; no primary data or student theses were used. The synthesis yields a six-construct integrative model - AI Recommendation Quality, Personalisation, Consumer Trust, Perceived Value, Perceived Privacy Risk, and Purchase Intention - in which trust operates as the central mediating mechanism, explainability acts as a positive boundary condition, and perceived privacy risk acts as a negative boundary condition on the recommendation-to-trust pathway, alongside a recommended two-stage hybrid SEM-ANN validation procedure. Rather than merely cataloguing prior findings, the paper derives a novel integrative structure, articulates eight falsifiable propositions, and proposes a concrete methodological agenda - including sample sizes, measurement sources, and analysis sequence - for scholars testing the model empirically, particularly in under-studied markets such as India.