Been There, Done That: How Episodic and Semantic Memory Affects the Language of Authentic and Fictitious Reviews.

This article focuses on a theory-driven approach to distinguishing authentic from fictitious online product reviews by leveraging the cognitive distinction between episodic and semantic memory. It posits that authentic reviews rely more on episodic memory—recollections of specific lived experiences—while fictitious reviews depend primarily on semantic memory, which involves general knowledge. The authors identify five linguistic features—concreteness, uniqueness, informativeness, reduced reliance on external cues, and integration of what–where–when details—that characterize authentic reviews and validate these features through experimental data, analyses of real-world review datasets, and machine learning classification. Their theory-based algorithm, using a small set of psycholinguistically grounded features, demonstrated better cross-domain predictive accuracy in detecting fake reviews than several data-driven benchmark algorithms, suggesting greater generalizability across contexts. This work contributes a coherent psycholinguistic framework for understanding deceptive word-of-mouth communication and offers practical tools for improving the detection of fraudulent online reviews.