Ai Unveils the Mole? Language Algorithm Predicts Reality Show Saboteur

A language technology researcher at UAntwerpen, Sara Budts, along with master's student Elin Dillen, developed an AI algorithm to identify the saboteur in the reality show "Wie is de Mol?" The algorithm analyzes distorted diary entries for linguistic patterns. Budts noted that police services sometimes use AI to compare language in ransom letters with potential suspects. She applied a similar approach to the show, using stylometry to analyze candidates' language. The AI was trained by meticulously transcribing two previous seasons of the show. The algorithm achieved success in identifying the mole in four recent seasons, except for Alina. The key is precise transcription, focusing on what candidates actually say, not cleaned-up subtitles. Linguistic fingerprints are found more in grammar than in easily imitated filler words. According to the AI, Michèle is the most likely saboteur among the finalists, with an 80% certainty. This prediction aligns with that of Dutch mathematics teacher Jasper de Jong, who uses a different method. Budts humorously anticipates potential backlash from the show's host, Gilles De Coster. She also notes that the introduction of the "molkoffer" provided more diary entries, aiding the algorithm's analysis. Observations from the analysis include Pedro's actions appearing even more conspicuous and Hilde's frequent use of the expression "getverdemme" (a Dutch expression of annoyance or disgust). Budts also jokingly declined to create an algorithm to translate their cat Svetlana's meows.