Abstract

Effective methods for evaluating the reliability of statements issued by witnesses and defendants in hearings would be an extremely valuable support to decision-making in court and other legal settings. In recent years, methods relying on stylometric techniques have proven most successful for this task; but few such methods have been tested with language collected in real-life situations of high-stakes deception, and therefore their usefulness outside lab conditions still has to be properly assessed. In this study we report the results obtained by using stylometric techniques to identify deceptive statements in a corpus of hearings collected in Italian courts. The defendants at these hearings were condemned for calumny or false testimony, so the falsity of (some of) their statements is fairly certain. In our experiments we replicated the methods used in previous studies but never before applied to high-stakes data, and tested new methods. We also considered the effect of a number of variables including in particular the homogeneity of the dataset. Our results suggest that accuracy at deception detection clearly above chance level can be obtained with real-life data as well.