Keyboard dynamics discrepancies between baseline and deceptive eyewitness narratives

Abstract

Information manipulation and cognitive load imposition make the production of deceptive narratives difficult. But little is known about the production of deception, and how its mechanisms may help distinguish truthful from deceitful narratives. This study focuses on the measurement of keystroke dynamics while typing truthful and deceptive eyewitness testimonies after a baseline assessment. While typing their narrative, some participants would undergo an auditory cognitive load. Results show that liars typed their story slower, and in less time than the truthful participants when compared to their respective baselines. The imposition of the auditory cognitive load showed adverse results, enhancing the amount of keystrokes and the time necessary to type the narrative. Classification shows better results for deceptive narrative when no auditory cognitive load is imposed. These results are discussed in terms of expanding current models to include the cognition of linguistic production and writing strategies.

Computational Measures of Deceptive Language: Prospects and Issues

“…This accuracy is particularly interesting as it includes field data and experimental data. If experimental results compose most of these studies (e.g., Masip et al, 2012;Tomas et al, 2021c), other studies have focused on real-life data and shown similar results (Fornaciari and Poesio, 2013;Pérez-Rosas et al, 2015a, 2015b. For instance, an 84% accuracy rate in detecting misleading opinions was reached in a study using lexical (e.g., vocabulary richness) and syntactic (e.g., punctuation) stylometric features, as well as supervised learning methods (i.e., classification by algorithmic methods based on texts labeled as authentic or misleading, Shojaee et al, 2013).…”

“…Indeed, lie detection from stylometric cues is a testable hypothesis: It is possible to determine, through human or algorithmic classification methods, whether the use of computational stylometry detects deception or not, in accordance with the falsifiability principle (Popper, 1959). Second, computational stylometry has been tested for its ability to detect deception with various tools and methods (e.g., for LIWC, see Ali and Levine, 2008;Fornaciari and Poesio, 2013;Newman et al, 2003;Tomas et al, 2021c; for named entity recognition, see Kleinberg et al, 2018; for morpho-syntactic labeling, see Banerjee and Chua, 2014; for n-grams, see Cagnina and Rosso, 2017;Hernández Fulsilier et al, 2015;Ott et al, 2013; for vector representations, see Nam et al, 2020; for BERT, see Barsever et al, 2020). Third, it has been the subject of over 20 peer-reviewed publications (e.g., Hauch et al, 2015;Forsyth and Anglim, 2020;Tomas et al, 2021a).…”

“…These commentaries covered the essential subjects regarding human judgements of deception and highlighted paramount issues on which researchers should focus. However, computational linguistic deception detection is barely mentioned in the publication of this 3-day workshop, while the interest for these tools and methods increases (e.g., Jupe et al, 2018;Kleinberg et al, 2018;Tomas et al, 2021c).…”

A Systematic Literature Review on Latest Keystroke Dynamics Based Models

“…A study [213] included 283 subjects, while some of the few studies [102], [214]- [217] considered more than 300 subjects. A recent study [218] used GPower for estimating sample size. In KD research, most of the studies included patterns collected from college or university students, teachers, support staff, etc.…”

Investigative Interviewing Research: Ideas and Methodological Suggestions for New Research Perspectives

“…Yet, this approach appears controversial and mostly ineffective (Vrij, 2016;Caso et al, 2019a,b). Efforts have been made to improve its efficacy (Palena et al, 2019;Verigin et al, 2020;Tomas et al, 2021b), and a recent study has questioned its relevance, without rejecting it either (Tomas et al, 2021a). Furthermore, although the baseline attempts to deal with personal characteristics, almost all research studies in investigative interviewing used a variable-centred approach (Magnusson, 1992(Magnusson, , 1998.…”