Truth or lie: Exploring the language of deception.

Area of Science:

• Psycholinguistics
• Computational Linguistics
• Forensic Linguistics

Background:

• Lying is prevalent in communication, necessitating reliable detection methods.
• Previous research suggests linguistic differences exist between true and false statements.
• Cross-linguistic validation of these indicators is needed.

Purpose of the Study:

• To confirm previous findings on statement complexity and sentiment differences between truths and lies.
• To investigate if English-derived linguistic deception indicators apply to the Polish language.
• To develop and evaluate automated classifiers for truth-lie detection.

Main Methods:

• Analysis of approximately 1,500 true and false statements (spoken transcripts and written texts).
• Utilized tools for measuring linguistic complexity, sentiment, utterance length, concreteness, and parts-of-speech.
• Development and testing of machine learning classifiers for statement veracity.

Main Results:

• False statements exhibited lower vocabulary complexity, greater conciseness, and higher concreteness.
• Deceptive statements contained more positive and fewer negative words compared to truthful ones.
• No significant differences were found between spoken and written deception; classifiers achieved 60% accuracy.

Conclusions:

• Linguistic features associated with deception in English are also observable in Polish.
• Automated detection of deception using linguistic analysis is feasible, though accuracy requires improvement.
• Findings contribute to understanding deception indicators in cross-linguistic contexts.