Truth Bias
With few exceptions, people have great difficulty in classifying another person as truth teller or liar. Many researchers have documented an inborn "truth bias" that prompts a listener or reader to believe what is said or written. This bias may be needed to facilitate social, familial and work relationships but it fails to help when deception detection becomes critical for making decisions that affect legal proceedings, hiring decisions and national security.
Statement Analysis
For decades, psychologists and forensic experts have studied the connection between deception and language in an effort to understand how word choice, sentence structure, amount of detail, and other factors can be used to separate truth from lies. Statement analysis refers to a collection of methods for associating linguistic behavior with deception. It rests on the hypothesis developed by Undeutsch (1989) that statements based on true memories differ in their use of language and detail from statements based on fabrication. Versions of statement analysis have been developed and applied by the FBI (Adams 1996), commercial organizations (LSI, QED), and academic researchers (Porter & Yuille 1996, Miller & Stiff 1993, Zhou et al. 2004) working with English and other Western European languages. In Germany, Criteria Based Content Analysis is an application of Undeutsch's proposals that has long been used for investigating child molestation cases.
Deception
Discovery Technologes
Deception Discovery has developed a statement analysis technology that can be applied to any unscripted narrative or interview. Two features of the technology set it apart: (I) The analysis uses exclusively- linguistic features that can be implemented in an automated system. Because the analysis is machine-generated instead of human-generated, no training is required and subjective judgments are minimized. (II) The focus of the analysis is on particular statements, not on the person making them. Rather than attempting to characterize a narrator or interviewee as deceitful, the goal of Deception Discovery analysis is to identify which parts of a narrative or interview are likely to be deceptive and which parts are likely to be truthful. This produces a more fine-grained analysis than is typically seen in other approaches to statement analysis. One benefit of the approach is that it can provide an interviewer with specific guidelines for further questioning or investigation.
Experimental results of Deception Discovery analysis show that our technology identified false statements 93% of the time. The results of less formal field tests on new data (depositions and interviews) indicate a high rate of accuracy in labeling true vs. false sections. For these documents, Deception Discovery used its proprietary automated system, which performed between 70%-80% of the analysis accurately. These results are expected to improve significantly as more sophisticated linguistic features are incorporated into the automated system.
References
Adams, S. 1996. Statement analysis: What do suspects’ words really reveal? The FBI Law Enforcement Bulletin. 65(10).
Miller, G. R. and J. B. Stiff. Deceptive Communication. 1993. Thousand Oaks, CA: Sage Publications.
Porter, S. & Yuille, J. (1996). The language of deceit: An investigation of the verbal clues in the interrogation context. Law & Human Behavior, 20(4) 443-458.
Undeutsch, U. 1989. The development of statement reality analysis. In J. Yuille (Ed.), Credibility Assessment. Dordrecht (NL): Kluwer, 101-121.
Zhou, L., Burgoon, Judee K., Nunamaker, Jay F. Jr. and D. P. Twitchell. 2004. Automating linguistics-based cues for detecting deception in text-based asynchronous computer-mediated communication. Group Decision and Negotiation, 13.