In the midst of Washington headlines dominated by the work of the National Commission on Terrorist Attacks Upon the United States (also known as the 9/11 Commission), COSSA focused its first Congressional Briefing of 2004 on a related topic.  The March 19 seminar, Detecting Deception:  Research to Secure the Homeland, highlighted three experts who have been studying techniques and approaches that security agencies can use to uncover threats to the nation.

                                                                                                                                Photograph by Charles Votaw

Pictured from left to right:  Howard Silver, Mark Frank, Charles Bond, and Judee Burgoon

Automated Detection

Judee Burgoon, Professor of Communication, Professor of Family Studies and Human Development, and Director of Human Communication Research for the Center for the Management of Information at the University of Arizona, discussed “Automating Detection of Deception and Hostile Intent.”  She noted that her presentation was built from her work on two major, multi-institutional projects:  1) a five-year multidisciplinary project funded by the Department of Defense on detecting deception in the military infosphere; and 2) an automated identification of intent project funded by the Department of Homeland Security.

For purposes of her work, Burgoon defined deception as “messages that are knowingly transmitted to create some kind of false impression,” often with hostile intent.  Intent can be defined as a cognitive state that causes external behaviors to be exhibited, which can then be interpreted by others.  Key issues, according to Burgoon, are whether we can tie noted behaviors to intent, whether we can accurately determine if a person is being deceptive or truthful and whether they pose a risk.

Traditionally, she stated, “humans have very poor ability to detect deception.”  To this end, one objective is to automate detection to augment human capabilities.  This could especially enhance efforts to detect deception in electronic, audio, and video messages – all common forms of communication in this day and age. 

One aim of Burgoon’s research has been to go beyond typical methods of judging deception.  For example, many people correlate speaking more rapidly or in a higher pitch with deception.  The same can be said for emotional indicators that show fear or guilt.  But not all deceivers, of course, feel these emotions or experience the type of arousal that causes changes in speech quality.  As a result, Burgoon and her collaborators have had to consider other possible signs, such as lacking details, the absence of emotional language, and false smiles.

Burgoon also discussed analyzing text messages in the context of an experiment involving a mock theft.  She noted that, in responding to questions, deceivers use much longer responses in text format than in face-to-face contact.  As a result, her team concluded that “text-based deception allows for planning, rehearsal, and editing in

ways that are not possible in spontaneous, face-to-face communication and deceivers can use text communication to their advantage.”

Based on what they learned, Burgoon explained that her team developed a set of automated tools to detect deception in text messages.  This set of tools, dubbed Agent 99 (after the character from the “Get Smart” television show), has scored between 61 and 91 percent in judging truthful messages as the truth and between 61 and 88 percent in judging deceptive messages as deceit. 

Burgoon also discussed a form of video analysis based on motions made by an individual’s hands and face.  It can measure levels of agitation, and using an algorithm, differentiate between individuals telling the truth and those trying to deceive.

In conclusion, Burgoon asserted that “humans are very fallible at detecting deception and hostile intent, but we think we can develop automated tools to augment human judgment that can greatly increase detection’s accuracy by using them.”  This can be studied across cultures and developed as more research is done over time.

Cross Cultural Deception

Charles Bond, Professor of Psychology at Texas Christian University, focused his presentation on “International Deception.”  He noted that his research has focused on judgments of deception across cultures and that “international judgments of deception might in principle prevent future terrorist attacks.”  Bond also explained that the key questions for his research are whether there are universal cues of deception – “behaviors that liars display in every culture” – or if there is a good way to detect lies across cultures. 

To give some context, Bond stated that deceptive behaviors have primarily been studied in the U.S., Western Europe, Japan, and Australia.  He did the only study to date on deceptive behavior in the Arab world – it was done a long time ago involving Jordanians.  Bond has also done work about judgments of deception involving citizens of Jordan, the U.S., and India.  Individuals of all three nationalities were videotaped and other people from all three counties watched and were asked to judge whether they were watching a liar or a truth teller. 

Bond’s results were statistically significant:  people have an ability to detect lies told not only by people from their own country, but also “a highly significant ability to detect lies across cultures.”  In fact, he explained, people could detect lies across cultures when they were listening to a language that they couldn’t understand. 

Bond noted that his current work is a very large study with 88 collaborators on beliefs about deception in 75 different countries.  For the first part of this study, which involves 4,800 participants, people were asked “How can you tell if people are lying?”  Common beliefs (in response to this question) are that liars are inconsistent, they shift their posture often, their speech is somewhat incoherent (i.e. their speech contains “uh”s, “ah”s, and long pauses), and liars tend to show signs of nervousness.  “The most common belief about deception worldwide, however, is that liars can’t look you in the eye,” Bond explained.

The second part of the study involved responses to a questionnaire on beliefs about deception.  Again, the participants indicated that liars make less eye contact than truth tellers.  Other common responses emphasize that liars tell long stories and touch and scratch themselves and shift their posture more than usual.  The questionnaire also asked how many lies the typical person tells in a week and how accurate the participants thought they were at detecting lies.  Results varied widely across countries.

Bond concluded by asserting that “We should be researching deception in the countries of greatest interest to the U.S. security community.  Studies there might in principle prevent future terrorist attacks.”

Face-to-Face Intelligence Gathering

Mark Frank, Associate Professor of Communication at Rutgers, addressed “Practical Approaches to Detecting Deception in Counter-Terrorism.”  To introduce his topic, Frank told the audience that “Prior to September 11, 2001, human to human intelligence gathering had been seen as the weird uncle of the intelligence gathering world.”  With the “limits of technological approaches to gathering intelligence laid bare on September 11,” however, “there has been a renewed interest in gathering information from people in situations where two people meet face to face.” 

Obviously, it is vital for the agents collecting face-to-face intelligence to be able to judge whether they have been told the truth.  Frank explained that the key questions are how we know when we are being lied to and if there’s any technology that can assist in this assessment. 

In trying to answer these questions, Frank has taken care to look at situations involving “minimally implicit, but better explicit, rewards or punishments.” (i.e., not white lies)  This makes the studies more applicable to real counter-terrorism situations.  In addition, Frank noted that he has worked closely with individuals from counter-terrorism and intelligence fields in the planning, design, and interpretation of results. 

To determine how individuals can best detect lying, Frank conducted extensive studies of human expressive behavior.  Research has shown conclusively that certain muscles, including ones in the face, are “reliable muscles,” ones that are very hard to control.  These muscles can give away someone’s attempting to conceal his or her feelings.  Frank’s work has shown that a person’s ability to spot these micro-expressions is correlated with their abilities to spot high stake lies.

Frank also discussed individuals in his studies who have been able to detect deception at a rate of 90 percent or better.  He is working to determine why these people are particularly gifted and urged that those in law enforcement and related communities with this ability should be identified and utilized in critical interviews and situations. 

Touching on the technology question, Frank stated that the Facial Action Coding System (FACS), developed by Paul Ekman and Wally Friesen, has been used by over 300 scientists in about a dozen countries and can identify micro-expressions that even highly qualified and trained experts have missed.  The problem is that it can take up to three hours to code one minute of behavior into FACS.  Thus, developing a fast-working computer system capable of the same identification could be a key to detection efforts. 

Franks has been working towards this goal with help of an NSF grant and has been able to replicate “90 percent agreement for all upper face units.”  He noted that it is still a work in progress, but that it could be ready for field use within five years. 

To conclude, Frank mentioned that ongoing collaborations between law enforcement personnel and those in academe have proven successful and should be continued and expanded.  He urged renewed attention to and funding for these efforts.

COSSA will prepare edited transcripts of the seminar, which included a lively question and answer period.  These should be available by late May.  If you would like to request a copy, please e-mail cossa@cossa.org.