Saturday, November 12, 2011

Lie Detection Software: Here at Last!

Imagine if a rigorous lie detection software became available and could not only expose the lies of current politicians, but also sift truth from fiction in statements from historical characters, presumed involved in historical crimes. Such a find would be monumental to say the least, and easily replace current lie detectors which had their shortcomings well documented over the last 40 years or so.

Now, the vision may soon be a reality with a lie catching software developed by two Stevens Institute of Technology professors, Rajarathnam Chandramouli and Koduvayur Subbalakshmi. The core concept behind the software, which is lingual as opposed to electrical signal- based, isn't novel by any means. It goes back to the old "Freudian slip" and the fact that no human brain, no matter how well disciplined or trained, can retain a liar's matrix over time. Ultimately, the brain circuits fry and the slip emerges!

In like manner, this new software parses liar's lingual predilections by combing language clues and statements for no fewer than 88 psycho-linguistic cues that indicate whether the subject is open or trying to cover something up.

As with all software, it was beta-tested, using some 1,000 known email hoaxes out of As a result, an 86-99% success rate emerged by sifting through and identifying the falsehoods and hoaxes. This compares to a 60-65% success rate for lie detectors.

In addition, it was ex post facto tested on statements made in the past by both still living and dead presumed perps. It parsed Casey Anthony's statement to the police about her missing daughter, and arrived at the verdict 'lie'. It also parsed Lee Harvey Oswald's now famous statement to the Dallas cops, when asked if he had killed John F. Kennedy, and he replied:

"No, I have not been charged with that. In fact, nobody has said that to me yet. The first thing I heard about it was when the newspaper reporters in the hall asked me that question...I did not do it. I did not shoot anyone. I am just a patsy!"

Verdict of the software: TRUTH

There are also cross-checks possible here!

In a previous blog to do with Stephen King's new novel, '11/22/63', I already examined multiple ways that Oswald could not have been guilty, i.e.:

There were, in addition, reports of an "Oswald double" in Dallas seen after the assassination. Air Force Sgt. Robert Vinson, on Nov. 22, was trying to get home to Colorado Springs. He took a bus to Andrews AFB in Washington, DC, and got on a C-54 (the first available flight to "the vicinity" of the Springs is what he wanted) and was informed it was to depart for Lowry AFB near Denver.

What Vinson recalled most about the particular flight, is that it had no crew chief or manifest. (Usually the crew chief always asked him to sign a manifest). The flight itself didn't land at Lowry but in Dallas, as Vinson clearly made out the skyline. Vinson noted a lot of dust blew up as the craft landed near the Trinity River.

Two men boarded, just after 3:30 p.m. central time, one a "Latino", 6' or 6'1" weighing 180-190 lbs. and wearing a mustache, the other 5'7" to 5'9" and Caucasian, 150-160 lbs. Only later when press photos materialized did Vinson realize the latter guy was the spitting image of Lee Harvey Oswald. But how could Oswald have boarded this plane while in Dallas police custody? The Oswald Double aspect of the assassination then materialized. Subsequently, other witnesses of this Oswald emerged, as documented by James Douglass in his book, JFK and the Unspeakable (2008, Orbis Books, p. 335).

Subsequently, putative images of the double surfaced (see inset images at top of center graphic) and these were photo analyzed with the images of the actual Oswald. One startling find - shown to the right of the graphic - is that the facial images didn't match even when integrated at the chin. The real Oswald was at least 2½ inches taller than the double. All of these in concert bolster the software finding that Oswald told the truth in the Dallas PD station.

Some of the key aspects of the software's lie parsing, after a subject is asked point blank questions, are based on:

1) Words longer than eight letters. The reason is that such words are not normally used in day to day conversation, so people who use them could be trying too hard to sound authentic or invoke a gravitas they don't otherwise have.

2) A lack of personal pronouns, me, myself, I:

In deceptive responses, the subject usually attempts to dissociate himself from his own words, and thus more often employs passive speech if he isn't present or the active agent.

3) Excessive use of 'you' references:

This is often an effort to deflect attention from the liar toward the person he's attempting to dupe.

Another benefit of the software arising out of its beta tests is that it can discriminate the gender of anonymous text senders as well. For example, males are more apt to use 'I' statements. Females also often use adverbs to try to intensify their statements (i.e. appending 'very' to them).

Can this lie detection software be the answer to confronting a society largely based on prevarication? Maybe, but for sure more tests need to be done, especially on real life subjects in real time. While I am convinced ex post facto tests on historical events and subjects disclose it to be of some significant value, the proof is in the pudding of real life crimes, and transgressions.

Maybe we can have it in place to parse the claims of each of the Republican candidates at their next 14 debates?

No comments: