HOW ACCURATE IS A LIE DETECTOR?
You may have heard or read that some say polygraph tests are not accurate, while others say they are. The vast majority make their statements without referring to the research that exists. For example, if you read on Wikipedia about lie detector in several different languages, you will find that different accuracy rates are stated for lie detector tests in almost all languages. In the vast majority of places there is no reference to research results on lie detectors.
One of the reasons is that the research on lie detector tests has not been readily available. American authorities and industry organizations previously wanted to keep the methods and techniques of various lie detector tests secret. Today it is not like that.
What many people are not aware of is that there is relatively a lot of research on lie detector tests. After 2002, there was a major shift within the field, where, among other things, the American Polygraph Association and American authorities' research on lie detector tests decided to have to follow scientific criteria for peer review and publication in scientific journal.
There are many different types of lie detector tests. We generally distinguish between the methods "recognition tests" and "lie tests". Under each of the two methods there are subcategories, and within each subcategory there are lie detector test techniques. The most commonly used lie detector tests are tests approved by the American Polygraph Association and the standards society ASTM. Each technique is different, and each technique achieves different accuracy in scientific studies.
Here follows an overview from a meta-study published in 2012 on the accuracy of techniques approved by members of the American Polygraph Association. The average accuracy of the 5 most accurate tests is 95.76%, and 8 techniques achieve higher than 90% accuracy. Research published after 2012 shows fairly similar accuracy for the same methodologies as the accuracy published in the study from 2012. The average accuracy for IZCT is currently about 97%, against 99.4% in the meta-analysis from 2012.
Source: APA Meta-Analytic Survey, 2012
CD = Correct decisions. INC = Inconclusive (Tests where data were not clear enough to conclude that the person being tested lied or answered truthfully).
For example, when it says "CD = .994, it means that 99.4% of the tests have concluded correctly.
The accuracy of the lie detector techniques in the overview above is calculated from 11,737 analyzes of lie detector tests, which were analyzed by 295 different polygraphists. The studies that have been carried out have been published in peer-reviewed scientific journals.
In all of these 11,737 lie detector tests, the "ground truth" is confirmed. In other words, the researchers have been able to confirm with other reliable means of proof that the person who was tested with a lie detector either lied or answered truthfully in the lie detector test (For example, by the police having found DNA or video evidence after the lie detector test_cc781905-5cde -3194-bb3b-136bad5cf58d_was executed).
The more and better data we have from a test, the greater the probability that the test will show correctly. In all the scientifically published studies included in the meta-analysis from 2012, the criterion is that only 3 series of questions should be included in the analysis (to standardize and enable relevant comparisons). In daily use, we use 3-5 series of questions that can be included in an analysis. This means that we get more data, which both increases accuracy and reduces the proportion of tests with "inconclusive" answers.
After 2012, several scientific studies have been published that strengthen the findings from the meta-analysis published in 2012. In addition, further development has taken place for 3 of the analysis tools we use. Both the Horizontal Scoring System (HSS), the Empirical Scoring System (ESS), and the Objective Scoring System (OSS3) have implemented changes that increase accuracy and reduce the proportion of tests with inconclusive answers.
Our tests usually achieve an accuracy in the range of 97 - 99.9%, and we believe this is mainly because we spend a lot of time - and more time than many other polygraphists on making sure that the questions are as good as possible. In addition, we have a strong focus on exposing attempts at cheating.
In addition to the polygraph test, we use other tests that can be combined with the polygraph test, and which are methodologically independent of the polygraph test. We are currently one of the few companies in the world that can offer such combination tests.
It is entirely possible to achieve statistically expected accuracy in individual tests that is significantly higher than 99.9% for conclusive tests when we use combination tests.
.