Silent Talker was invented between 2000 and 2002 by a team at Manchester Metropolitan University, Zuhair Bandar, James O'Shea, David McLean and Janet Rothwell. Following its invention, the Silent Talker Adaptive Psychological Profiling architecture and its specific instantiation as a lie detector, were patented internationally. In the interim, the inventors have been involved in raising investment funding and the code has been ported to various programming languages and speeded up from near real-time to real-time response. Current research includes adapting the technology to the measurement of comprehension amongst participants giving informed consent to take part in clinical trials Silent Talker Limited was incorporated on 9 April 2015 to commercialize this technology worldwide.
Testing procedure
The subject of the interview is observed by one or more cameras, which input the video stream to a conventional computer. As the interview takes place, Silent Talker's model of truthful vs. deceptive behaviour is used to classify the answers to the questions as truthful or deceptive in real-time. This can be as a classification at the end of the answer to a question or as a continuous monitoring stream during the interview. No calibration is required to tune the system to individuals and no training of the interviewer is required to interpret the Silent Talker classifications.
Countermeasures
As other lie detectors detect changes in stress, the most common approach is to disrupt this either by practicing calming techniques during lying or artificially raising stress during truth-telling. Because Silent Talker is based on a multi-factor model including cognitive load, duping delight and behaviour control, its inventors claim that it is robust to countermeasures. In fact it is believed that because a large number of channels are used, attempts at behaviour control will generate more incongruities between channels which can be detected. Further experimental trials are required to investigate this hypothesis.
Challenges
Challenges with this technological approach include:
it relies on only one channel, the face; excludes body language, voice, verbal content, verbal style and psychophysiology
it concludes from data without hypothesizing; operators play no role in the decision making
it relies on an intrusive camera within a few metres of the face
it doesn't analyse what the person is saying so expression can be correlated with the account.
