COMMENTARY
One of the most powerful aspects of artificial
intelligence is that it can see things that humans can’t.
More precisely, it can analyze data and find
previously unseen correlations, some likely never to be discovered manually on
a recurring basis.
Now researchers are tapping AI to tell if a
person is lying, such as in a courtroom trial.
The system for covert automated deception
detection, named DARE (Deception Analysis Reasoning Engine), is an AI engine
that uses video features that detect human micro-expressions.
Researchers from the University of Maryland and
Dartmouth College used real-life courtroom trial videos. Their system used
classifiers trained on video features that predict human micro-expressions,
which they determined could be used to predict deception.
The researchers created a three-step process
involving capturing audio and video and then data encoding.
They trained their AI engine to identify the
five most predictive micro-expressions: frowning, eyebrows raising, lip corners
up, lips protruded and head side turn.
The researchers created a database of trial
video clips, with 50 truthful videos and 54 that were deceptive.
They then had 15 people watch the videos and the
DARE system ended up being more accurate than the people in predicting when
someone in the video was lying.
The AI engine detected deception at a rate of
92%.
Artificial intelligence can combine audio, video
and transcriptions in real-time to make a determination. It can see and process
more things simultaneously and conclude which statements are likely not true,
with a high degree of accuracy.
In the future, it may be difficult to lie to AI.
No comments:
Post a Comment