The Force is Strong with This One: IARPA Picks ‘Jedi Mind’ to Predict Intelligence Trustworthiness

Editor's note: This story previously misstated the challenge was launched following the Edward Snowden leaks. The story has been updated. 

What if members of the intelligence community could use a combination of physiological data and computer algorithms to anticipate behavior and gauge trustworthiness of individuals?

That was the public challenge put forth by the Intelligence Advanced Research Projects Activity back in February in an effort called INSTINCT, which stands for Investigating Novel Statistical Techniques to Identify Neurophysiological Correlates of Trustworthiness.

IARPA today announced the challenge winners – a duo from Massachusetts-based BAE Systems’ Adaptive Reasoning Technologies Group. The team developed JEDI MIND, a "Star Wars"-themed acronym, which stands for Joint Estimation of Deception Intent via Multisource Integration of Neuropsychological Discriminators.

Troy Lau and Scott Kuzdeba used a combination of “innovative statistical techniques” to improve predictions of human behavior by 15 percent over baseline analysis, according to an IARPA statement. The duo found a strong correlation between physiological characteristics, such as heart rate and reaction time, and predicting how likely their partner was to keep a promise.

In short, JEDI MIND indicates nonvoluntary physiological actions occur within the human body that might make one’s behavior, intentions or trustworthiness much easier to predict.

”We’re delighted with Lau and Kuzdeba’s insight into the data,” said Adam Russell, the TRUST program manager. “Their performance under the rigorous evaluation process of the INSTINCT Challenge provides additional evidence in support of one of the TRUST program’s basic hypotheses: that the self’s own, often nonconscious signals – if they can be detected and leveraged appropriately – may provide additional valuable information in trying to anticipate the intentions of others.” 

IARPA lauded Lau and Kuzdeba’s efforts as the only solvers among 39 contestants who were able to overcome analytic techniques for multimodal data sets that could “generalize well.” Over the 70-day challenge, solvers were asked to create algorithms based on training data and to submit those algorithms to be used against a test set of data. The final part entailed submitting a “best and final” algorithm for final evaluation against an unreleased data set.

JEDI MIND’s algorithms led to the most accurate predictions. In addition to the clout of winning IARPA’s first challenge, Lau and Kuzdeba will receive $25,000 in funding from IARPA. 

There are many potential applications for analytic and algorithmic techniques such as those under JEDI MIND, ranging from security clearance processes to gauging the trustworthiness of intelligence agents, analysts or potentially captured adversaries.