Toward Interfaces that Help Users Identify Misinformation Online: Using fNIRS to Measure Suspicion

Abstract

With terms like ‘fake news’ and ‘cyber attack’ dominating the news, skepticism toward the media and other online individuals has become a major facet of modern life. This paper views the way we process information during HCI through the lens of suspicion, a mentally taxing state that people enter before making a judgment about whether or not to trust information. With the goal of enabling objective, real-time measurements of suspicion during HCI, we describe an experiment where fNIRS was used to identify the neural correlates of suspicion in the brain. We developed a convolutional long short-term memory classifier that predicts suspicion using a leave-one-participant-out cross-validation scheme, with average accuracy greater than 76%. Notably, the brain regions implicated by our results dovetail with prior theoretical definitions of suspicion. We describe implications of this work for HCI, to augment users’ capabilities by enabling them to develop a ‘healthy skepticism’ to parse out truth from fiction online.