Pupil size predicts the level of trust or mistrust in news consumption

Introduction: Over the last few years, the rise of social media has been accompanied by an increase in the spread of fake news. It has been suggested that the human predisposition to seek information aligning with pre-existing beliefs, i.e., confirmation bias, is one of the main mechanisms underlying misinformation [1]. CBias contributes to maintaining maladaptive beliefs, even in the face of contrary evidence. However, many questions remain open. In particular, maintaining maladaptive beliefs challenges classic reinforcement learning theories, emphasising external feedback as the drive of our behaviours [2]. Here, we sought to understand the role of Confirmation Bias in reinforcement learning within contexts related to news consumption. First, we mapped participants’ prior beliefs by asking them to make binary judgments about the truthfulness of news headlines and wager virtual money on their own judgements to gain a measure of confidence. Second, participants were asked to choose one out of two previously judged headlines to gain a reward (two-armed bandit task). Finally, they were asked to confirm or disconfirm their initial judgments. Methods: 31 healthy volunteers (m.age = 23.7(4.2); M:10, F:21) were tested. In the mapping phase, d' and c indexes were analysed with one-sample T-tests, to disentangle the ability to distinguish between real and fake news from the individual tendency to judge them regardless of their veracity. Learning rates during the two-armed bandit task were analysed through repeated measures ANOVA, as a function of reward probability, which, in different blocks of trials, was linked to news subjective truthfulness, or rather to individual’s confidence score. Finally, using paired t-tests, we quantified participants’ belief updating as the measure of opinion changes during the last experimental phase. Pupil dilation was recorded throughout the whole experiment and analysed using permutation tests. Results: We found that Confirmation Bias affected reinforcement learning as participants showed higher accuracy and learning rates only when rewards were likely associated with news judged as real with high confidence scores. Notably, the association between rewards and confidence scores induced large proportions of opinion changes as regards news veracity. Pupil dilation predicted both the outcomes of veracity judgments and confidence strength, while during the learning task, pupil size was associated with rewarded choices and reward evaluation. Finally, variations in pupil size also marked opinion changes as regards the likely truthfulness of news judgments. Discussion: Our study confirmed the impact of Confirmation Bias on reinforcement learning [3]. Interestingly, in the context of news consumption, confirmation bias seems tied more closely to the perceived truth of a headline rather than to the level of confidence an individual holds in his own judgment. On the other hand, we found that it was possible to induce people to consider the contents of information critically and eventually change their beliefs by associating reward with individual confidence ratings rather than truthfulness judgments. Such a result provides a clue for contrasting the spread of misinformation. Results from pupil dilation complemented behavioural results, revealing higher neurophysiological activation in correspondence of larger confirmatory bias.