Extended Data Fig. 2 | Distribution of sharing intentions in studies 3 and 4, by condition and headline veracity. Whereas Fig. 2 discretizes the sharing intention variable for ease of interpretation such that all 'unlikely' responses are scored as 0 and all 'likely' responses are scored as 1, here the full distributions are shown. The regression models use these non-discretized values. Extended Data Fig. 3 | Distribution of sharing intentions in study 5, by condition and headline veracity. Whereas Fig. 2 discretizes the sharing intention variable for ease of interpretation such that all 'unlikely' responses are scored as 0 and all 'likely' responses are scored as 1, here the full distributions are shown. The regression models use these non-discretized values.

The spread of false and misleading news content on social media is of great societal concern. Why do people share such content, and what can be done about it? In a first survey experiment (N=1,015), we demonstrate a disconnect between accuracy judgments and sharing intentions: even though true headlines are rated as much more accurate than false headlines, headline veracity has little impact on sharing. Although this may seem to indicate that people share inaccurate content because they care more about furthering their political agenda than they care about truth, we propose an alternative attentional account: Most people do not want to spread misinformation, but the social media context focuses their attention on factors other than truth and accuracy. Indeed, when directly asked, most participants say it is important to only share news that is accurate. Accordingly, across four survey experiments (total N=3,485) and a digital field experiment on Twitter in which we messaged users who had previously shared news from websites known for publishing misleading content (N=5,379), we find that inducing people to think about the concept of accuracy increases the quality of the news they subsequently share. Together, these results challenge the narrative that people no longer care about accuracy. Instead, the results support our inattention-based account wherein people fail to implement their preference for accuracy due to attentional constraints. Furthermore, our research provides evidence for scalable anti-misinformation interventions that are easily implementable by social media platforms.

When faced with the chance to help someone in mortal danger, what is our first response? Do we leap into action, only later considering the risks to ourselves? Or must instinctive self-preservation be overcome by will-power in order to act? We investigate this question by examining the testimony of Carnegie Hero Medal Recipients (CHMRs), extreme altruists who risked their lives to save others. We collected published interviews with CHMRs where they described their decisions to help. We then had participants rate the intuitiveness versus deliberativeness of the decision-making process described in each CHMR statement. The statements were judged to be overwhelmingly dominated by intuition; to be significantly more intuitive than a set of control statements describing deliberative decision-making; and to not differ significantly from a set of intuitive control statements. This remained true when restricting to scenarios in which the CHMRs had sufficient time to reflect before acting if they had so chosen. Text-analysis software found similar results. These findings suggest that high-stakes extreme altruism may be largely motivated by automatic, intuitive processes.

The recent emergence of machine-manipulated media raises an important societal question: How can we know whether a video that we watch is real or fake? In two online studies with 15,016 participants, we present authentic videos and deepfakes and ask participants to identify which is which. We compare the performance of ordinary human observers with the leading computer vision deepfake detection model and find them similarly accurate, while making different kinds of mistakes. Together, participants with access to the model’s prediction are more accurate than either alone, but inaccurate model predictions often decrease participants’ accuracy. To probe the relative strengths and weaknesses of humans and machines as detectors of deepfakes, we examine human and machine performance across video-level features, and we evaluate the impact of preregistered randomized interventions on deepfake detection. We find that manipulations designed to disrupt visual processing of faces hinder human participants’ performance while mostly not affecting the model’s performance, suggesting a role for specialized cognitive capacities in explaining human deepfake detection performance.

Recent research suggests that shifting users’ attention to accuracy increases the quality of news they subsequently share online. Here we help develop this initial observation into a suite of deploy-able interventions for practitioners. We ask (i) how prior results generalize to other approaches for prompting users to consider accuracy, and (ii) for whom these prompts are more versus less effec-tive. In a large survey experiment examining participants’ intentions to share true and false head-lines about COVID-19, we identify a variety of different accuracy prompts that su¬ccessfully increase sharing