No abstract

Recent studies have shown that the spread and consumption of misinformation online can be attributed to errors in human decision making, facilitated by cognitive biases. The field of Behavioral Economics has contributed a repertoire of such cognitive biases that can be leveraged for the design of technological interventions. In particular, the concept of nudging refers to subtle changes in the 'choice architecture' that can alter people's behaviors in predictable ways. In this paper we present our ongoing work on the design of nudging interventions in the context of misinformation, including a systematic review of the use of nudging in HCI that has led to a design framework consisting of 23 mechanisms of nudging tapping to 15 different cognitive biases, the translation of this framework into a set of design cards, the Nudge Deck, and its use in a planned workshop that aims to explore the design space of misinformation in the context of nudging.

Social media have become online spaces where misinformation abounds and spreads virally in the absence of professional gatekeeping. This information landscape requires everyday citizens, who rely on these technologies to access information, to cede control of information. This work sought to examine whether the control of information can be regained by humans with the support of a co-created browser plugin, which integrated credibility labels and nudges, and was informed by artificial intelligence models and rule engines. Given the literature on the complexity of information evaluation on social media, we investigated the role of technological, situational and individual characteristics in “liking” or “sharing” misinformation. We adopted a mixed-methods research design with 80 participants from four European sites, who viewed a curated timeline of credible and non-credible posts on Twitter, with (n=40) or without (n=40) the presence of the plugin. The role of the technological intervention was important: the absence of the plugin strongly correlated with misinformation endorsement (via “liking”). Trust in the technology and technology acceptance were correlated and emerged as important situational characteristics, with participants with higher trust profiles being less likely to share misinformation. Findings on individual characteristics indicated that only social media use was a significant predictor for trusting the plugin. This work extends ongoing research on deterring the spread of misinformation by situating the findings in an authentic social media environment using a co-created technological intervention. It holds implications for how to support a misinformation-resilient citizenry with the use of artificial intelligence-driven tools.