Examining the effects of information pollution using SDT

Psychological research on misinformation susceptibility often relies on experimental paradigms that use a series of isolated news headlines as its experimental stimuli (Kozyreva et al., 2024; Pennycook et al., 2021; Roozenbeek et al., 2024). However, in the real-world social media environments where people typically encounter these headlines, any given claim is encountered amidst other content - other posts, headlines, and images that are also present in the information environment. As such, the task of discerning truth from falsehood on social media does not involve just the task of discernment itself, but also the task of navigating through a variety of unrelated, unsolicited, and potentially distracting content. The term ‘information pollution’ has been used to refer to the proliferation of such content in information environments (Orman, 1984), also specifically with regard to the internet and social media (Hwang, 2020; Nielsen, 2003; Wardle & Derakhshan, 2017). Crucially, the presence of information pollution may interfere with people’s ability to distinguish true from false information. For instance, Pennycook & Rand (2022) argue that aimlessly scrolling through this type of content diminishes attention to accuracy when evaluating news. Experimental paradigms that test headlines in isolation do not account for such effects, even though there are various ways in which information pollution might affect information processing, both in quantitative (how much pollution there is) and qualitative (how disruptive the pollution is) terms (Shahrzadi et al., 2024). It is important to note here that developments in the real world can affect how ‘polluted’ social media environments are (Hwang, 2020), which in turn could have implications for people’s truth discernment on those social media platforms. For instance, the emergence of generative AI has increased information pollution along both dimensions: it enables the rapid production of content (high quantity) that is dubious and unreliable (low quality) (Barman et al., 2024; Zhou et al., 2023). Accounting for these types of information pollution in an experimental setting allows us to test theory-driven predictions that cannot be tested using paradigms that present headline stimuli in isolation. For instance, the Misinformation Receptivity Framework (MRF) by Zmigrod et al. (2023) predicts that navigating polluted information environments (which it refers to as ‘noisy’) makes people uncertain about the reliability of incoming information, resulting in increased reliance on prior beliefs when making veracity judgments. This increased reliance on prior beliefs makes people more susceptible to information that confirms their existing worldviews while making them more resistant to belief-incongruent information, regardless of the veracity of such information. To investigate this mechanism more precisely, we can draw upon Signal Detection Theory (SDT; Green & Swets, 1966). For previous misinformation research, SDT has proven valuable by allowing researchers to separate people's ability to accurately distinguish true from false information from their baseline response tendencies when evaluating news content (Batailler et al., 2022; Gawronski et al., 2023; Sultan et al., 2022). It does so by conceptualizing information processing as involving two distinct cognitive components: discrimination ability (d') – how well someone can distinguish between true and false information – and response criterion (c) – the bias towards responding "true" versus "false" regardless of signal quality. Crucially, when discrimination becomes difficult (low d'), people must rely more heavily on their response criteria (c), which can vary depending on (in)congruence with prior beliefs. Past research has shown that people have more liberal (in the sense of being more likely to respond ‘true’) response criteria for belief-congruent information, while also having more conservative (in the sense of being more likely to respond ‘false’) response criteria for belief incongruent information (Batailler et al., 2022; Gawronski et al., 2023). This framework can provide an account of how the MRF's predictions might unfold in practice. Information pollution could impair discrimination ability (decreasing d'), forcing greater reliance on response criteria. Information pollution should also lead to general skepticism, shifting people's overall response criterion (c) towards being more conservative. Prior beliefs can systematically bias this criterion, with people setting lower thresholds for belief-congruent information (lower ccongruent) and higher thresholds for belief-incongruent information (higher cincongruent). The difference between these criteria (Δc = ccongruent - cincongruent) provides a direct behavioral measure of how heavily people rely on prior beliefs when making veracity judgements in this context. In polluted information environments, when discrimination ability is limited, this difference should become larger as people compensate for uncertainty by relying more heavily on their existing beliefs.