Social hierarchies are fundamental in determining the amount of power allotted to identity groups. Currently, we examine the amount of power different sexual identity groups are perceived to be allotted by society versus how much power participants indicate each group should ideally have. Across 2 related studies conducted in 2024, LGBTQIA+ and cishet participants first rank-ordered six (Study 1b) or eight (Study 1a) sexual identity groups into a social hierarchy. Next, participants were asked to assign power to each group based on (a) how much power they feel society gives each group and (b) how much power they feel that each group should ideally have in society. We found differences between how LGBTQIA+ and cishet participants felt power should ideally be distributed and how it actually is distributed. LGBTQIA+ participants, compared with cishet participants, believed that power should be more equally allocated across sexual identity groups, and that society gives significantly more power to those identifying as straight. These findings provide insight into the differences between hierarchy within and outside of the LGBTQIA+ community and increase visibility of marginalized sexual identities. Public Significance Statement This study suggests that cisgender heterosexual individuals seek to maintain their dominant position in their perceived sexual identity hierarchy, whereas LGBTQIA+ individuals advocate for equalizing power. These findings are crucial for informing efforts to empower LGBTQIA+ identities despite their marginalized position and resistance from those in power. Addressing these disparities requires shifting cishet perspectives on power and privilege to foster societal change toward equity.
With implications for investor behavior and perceptions of risk, social media has transformed how individuals obtain financial information. Using restricted 2021 National Financial Capability Study (NFCS) data and a three-step empirical investigation, we find social media use is positively related to fear of financial investment fraud, with financial education positively moderating this relationship. Contrary to expectations, the fear-increasing effect is stronger for those with a financial education background. We propose a “knowledge dilution effect” to explain this paradox: exposure to misinformation and low-credibility financial content on social media may undermine the confidence and protective benefits typically conferred by financial education, amplifying anxiety about fraud. These findings highlight an urgent need to address the quality of online information and to supplement financial education with digital literacy and fraud awareness initiatives in the digital age.
Concerns that AI homogenizes human thinking appear at odds with findings that LLM writing is perceived as more creative than human writing. We propose that LLMs enhance superficial semantic diversity while simultaneously homogenizing underlying ideas. Four large-scale natural experiments tested this preregistered “disjunctive homogenization” hypothesis in 372,793 personal statements written in high-stakes college admissions contexts. Comparing before-versus-after ChatGPT’s release revealed increasing word-level diversity simultaneous with decreasing conceptual diversity at sentence and whole-document levels. Controlled experiments provided causal evidence of AI’s influence and identified a plausible mechanism: a positive association of word-to-idea diversity in humans was reversed in GPT. Despite conceptual homogenization, raters perceived post-ChatGPT essays as more creative due to increased word-level diversity, even when considering multiple essays. Disjunctive trends were strongest among minoritized applicants.
We examined whether the behavior of a witness at the time of a lineup procedure could help to sort accurate from inaccurate witnesses over and above confidence and decision time. We videorecorded witness-participants (N = 1,496) as they completed lineups and then trained coders scored them on numerous behavioral manifestations of recognition strength and fluency (how easy the witness found the lineup task). We then used machine learning to determine which of these behaviors, if any, distinguished accurate witnesses from inaccurate witnesses. As predicted, both recognition strength and fluency proved informative. Whereas witnesses who made accurate decisions appeared to have a strong recognition experience and completed the lineup with ease, witnesses who made inaccurate decisions appeared to have a weak recognition experience, and completed the lineup with difficulty. Lineup behaviors discriminated accurate from inaccurate witnesses over and above confidence and decision time. In fact, the behavior of the witness at the time of the lineup also diagnosed high-confidence mistaken identifications. These mistaken witnesses expressed high confidence despite appearing disfluent during the lineup procedure. Of course, the only way to create a complete and veridical record of everything that transpired during a lineup procedure is by videorecording the lineup in its entirety. For the approximately 80% of police agencies that do not mandate the videorecording of lineups, this valuable information is lost (Police Executive Research Forum, 2013). These findings establish an evidence-based imperative to videorecord lineup procedures.
