SWSterling Williams-Ceci

Sterling Williams-Ceci

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Sterling Williams-Ceci has a total of 27 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Humanlike AI May Strengthen Women’s Agreement With Sexist Stereotypes

    by Sterling Williams-Ceci, Rachel Minkowitz, Lior Zalmanson, Michael Macy, Mor Naaman

    AI can influence societally-consequential attitudes by generating biased arguments – but can it influence our attitudes without changing its arguments at all? In four experiments (N = 2,762), we show that the mere presence of humanlike features in an AI chatbot influences women to more strongly accept ideologically-compatible sexist stereotypes, biases that often resist influence and have societal harm. This influence operates indirectly by increasing “mind perception” – the belief that AI is conscious and thinking – which increases perceptions of its trustworthiness. Our contribution is that anthropomorphism’s influence extends beyond increased trust toward AI itself, impacting women’s broader attitudes about their competence and status. This influence happens when women are attentive during conversations with AI. As developers add more anthropomorphic features to AI, the harms of stereotypes it repeats will increase for vulnerable groups – but making AI appear less humanlike can mitigate this risk.

  • Open Access

    AI Overviews Reduce Search Result Engagement & May Influence Attitude Polarization

    by Saloni Dash, Yiwei Xu, Lei Cai, Sterling Williams-Ceci, Wang Liao, Emma Spiro

    AI-generated overviews (AIOs) increasingly appear at the top of search result pages, making them the first, and potentially only, content users encounter during search. Despite their widespread adoption, their impact on information seeking, processing, and downstream attitudes remains critically underexplored. Across two preregistered experiments (N = 2,797), we randomly assigned participants to view AIOs on one of four contested topics manipulated to be attitude-congruent, attitude-incongruent, or balanced (only Study 2) relative to their prior attitudes. Participants who viewed an AIO, irrespective of content, were significantly less likely to engage with subsequent search results than control (no AIO). Participants perceived attitude-incongruent AIOs as more biased and weaker than attitude-congruent AIOs. Despite this, attitude-incongruent AIOs reduced attitude polarization relative to attitude-congruent AIOs in Study 1, but not in Study 2 (although directionally consistent). Taken together, these findings suggest that AIOs significantly influence information seeking and processing and raise critical concerns for democratic deliberation.

  • Top Cited Papers
    Open Access

    Biased AI writing assistants shift users’ attitudes on societal issues

    by Sterling Williams-Ceci, Maurice Jakesch, Advait Bhat, Kowe Kadoma, Lior Zalmanson, Mor Naaman

    Artificial intelligence (AI) writing assistants powered by large language models (LLMs) are increasingly used to make autocomplete suggestions to people as they write text. Can these AI writing assistants affect people’s attitudes in this process? In two large-scale preregistered experiments ( N = 2582), we exposed participants writing about important societal issues to an AI writing assistant that provided biased autocomplete suggestions. When using the AI assistant, the attitudes participants expressed in a posttask survey converged toward the AI’s position. However, a majority of participants were unaware of the AI suggestions’ bias and their influence. Further, the influence of the AI writing assistant was stronger than the influence of similar suggestions presented as static text, showing that the influence is not fully explained by these suggestions, increasing accessibility of the biased information. Last, warning participants about assistants’ bias before or after exposure does not mitigate the attitude-shift effect.


  • Bring Wisdom into AI Education (Before it’s Too Late)

    by Sterling Williams-Ceci, Wendy M. Williams

  • Open Access

    Biased AI Writing Assistants Shift Users’ Attitudes on Societal Issues

    by Sterling Williams-Ceci, Maurice Jakesch, Advait Bhat, Kowe Kadoma, Lior Zalmanson, Mor Naaman

    AI writing assistants powered by Large Language Models (LLMs) are increasingly used to make autocomplete suggestions to people as they write. Can these AI writing assistants impact people’s attitudes in this process? In two large-scale preregistered experiments (N=2,582), we exposed participants writing about important societal issues to an AI writing assistant that provided biased autocomplete suggestions. When using the AI assistant, the attitudes participants expressed in a post-task survey converged towards the AI’s position. Yet, a majority of participants were unaware of the AI suggestions’ bias and their influence. Further, the influence of the AI writing assistant was stronger than the influence of its suggestions when presented as static text, showing that the influence is partially due to AI writing assistants leading people to incorporate biased viewpoints into their own writing. Lastly, warning participants about assistants’ bias before or after exposure does not mitigate the influence effect.


  • Open Access

    Perceptions of AI as a Misinformation Moderator: The Roles of Argument Type and Group Chat Size

    by Benjamin Silver, Sterling Williams-Ceci, Mor Naaman

    Content moderators aim to prevent misinformation and abuse in online communities. Human moderators have limitations like limited capacity that can potentially be addressed by AI moderators. However, in order for AI to be effective as a moderator, we need to understand how people perceive an AI moderator compared to a human moderator. We report two experiments (N = 698) that showed a human or AI moderating misinformation in online groups, varying the moderator's arguments and the group's size. In Experiment~1, AI was seen as less effective when it used harm-based arguments to refute misinformation; however, this gap disappeared a year later in Experiment~2, perhaps as attitudes towards AI shifted. Perceptions of AI moderator effectiveness were unaffected by the moderated group's size. We suggest that there may be growing acceptance of AI as content moderators in online groups, and highlight some of the potential challenges their use may raise.

  • Open Access

    Bring Wisdom into AI Education (Before It’s Too Late)

    by Sterling Williams-Ceci, Wendy M. Williams

    Generative Artificial Intelligence (AI) is a booming part of the technology industry, and college students majoring in disciplines such as Computer and Information Sciences are increasingly pursuing lucrative jobs to develop AI applications for major companies. However, research is unveiling harmful effects of AI products, which limit people’s freedom of thought. AI developers are largely failing to anticipate these harms as companies incentivize quick rollouts and profitable products. We review the literature revealing AI’s cognitive harms, arguing that they are partially caused by students failing to exercise wisdom once they join the workforce. We propose strategies for STEM educators to encourage the development of STEM wisdom in their students, modeled after Sternberg’s (2024c) 8-step process of ethical reasoning. This article provides a practical guide to educators, with actionable recommendations to prevent further downstream harms of AI products their students build.

  • Open Access

    Misinformation does not reduce trust in accurate search results, but warning banners may backfire

    by Sterling Williams-Ceci, Michael W. Macy, Mor Naaman

    People rely on search engines for information in critical contexts, such as public health emergencies–but what makes people trust some search results more than others? Can search engines influence people’s levels of trust by controlling how information is presented? And, how does the presence of misinformation influence people’s trust? Research has identified both rank and the presence of misinformation as factors impacting people’s search behavior. Here, we extend these findings by measuring the effects of these factors, as well as misinformation warning banners, on theperceived trustworthinessof individual search results. We conducted three online experiments (N = 3196) using Covid-19-related queries, and found that although higher-ranked results areclickedmore often, they are not moretrusted. We also showed that misinformation does not damage trust in accurate results displayed below it. In contrast, while a warning about unreliable sources might decrease trust in misinformation, it significantly decreases trust inaccurateinformation. This research alleviates some concerns about how people evaluate the credibility of information they find online, while revealing a potential backfire effect of one misinformation-prevention approach; namely, that banner warnings about source unreliability could lead to unexpected and nonoptimal outcomes in which people trust accurate information less.