Scholarlink

<p>Drawing on an online experiment with a nationally representative sample of U.S. adults (<i>N</i> = 948), this study examined how individuals’ experience with AI and self-assessed AI literacy, spanning perceived knowledge, capabilities, and evaluations, were associated with their attribution of AI authorship to college admissions essays. The study focused on the attributional heuristics people apply, such as writing style or references to personal experience, when inferring whether a piece of content was generated by AI. Results show that participants with greater AI experience were more likely to use experience-related cues, while those with higher subjective knowledge of AI relied on all content heuristics, including those previously shown to be unreliable indicators of perceived AI authorship, potentially exhibiting a confidence–competence mismatch (i.e., Dunning–Kruger Effect). Additionally, people with higher ethical concerns about AI are less likely to attribute writings to AI, regardless of content heuristics. These findings highlight the role of individual differences in AI authorship attribution, reveal limitations of human judgment in high-stakes contexts, and underscore the importance of critical AI literacy and value-based biases in public perceptions of generative AI.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

The more you think you know: AI experience, literacy, heuristics, and perceived AI authorship

  • 期刊论文

  • 发表日期：

    2025年9月27日

查看全文
  • Sixiao Liu,
  • Haoran Chu,
  • Shupei Yuan,
  • Yuan Sun,
  • Linjuan Rita Men

摘要

Drawing on an online experiment with a nationally representative sample of U.S. adults (N = 948), this study examined how individuals’ experience with AI and self-assessed AI literacy, spanning perceived knowledge, capabilities, and evaluations, were associated with their attribution of AI authorship to college admissions essays. The study focused on the attributional heuristics people apply, such as writing style or references to personal experience, when inferring whether a piece of content was generated by AI. Results show that participants with greater AI experience were more likely to use experience-related cues, while those with higher subjective knowledge of AI relied on all content heuristics, including those previously shown to be unreliable indicators of perceived AI authorship, potentially exhibiting a confidence–competence mismatch (i.e., Dunning–Kruger Effect). Additionally, people with higher ethical concerns about AI are less likely to attribute writings to AI, regardless of content heuristics. These findings highlight the role of individual differences in AI authorship attribution, reveal limitations of human judgment in high-stakes contexts, and underscore the importance of critical AI literacy and value-based biases in public perceptions of generative AI.