Scholarlink

<p>The pervasive threat of online disinformation challenges the integrity of the digital public sphere and the resilience of liberal democracies. This study conceptualizes and evaluates an explainable artificial intelligence (XAI) artifact specifically designed for disinformation detection, integrating confidence scores, visual explanations, and detailed insights into potentially misleading content. Based on a systematic empirical literature review, we establish theoretically informed design principles to guide responsible XAI development. Using a mixed-method approach, including qualitative user testing and a large-scale online study (<i>n</i> = 344), we reveal nuanced findings: while explainability features did not inherently enhance trust or usability, they sometimes introduced uncertainty and reduced classification agreement. Demographic insights highlight the pivotal role of age and trust propensity, with older users facing greater challenges in comprehension and usability. Users expressed a preference for simplified and visually intuitive features. These insights underscore the critical importance of iterative, user-centered design in aligning XAI systems with diverse user needs and ethical imperatives. By offering actionable guidelines and advancing the theoretical understanding of explainability, this study contributes to the development of transparent, adaptive, and effective solutions for disinformation detection in digital ecosystems.</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

Explainable AI for online disinformation detection: Insights from a design science research project

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年7月16日

查看全文
  • Isabel Bezzaoui,
  • Carolin Stein,
  • Christof Weinhardt,
  • Jonas Fegert

摘要

The pervasive threat of online disinformation challenges the integrity of the digital public sphere and the resilience of liberal democracies. This study conceptualizes and evaluates an explainable artificial intelligence (XAI) artifact specifically designed for disinformation detection, integrating confidence scores, visual explanations, and detailed insights into potentially misleading content. Based on a systematic empirical literature review, we establish theoretically informed design principles to guide responsible XAI development. Using a mixed-method approach, including qualitative user testing and a large-scale online study (n = 344), we reveal nuanced findings: while explainability features did not inherently enhance trust or usability, they sometimes introduced uncertainty and reduced classification agreement. Demographic insights highlight the pivotal role of age and trust propensity, with older users facing greater challenges in comprehension and usability. Users expressed a preference for simplified and visually intuitive features. These insights underscore the critical importance of iterative, user-centered design in aligning XAI systems with diverse user needs and ethical imperatives. By offering actionable guidelines and advancing the theoretical understanding of explainability, this study contributes to the development of transparent, adaptive, and effective solutions for disinformation detection in digital ecosystems.