Abstract:

Large Language Models (LLMs) and other generative AI systems have demonstrated unprecedented capabilities in producing fluent and persuasive text. However, these systems are prone to hallucination, generating factually incorrect, fabricated, or misleading content with high confidence. While technical literature has focused extensively on detecting and mitigating hallucinations, the organizational and academic consequences of uncritically adopting such outputs remain underexplored. This study proposes to examine how hallucinations influence decision-making within organizational settings and compromise academic integrity through automation bias and misplaced trust. Employing a mixed-methods design, the research will combine controlled experiments, organizational case vignettes, and qualitative interviews with decision-makers and academics. The goal is to develop the Hallucination Impact Framework (HIF), a socio-technical model connecting hallucination phenomena, human cognitive biases, and institutional outcomes. This work will provide both theoretical contributions to human–AI interaction research and practical guidelines for mitigating risks in professional and academic domains.

Generative AI has rapidly diffused into organizational workflows and academic practices, offering productivity gains and novel capabilities. However, these benefits are counterbalanced by risks associated with hallucinations outputs that are syntactically plausible but semantically false (Huang et al., 2023; Sahoo et al., 2024). Such errors have already resulted in serious consequences, including fabricated legal citations submitted in U.S. courts (Washington Post, 2025) and erroneous medical references (Gravel et al., 2023). Beyond isolated incidents, hallucinations represent a systemic challenge to trust, transparency, and reliability in decision-making contexts.

Although technical research has advanced in detecting and mitigating hallucinations (Bai et al., 2024; Zhang et al., 2025), little is known about how hallucinations affect organizational judgment processes and academic integrity. This study seeks to address this gap by systematically examining the cognitive and institutional pathways through which hallucinations shape human decisions, and by developing evidence-based strategies to mitigate their downstream impact.

Keywords:

Hallucination, Education, Decision-Making, Academic Integrity.