Institutional Repository of Vilnius University
Human vs ai-generated metaphors: a case study of illness and death /
Abstract
dc:descriptionThis paper aims to investigate how humans and AI use conceptual metaphors in the light of euphemistic topics. It also distinguishes whether people can confidently and accurately identify metaphors generated by humans when juxtaposed to an AI variant, and what factors influence their judgment. The study analyses a set of metaphors on sensitive topics such as depression, cancer, and death, collected from 154 human participants and 6 large language models. Using a thematic categorization, the metaphors were evaluated for common patterns and then tested in a survey where respondents were asked to identify the human text, mark the confidence level and justification of their choices. The results reveal similar human and AI use of metaphor categories for depression, cancer, and death, but human-created metaphors show greater emotional depth, vividness, and creativity. Moreover, participants were often confident and right in their choices, with occasional difficulties. Emotional tone, linguistic vividness, and perceived authenticity were frequently cited as reasons behind their decisions. However, these judgments did not consistently correspond to the actual authorship of the metaphors, suggesting that perceived emotional depth is not a reliable indicator in humans. The findings contribute to a deeper understanding of how humans versus AI metaphorically describe sensitive topics and how humans tackle metaphors when their authorship is unclear. The study offers insights into linguistic intuition, trust in AI-generated content, and the blurred boundaries of creativity in the age of technology.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2025
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Medutytė, Ema,
Subjects
dc:subject × 1Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- eng
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD239013897&prefLang=en_US