“Between the Lines”: Perceptions of Poetry With Authorship Attributed to Artificial Intelligence or Humans – A Comparative Analysis

In the rapidly evolving field of artificial intelligence (AI) literature generation, understanding how society perceives AI-generated content, compared with human-produced literature is of paramount importance. This study investigated societal perceptions and biases toward AI-generated versus human-produced poetry. A sample of 123 participants was subjected to a controlled experiment in which they evaluated a human-generated poem that was randomly attributed to either a human, an AI, or an unspecified author. The assessment metrics comprised five categories: originality, aesthetic appeal, emotional engagement, coherence, and interpretive difficulty. An analysis of variance was used to analyze the survey results. Our findings revealed that poems attributed to an AI consistently received lower scores for originality, aesthetic appeal, and emotional engagement compared to those attributed to a human author. However, AI-generated content was perceived as more complex and was rated higher in terms of interpretive difficulty. Interestingly, perceived authorship did not significantly influence coherence as a metric. When the poem was believed to be AI-generated, it faced more critical evaluations than when it was human-attributed. When authorship was ambiguous, feedback was distributed uniformly across negative, positive, and neutral sentiments, suggesting a potential mitigating effect of ambiguity on bias.