Self-reported perceived quality is largely decoupled from quiz performance in AI-generated educational interactive fiction¶

Claim · All claims
Evidence · 1 study · design r2 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i? Spearman correlations between sub-dimension ratings and gameplay metrics were small-to-moderate (|ρ| ≤ 0.38) and none reached uncorrected p < 0.05. → F. Rogosch 2026 q2 i? Correlations with the self-reported English-language barrier were negligible (|ρ| ≤ 0.10), so language issues do not plausibly explain the coherence result. → F. Rogosch 2026

Evidence¶

F. Rogosch 2026¶

F. Rogosch, A. Schrader. (2026). AI-Generated Interactive Fiction for Educational Use: A Pilot Study of Perceived Comprehensibility, Coherence, and Engagement. Proceedings of EDULEARN26 Conference. https://doi.org/10.21125/edulearn.2026.1075

q2 · i? · design · r2

Exploratory correlation analysis in the pilot (N = 22) linking questionnaire sub-dimensions to gameplay telemetry. Correlations were "all small-to-moderate (|ρ| ≤ 0.38)" and none significant; the strongest was engagement with extra clicks (ρ = -0.38, p = 0.08).

"Spearman rank correlations between each sub -dimension and the gameplay metrics (duration, extra clicks, first-try-correct rate, total wrong clicks) were all small-to-moderate (|ρ| ≤ 0.38) and none reached the uncorrected p < 0.05 threshold."

Discussion¶

Related Claims¶

• Story-content coherence is the weakest perceived-quality dimension of AI-generated educational interactive fiction in this pilot — related
• AI-generated interactive-fiction learning scenarios are perceived as narratively clear and of appropriate length by STEM higher-education participants — related
• In a small pilot of one AI pipeline's educational interactive fiction, the dominant free-text criticism was the artificial in-fiction motivation for quiz prompts — related
• Students associate poor slide quality with AI as the source: perceived quality correlates negatively with AI-source guesses — related
• Pedagogy and production dimensions associated most strongly with global quality (Spearman ρ = 0.68 and 0.66), while content was relatively independent (ρ = 0.46) — related
• Perceived tool quality was perfectly rank-associated with panel-rated presentation score among the four GenARch teams (exploratory) — related
• Engagement with AI-generated educational interactive fiction is marginal, near the neutral midpoint of the scale — related
• Abrupt location or character changes mid-scenario co-occur with the lowest story-content coherence ratings — related