EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Artistic-Value Descriptions Are Associated with Higher Aesthetic Ratings than Social-Value Descriptions for AI-Generated Variants of Classical Paintings

Qi Li iD, Qi Li, Chaoyuan Zhang, Ziyu Feng

Behavioral Sciences · 20.08.2026 · 16 (8) · 1437

DOI: 10.3390/bs16081437 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Behavioral Sciences, online erschienen am 20.08.2026. Verfasst von Qi Li und 3 weiteren Autor:innen. Beteiligte Einrichtung: School of Fine Arts, Shanxi University, Taiyuan 030006, China. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 31 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Watershed Vernacular Built Environment Shanxi Provincial Key Laboratory Open Research Fund, 2025 Shanxi Province Art Science Planning Project.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Contextual meaning can influence art evaluation even when viewers encounter images within a structured visual stimulus family. This exploratory questionnaire-version study examined whether value-oriented textual descriptions were associated with aesthetic ratings of AI-generated variants of classical paintings in three Chinese-language questionnaire versions emphasising cultural value, artistic value, or social value. The available dataset contained 61 participants, each rating 40 images on a 1–10 aesthetic scale, yielding 2440 complete observations. A post hoc three-rater manipulation check of 30 image–description pairs found that all evaluated pairs satisfied the applied validation criterion; inter-rater reliability across item–dimension judgments was high at ICC(2,k) = 0.936. The primary analysis used a linear mixed-effects model with crossed random intercepts for participant and stimulus identity, and a fuller by-stimulus random-slope model was estimated as a diagnostic response to review. Relative to social-value descriptions, artistic-value descriptions were associated with higher ratings in the random-intercept model, where b = 1.74, SE = 0.57, z = 3.03, p = 0.002, 95% CI [0.62, 2.87], and this contrast also persisted in the diagnostic by-stimulus random-slope model, although that model converged with a non-positive-definite Hessian. Cultural-value descriptions were intermediate and should not be interpreted as showing a stable general advantage over social-value descriptions. The results therefore provide qualified evidence for an artistic-versus-social framing association, not a confirmed causal effect of all value-cue conditions. Interpretation remains bounded by the modest sample, undocumented allocation procedure, rapid-completion records, incomplete prompt metadata, and post hoc validation of a representative rather than exhaustive stimulus set.

Schlagwörter

Stichwörter der Autor:innen

AI-generated imagesaesthetic cognitionart perceptioncontextual informationempirical aestheticsmixed-effects modelvalue cues

Autor:innen und Institutionen

  • School of Fine Arts, Shanxi University, Taiyuan 030006, China.

ORCID-Profile: Qi Li

Förderung

  • Watershed Vernacular Built Environment Shanxi Provincial Key Laboratory Open Research Fund
  • 2025 Shanxi Province Art Science Planning Project

Zitieren

APA

Li, Q., Li, Q., Zhang, C., & Feng, Z. (2026). Artistic-Value Descriptions Are Associated with Higher Aesthetic Ratings than Social-Value Descriptions for AI-Generated Variants of Classical Paintings. Behavioral Sciences, 16(8), 1437. https://doi.org/10.3390/bs16081437

Vancouver

Li Q, Li Q, Zhang C, Feng Z. Artistic-Value Descriptions Are Associated with Higher Aesthetic Ratings than Social-Value Descriptions for AI-Generated Variants of Classical Paintings. Behavioral Sciences. 2026;16(8):1437. doi:10.3390/bs16081437

BibTeX
@article{Li2026,
  author = {Li, Qi and Li, Qi and Zhang, Chaoyuan and Feng, Ziyu},
  title = {Artistic-Value Descriptions Are Associated with Higher Aesthetic Ratings than Social-Value Descriptions for AI-Generated Variants of Classical Paintings},
  journal = {Behavioral Sciences},
  year = {2026},
  volume = {16},
  number = {8},
  pages = {1437},
  doi = {10.3390/bs16081437}
}
Kostenlos starten →