EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-ND 4.0

Can AI Assess Creativity? The Validity of Machine Learning Models in Scoring Open-Ended Art and Design Tasks

Bei TU

Artificial Intelligence Education Studies · 2025

DOI: 10.6914/aiese.010403 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Artificial Intelligence Education Studies, 2025. Verfasst von Bei TU. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This paper critically examines the validity of Machine Learning (ML) models in assessing human creativity within open-ended art and design tasks. As AI tools increasingly permeate educational assessment, they offer scalability but face significant challenges in capturing the subjective and nuanced nature of creativity. Through an analysis of current ML paradigms—such as Convolutional Neural Networks (CNNs) and computational aesthetics—the study reveals that while AI achieves high accuracy in replicating human scores, it often relies on surface-level pattern recognition rather than a deep understanding of context, intent, or "appropriateness." The paper highlights critical limitations, including algorithmic bias, the "black box" opacity of deep learning, and the risk of "teaching to the algorithm." Ultimately, it argues that AI lacks the construct validity to fully replace human evaluators and should instead be positioned as an assistive tool to enhance human judgment and pedagogical feedback.

Zitieren

APA

TU, B. (2025). Can AI Assess Creativity? The Validity of Machine Learning Models in Scoring Open-Ended Art and Design Tasks. Artificial Intelligence Education Studies. https://doi.org/10.6914/aiese.010403

Vancouver

TU B. Can AI Assess Creativity? The Validity of Machine Learning Models in Scoring Open-Ended Art and Design Tasks. Artificial Intelligence Education Studies. 2025. doi:10.6914/aiese.010403

BibTeX
@article{TU2025,
  author = {TU, Bei},
  title = {Can AI Assess Creativity? The Validity of Machine Learning Models in Scoring Open-Ended Art and Design Tasks},
  journal = {Artificial Intelligence Education Studies},
  year = {2025},
  doi = {10.6914/aiese.010403}
}
Kostenlos starten →