Can Image Generation Be Considered An Expert?

Abstract

This paper addresses foundational challenges in evaluating Generative Artificial Intelligence (GAI), focusing on the transition from expertise evaluation to intelligence evaluation. It critiques both quantitative and qualitative metrics for GAI, highlighting limitations in human-algorithm interaction environments. The study examines knowledge representation in neural network architectures and the processes of filtering versus tokenisation for image processing, emphasising inconsistencies and lack of standardisation in test design. Based on this finding the paper proposes a framework for future research to further explore the research questions.

Research Contribution

• Proposes a new framework for evaluating GAI systems
• Critically examines current quantitative and qualitative metrics
• Identifies key limitations in human-algorithm interaction evaluation
• Provides foundation for future research in AI expertise assessment

Research Areas

This work contributes to several research domains:

• Generative AI Evaluation: Novel frameworks for assessing AI capabilities
• Computer Vision: Understanding image generation model performance
• AI Ethics: Implications of considering AI systems as experts
• Human-Computer Interaction: Evaluation in interactive environments

BibTeX Citation

@inproceedings{quang2025image,
  title={Can Image Generative Models Be Considered Experts?},
  author={Quang, D.B.H. and Dorleon, G. and Colarik, A.},
  booktitle={Information and Communication Technology. SOICT 2024},
  pages={--},
  year={2025},
  publisher={Springer},
  address={Singapore},
  series={Communications in Computer and Information Science},
  volume={2353},
  doi={10.1007/978-981-96-4291-5_33}
}

Recommended citation: Quang, D.B.H., Dorleon, G., & Colarik, A. (2025). "Can Image Generative Models Be Considered Experts?" Information and Communication Technology. SOICT 2024. Communications in Computer and Information Science, vol 2353. Springer, Singapore.
Download Paper