Visual Representation of "Chinoiserie" in Western AI Image Models: A Case Study of Midjourney

Shurong Wang

doi:10.63313/IJSSEH.9063

Authors

Shurong Wang Hangzhou Normal University, Hangzhou 310000, China Author

DOI:

https://doi.org/10.63313/IJSSEH.9063

Keywords:

AI-Generated Imagery, Chinoiserie, Cultural Bias

Abstract

Using visual grammar as its theoretical framework, this study adopts a comparative experimental approach to perform a multimodal discourse analysis on 240 "Chinoiserie" (Chinese style) images generated by Midjourney. The analysis evaluates baseline and professionally constrained prompt groups across three dimensions: architectural space, costumed portraiture, and landscape atmosphere. Results show that in representational meaning, the model exhibits attribute ambiguity, structural dislocation, and cultural semantic rupture, achieving high adaptability only with abstract landscape styles. In interactive meaning, it demonstrates a spectacularized and Othered perspective, constructing a Eurocentric narrative through low-angle shots and demand-gaze interaction. In compositional meaning, stereotypical symbols occupy an exceptionally high proportion; professional constraints paradoxically exacerbate symbol stacking, and the aesthetic of negative space (Liu Bai) is entirely erased during entity generation. The study indicates that the "Chinoiserie" generated by Western AI is not an authentic cultural representation but a de-historicized digital simulacrum marked by significant cultural bias and aesthetic homogenization. Finally, governance recommendations regarding dataset construction and prompt optimization are proposed to serve as empirical references for the cross-cultural communication of Chinese culture and algorithmic bias correction in the AI context.

References

[1] Peng, L. (2025). When algorithms become the “mirror” of humans. Journalism and Writing, (9), 1.

[2] Zhang, K., & Wang, X. (2024). Media presentation and risk governance of generative AI in traditional cultural communication. Library and Information Knowledge, 41(4), 98–109. DOI:10.13366/j.dik.2024.04.098.

[3] Sun, T. (2021). Characteristics and landscape of visual communication research since the 21st century: A content analysis based on two English academic journals (2002–2019). Journalism Research, (11), 44–58+123. DOI:10.20050/j.cnki.xwdx.2021.11.005.

[4] UNESCO. (2024). Challengingtem Sysatic Prejudices: An Investigation into Bias Against Women and Girls in Large Language Models.

[5] HAI. Artificial Intelligence Index Report 2025[R/OL]. https://arxiv.org/pdf/2504.07139

[6] KRESS G R, VAN LEEUWEN T. Reading Images: The Grammar of Visual Design[M]. London: Routledge, 2006.

[7] FRIEDMAN B, NISSENBAUM H. Bias in computer systems[J]. ACM Transactions on Information Systems, 1996, 14(3): 330-347

[8] NOBLE S U. Algorithms of Oppression: How Search Engines Reinforce Racism[M]. New York: NYU Press, 2018.

[9] KELLNER D. Media Spectacle[M]. London: Routledge, 2003.

[10] SAID E W. Orientalism[M]. New York: Pantheon Books, 1978.

[11] BAUDRILLARD J. Simulacra and Simulation[M]. Ann Arbor,1994.

[12] FRIEDMAN B, NISSENBAUM H. Bias in computer systems[J]. ACM Transactions on Information Systems, 1996, 14(3): 330-347.

[13] MANOVICH L. Software Takes Command[M]. London: Bloomsbury Academic, 2013.

[14] CRAWFORD K. The Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence[M]. New Haven: Yale University Press, 2021.

[15] Pan, S., Luo, L., Wang, Y., & Chen, M. (2024). Unifying Large Language Models and Knowledge Graphs: A Roadmap. IEEE Transactions on Knowledge and Data Engineering (TKDE).

[16] Blodgett, S. L., Barocas, S., Daumé III, H., & Wallach, H. (2020). Language (Technology) is Power: A Critical Survey of "Bias" in NLP. Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, 5454–5476.