Varela-Rodríguez Miguel, Plage Stefanie
Department of Sociology and Social Work, Faculty of Commerce, University of Valladolid, Pl. Campus Universitario, 1, 47011, Valladolid, Spain.
School of Social Science, Faculty of Humanities, Arts and Social Sciences, The University of Queensland, Brisbane, Australia.
J Cancer Surviv. 2025 Jun 7. doi: 10.1007/s11764-025-01843-z.
Generative Artificial Intelligence (GAI) is transforming visual communication in the context of cancer survivorship, presenting opportunities to innovate advocacy while also posing risks for social representation. This study explores how GAI visualizes cancer and survivorship, focusing on its ability to reflect diverse experiences and its limitations.
We analyzed 262 images generated by Dall-E and Stable Diffusion using prompts related to breast, prostate, and pancreatic cancer. A mixed-methods approach examines how GAI utilizes cancer signifiers, visualizes the impact of cancer on individuals, and represents people with cancer.
GAI frequently reproduces cancer tropes, such as prescriptive positivity, and fails to depict medical treatments or embodied experiences unless explicitly prompted. AI-generated images predominantly featured White, female subjects, particularly in breast cancer contexts, reflecting broader biases in public discourse. While GAI tools can produce inclusive visuals, achieving this requires users to have nuanced knowledge of cancer and survivorship, limiting accessibility for lay GAI users.
GAI can support cancer communication but risks perpetuating stereotypes and excluding less visible experiences of cancer. Our findings offer practical insights to support the design of advocacy materials and campaigns, particularly through improved prompt literacy and inclusive image generation strategies.
Inclusive and respectful visual representation is critical for capturing the diverse realities of cancer survivorship, which in turn affects the wellbeing of cancer survivors and carers. Collaborative efforts among researchers, advocates, and GAI developers are necessary to improve datasets and foster accessible tools, ensuring that GAI supports rather than undermines cancer survivorship advocacy.
生成式人工智能（GAI）正在改变癌症幸存者背景下的视觉传播，为倡导创新带来机遇，同时也给社会表征带来风险。本研究探讨了GAI如何将癌症和幸存者状态可视化，重点关注其反映不同经历的能力及其局限性。
我们使用与乳腺癌、前列腺癌和胰腺癌相关的提示词，分析了由Dall-E和Stable Diffusion生成的262张图像。采用混合方法研究GAI如何利用癌症标识符、将癌症对个体的影响可视化以及呈现癌症患者。
GAI经常重现癌症比喻，如规定性的积极态度，除非明确提示，否则无法描绘医疗治疗或具体经历。人工智能生成的图像主要以白人女性为特征，尤其是在乳腺癌背景下，反映了公共话语中更广泛的偏见。虽然GAI工具可以生成包容性的视觉效果，但要实现这一点，需要用户对癌症和幸存者状态有细致入微的了解，这限制了普通GAI用户的可及性。
GAI可以支持癌症传播，但存在延续刻板印象和排除癌症不太明显经历的风险。我们的研究结果提供了实用的见解，以支持宣传材料和活动的设计，特别是通过提高提示词素养和包容性图像生成策略。
包容和尊重的视觉表征对于捕捉癌症幸存者的多样现实至关重要，而这反过来又会影响癌症幸存者和护理人员的福祉。研究人员、倡导者和GAI开发者之间的合作努力对于改进数据集和开发易于使用的工具是必要的，以确保GAI支持而不是破坏癌症幸存者倡导工作。
