Benjamin Ertman, Bridget Xia, Mona Sloane, Tom Hartvigsen, Paul B Perrin
{"title":"人工智能文本到图像生成中的残疾描述:语境的影响和残疾的医学化。","authors":"Benjamin Ertman, Bridget Xia, Mona Sloane, Tom Hartvigsen, Paul B Perrin","doi":"10.1037/rep0000634","DOIUrl":null,"url":null,"abstract":"<p><strong>Purpose/objective: </strong>Text-to-image (TTI) systems are artificial intelligence (AI) models that incorporate large amounts of data to produce high-resolution images. Although research has documented racial/ethnic and gender bias in TTI, little has examined disability bias. This study compared generated images of disabled people with no prompted setting to images of disabled individuals in health care settings.</p><p><strong>Research method/design: </strong>OpenAI's DALL-E-3 TTI generated 50 images for each of the following prompts: (a) \"person with a disability,\" (b) \"patient with a disability,\" (c) \"doctor with a disability,\" and (d) \"doctor with a disability and a patient without a disability.\" We calculated DALL-E's success in generating prompted images and coded disability type and demographics.</p><p><strong>Results: </strong>When prompted to create a \"person with a disability,\" DALL-E-3 was 100% successful, with a wide diversity of disabilities. When prompted to create a \"patient with a disability,\" DALL-E-3 was similarly 100% successful, although 70% of images portrayed an individual with a stereotypical physical disability. When prompted to create a \"doctor with a disability,\" DALL-E-3 did with 92% accuracy: 94% had a physical disability and 6% a sensory disability; no other disability types were portrayed. When prompted to create a \"doctor with a disability and a patient without a disability,\" in 64% of cases, DALL-E-3 generated images of doctors without disabilities, and 70% portrayed a disabled patient instead.</p><p><strong>Conclusions/implications: </strong>Disability diversity decreases dramatically when AI-generated images place disabled people in a medical environment. As TTI generation grows more ubiquitous, further work by model developers to mitigate representational harms is vital. (PsycInfo Database Record (c) 2025 APA, all rights reserved).</p>","PeriodicalId":47974,"journal":{"name":"Rehabilitation Psychology","volume":" ","pages":""},"PeriodicalIF":1.7000,"publicationDate":"2025-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Disability portrayals in artificial intelligence text-to-image generation: Influence of context and the medicalization of disability.\",\"authors\":\"Benjamin Ertman, Bridget Xia, Mona Sloane, Tom Hartvigsen, Paul B Perrin\",\"doi\":\"10.1037/rep0000634\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<p><strong>Purpose/objective: </strong>Text-to-image (TTI) systems are artificial intelligence (AI) models that incorporate large amounts of data to produce high-resolution images. Although research has documented racial/ethnic and gender bias in TTI, little has examined disability bias. This study compared generated images of disabled people with no prompted setting to images of disabled individuals in health care settings.</p><p><strong>Research method/design: </strong>OpenAI's DALL-E-3 TTI generated 50 images for each of the following prompts: (a) \\\"person with a disability,\\\" (b) \\\"patient with a disability,\\\" (c) \\\"doctor with a disability,\\\" and (d) \\\"doctor with a disability and a patient without a disability.\\\" We calculated DALL-E's success in generating prompted images and coded disability type and demographics.</p><p><strong>Results: </strong>When prompted to create a \\\"person with a disability,\\\" DALL-E-3 was 100% successful, with a wide diversity of disabilities. When prompted to create a \\\"patient with a disability,\\\" DALL-E-3 was similarly 100% successful, although 70% of images portrayed an individual with a stereotypical physical disability. When prompted to create a \\\"doctor with a disability,\\\" DALL-E-3 did with 92% accuracy: 94% had a physical disability and 6% a sensory disability; no other disability types were portrayed. When prompted to create a \\\"doctor with a disability and a patient without a disability,\\\" in 64% of cases, DALL-E-3 generated images of doctors without disabilities, and 70% portrayed a disabled patient instead.</p><p><strong>Conclusions/implications: </strong>Disability diversity decreases dramatically when AI-generated images place disabled people in a medical environment. As TTI generation grows more ubiquitous, further work by model developers to mitigate representational harms is vital. (PsycInfo Database Record (c) 2025 APA, all rights reserved).</p>\",\"PeriodicalId\":47974,\"journal\":{\"name\":\"Rehabilitation Psychology\",\"volume\":\" \",\"pages\":\"\"},\"PeriodicalIF\":1.7000,\"publicationDate\":\"2025-09-29\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Rehabilitation Psychology\",\"FirstCategoryId\":\"3\",\"ListUrlMain\":\"https://doi.org/10.1037/rep0000634\",\"RegionNum\":4,\"RegionCategory\":\"医学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q3\",\"JCRName\":\"PSYCHOLOGY, CLINICAL\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Rehabilitation Psychology","FirstCategoryId":"3","ListUrlMain":"https://doi.org/10.1037/rep0000634","RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"PSYCHOLOGY, CLINICAL","Score":null,"Total":0}
引用次数: 0
摘要
目的/目标:文本到图像(TTI)系统是人工智能(AI)模型,它包含大量数据以产生高分辨率图像。虽然研究记录了TTI中的种族/民族和性别偏见,但很少有研究调查残疾偏见。这项研究比较了没有提示设置的残疾人的生成图像和医疗机构中残疾人的图像。研究方法/设计:OpenAI的DALL-E-3 TTI为以下每个提示生成50张图像:(a)“残疾人”,(b)“残疾患者”,(c)“残疾医生”,(d)“残疾医生和无残疾患者”。我们计算了DALL-E在生成提示图像和编码残疾类型和人口统计数据方面的成功。结果:当提示创建一个“残疾人”时,DALL-E-3 100%成功,残疾的多样性很大。当提示创建一个“残疾患者”时,DALL-E-3同样100%成功,尽管70%的图像描绘的是一个典型的身体残疾患者。当提示创建“残疾医生”时,DALL-E-3的准确率为92%:94%的人有身体残疾,6%的人有感官残疾;没有描述其他残疾类型。当提示创建“残疾医生和非残疾患者”时,在64%的情况下,DALL-E-3生成的是非残疾医生的图像,而70%的情况下生成的是残疾患者的图像。结论/启示:当人工智能生成的图像将残疾人置于医疗环境中时,残疾多样性急剧下降。随着TTI生成变得越来越普遍,模型开发人员进一步减少表征危害的工作至关重要。(PsycInfo Database Record (c) 2025 APA,版权所有)。
Disability portrayals in artificial intelligence text-to-image generation: Influence of context and the medicalization of disability.
Purpose/objective: Text-to-image (TTI) systems are artificial intelligence (AI) models that incorporate large amounts of data to produce high-resolution images. Although research has documented racial/ethnic and gender bias in TTI, little has examined disability bias. This study compared generated images of disabled people with no prompted setting to images of disabled individuals in health care settings.
Research method/design: OpenAI's DALL-E-3 TTI generated 50 images for each of the following prompts: (a) "person with a disability," (b) "patient with a disability," (c) "doctor with a disability," and (d) "doctor with a disability and a patient without a disability." We calculated DALL-E's success in generating prompted images and coded disability type and demographics.
Results: When prompted to create a "person with a disability," DALL-E-3 was 100% successful, with a wide diversity of disabilities. When prompted to create a "patient with a disability," DALL-E-3 was similarly 100% successful, although 70% of images portrayed an individual with a stereotypical physical disability. When prompted to create a "doctor with a disability," DALL-E-3 did with 92% accuracy: 94% had a physical disability and 6% a sensory disability; no other disability types were portrayed. When prompted to create a "doctor with a disability and a patient without a disability," in 64% of cases, DALL-E-3 generated images of doctors without disabilities, and 70% portrayed a disabled patient instead.
Conclusions/implications: Disability diversity decreases dramatically when AI-generated images place disabled people in a medical environment. As TTI generation grows more ubiquitous, further work by model developers to mitigate representational harms is vital. (PsycInfo Database Record (c) 2025 APA, all rights reserved).
期刊介绍:
Rehabilitation Psychology is a quarterly peer-reviewed journal that publishes articles in furtherance of the mission of Division 22 (Rehabilitation Psychology) of the American Psychological Association and to advance the science and practice of rehabilitation psychology. Rehabilitation psychologists consider the entire network of biological, psychological, social, environmental, and political factors that affect the functioning of persons with disabilities or chronic illness. Given the breadth of rehabilitation psychology, the journal"s scope is broadly defined.