As artificial intelligence proliferates in American life, new research finds that the technology is poised to amplify stereotypes about people with intellectual disabilities.
ChatGPT, Claude and other leading AI systems consistently portray people with intellectual disabilities as dependent, childlike and requiring supervision, according to findings published recently in the Disability and Health Journal.
Researchers at Special Olympics International and Oregon State University asked five of the most popular AI models to create stories about people with and without intellectual disabilities in common situations such as crossing a street, playing a sport, going to a store or working at a job. Some 25,000 stories were generated, half about people with intellectual disabilities and half about those with no specified diagnosis. The researchers then used a separate AI instance to analyze the stories for bias.
Advertisement – Continue Reading Below
Overall, the study found that stories about people with intellectual disabilities tended to be paternalistic and were more likely to depict the subjects as younger even when no age was indicated. In addition, these stories featured people who had more supervision and less autonomy and involvement in decision-making.
AI systems were also more likely to portray people with intellectual disabilities as a source of inspiration, though the stories often exhibited hesitation about including the person and frequently included a negative perception of such individuals, the study found.
“Our findings revealed a pattern of representational discrepancies between groups that imply encoded implicit bias about people with (intellectual disabilities) in the products generated by current state (large language models),” the researchers said in their findings.
Those behind the study warn that the pattern of bias found in the AI-generated stories is consistent with previously documented ableist stereotypes and could lead to real life consequences for people with disabilities by influencing access to healthcare, employment and social opportunities. It could also potentially affect policy and programs or even laws, they said.
“What our research shows is that bias doesn’t have to be hateful to be harmful — it can be subtle, systematic and scaled globally. AI systems are learning stereotypes and repeating them at scale,” said Nathan Cook, chief information and technology officer at Special Olympics.
Cook said Special Olympics is “actively working” with Microsoft and welcomes other AI companies to engage with the group to make this population a priority.
“Artificial intelligence is quickly becoming foundational to how our world operates — from healthcare to education to economic opportunity. That makes this a pivotal moment. If we do not intentionally design for inclusion from the outset, we risk embedding bias into systems that will scale globally and persist for generations,” Cook said. “We want to disrupt automated ableism from the start.”
Nyasha Derera, a Special Olympics athlete from Zimbabwe, raised concerns about how statistics and data presented by AI chatbots might influence perceptions.
“Many people with (intellectual disabilities) have experienced exclusion, but there are also powerful stories of inclusion that deserve to be highlighted without bias. Bringing these stories to the forefront could encourage more positive attitudes and support broader inclusion,” Derera said.
