The Center for Democracy & Technology (CDT) has submitted a comment to the White House Office of Science and Technology Policy (OSTP) in response to its request for information on building the Federal Evidence Agenda on Disability Equity. The Agenda aims to enhance the federal government’s ability to make data-informed policy decisions that advance equity for people with disabilities.
CDT praised OSTP for seeking public input on this issue and for its dedication to achieving real equity for people with disabilities. As experts in technology policy, CDT has long recognized how underinclusive or non-inclusive datasets can contribute to algorithmic bias, impacting marginalized communities, particularly those with disabilities.
"Algorithmic systems are trained on datasets, where the system learns patterns included in the dataset, and then uses those patterns to create new outputs," CDT explained. "When datasets used to build and train algorithmic models are not fully representative of people with disabilities, they can amplify algorithmic bias and its impacts."
To mitigate these effects, CDT emphasized the importance of training algorithms on datasets that accurately represent people with disabilities and using these datasets to evaluate and mitigate an algorithmic system’s risks. A thoughtful federal strategy on disability equity in data will be instrumental in ensuring data is gathered in ways that reduce discrimination, including technology-facilitated disability discrimination.
In its response to OSTP's questions regarding disparities, data collection, public access, privacy, security, and civil rights, CDT highlighted several key points:
- There are at least four different models of defining disability (legal, medical, social, and identity/demographic), which can influence data gathering outcomes.
- Barriers such as disability-related stigma and disproportionate incarceration and institutionalization of disabled individuals complicate accurate data collection.
- Agencies should incorporate various privacy protection mechanisms into their practices when gathering disability-related data. These include encryption, de-identification, aggregation when possible, and retaining data only as long as necessary.
"We look forward to continuing to collaborate with OSTP in creating more equitable data collection practices for people with disabilities," CDT stated. "And we anticipate the publication of the Federal Evidence Agenda."
Read the full comments here.
