DASS: Designing Accountable Artificial Intelligence Services for People with Diverse Sensory Abilities

点击翻译按钮获取中文摘要

This project addresses the accessibility of information and communication technologies that include Artificial Intelligence (AI) services. These AI services are only as good as the data they’ve been fed, which often reflects historical or current biases either in the collection of data or in the data itself. In general, people with diverse visual and auditory disabilities are not well-represented in these datasets because although many people have some disability, the number with any one disability is usually a small percentage of the population. These disparities in representation can lead to disproportionate impacts of AI-based services on people with disabilities, and though regulations to protect people with disabilities exist, implementing them in AI-based systems and validating that the systems comply with those regulations in both the letter and spirit of the law are important, unsolved problems. To this end, the research team will develop evidence-based measurements, benchmarks, and guidelines for technologies that use AI services to ensure they comply with accessibility laws and are unbiased for people with visual or auditory disabilities.The project will use an interdisciplinary approach that integrates human-centered, social, and legal perspectives for interviews with, and surveys of, several groups of stakeholders critical to the design and validation of these systems. These include: 1) users with visual or aural disabilities, 2) information and communication technology developers that use AI services for accessibility, and 3) commercial AI services providers. The team will also survey the landscape of law and regulatory policy for using AI services for accessibility, as well as current practices around considering accessibility in developing these technologies. The project team will use the findings from these activities to analyze the design of AI-based technologies in terms of the types, collection, and access to data used to build these models, as well as in terms of the use cases, model performance, and transparency around the models themselves. Together, this analysis will help the team establish meaningful benchmarks and guidelines aimed at ensuring that AI-based information and communication services are fair for and comply with regulations that protect people with visual or auditory disabilities.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.