Millions of people with accessibility needs resulting from sensory, cognitive, and motor impairments are impacted by software every day. This software is increasingly using Artificial Intelligence (AI) and machine learning (ML) technologies that are used to personalize people’s technology experiences, make predictions about their likes, needs, and behaviors, and suggest courses of action. However, these AI services are only as good as their datasets, which can cause biases both because people with disabilities are likely to be underrepresented in a dataset relative to the population as a whole, and because the data collected may reflect social biases against those people.

Through interviewing people with disabilities, developers, and regulators of these systems, along with a thorough review of existing systems and regulations that target disabilities and accessibility issues, the project team will develop a model of the accessibility risks that can arise in developing systems with AI and ML components, along with guidelines for mitigating those risks and methods to assess them. This will in turn set up opportunities where people with disabilities play a more active role in co-designing these systems, in ways that are more likely to meet everyone’s needs. 

“This work matters because although there are laws that protect people with disabilities, how those laws should be realized in software systems is a hard problem,” said NSF Program Director Dan Cosley. “Realizing them in AI and ML algorithms is an especially interesting case, because questions of fairness and bias in general are increasingly on the radar for both developers and regulators around these algorithms—but the opaqueness of both the algorithms themselves and the processes for developing them raises extra challenges for assessing compliance.” 

School districts struggle to increase equal educational opportunities and avoid racial isolation for their students. Many districts use a school-assignment software system to further these goals. Today, more than 100 school districts across the US have adopted school-assignment software systems. Some of these systems have been subject to legal challenges, both successful and unsuccessful, to the approach the school districts used to advance equal opportunity.  

This study examines how best to ensure that these software systems are designed to respond to legal and economic constraints and democratic community participation in order to ensure both legal compliance and legitimacy in the eyes of the community. Improving these software systems will impact access to education for thousands of students.  

“This is timely research at the intersection of policy, education and software design that has the potential to enhance democratization and equality in school districts across the country.  It demonstrates the strength of collaborative expertise in social science and computer science coming together to advance a significant societal and educational goal,” said NSF Program Director Reggie Sheehan 

Software, regulation, and society interact with unpredictable and sometimes undesirable dynamic feedback effects. This project studies how agent-based models (ABMs) of social contexts can improve the design and regulation of accountable software systems to fill the gap between legal requirements and software design. The goal of this project is to help regulators, domain experts, software designers, and other stakeholders assess the potential societal implications of particular software and regulatory systems. Agent-based modeling (ABM) is a social scientific research method that involves bottom-up modeling of complex systems and computationally determining their emergent properties by running simulations. The investigators use ABMs to model elements of the social and regulatory environment in which a software system operates, and to guide the crafting and enforcement of technology regulations. 

NSF Program Director Nina Amla said, "this is a promising approach for improving the accountability of software deployed in social contexts like advertising software, fair housing, contact tracing, and privacy."

Berger-Wolf is among the founders of the field of computational ecology, leading research at the unique intersection of computer science, wildlife biology, and social sciences. She creates computational AI and data science solutions to generate, collect, analyze, and derive insight from data to answer questions and make decisions in ecology, environment, and wildlife conservation. Tanya is passionate about creating a human-AI partnership to address the greatest conservation challenges of our planet. As a legitimate part of her work, she gets to hang out of an ultra-light airplane in Kenya, taking a hyper-stereo video of zebra populations and learning how to identify each one of them by the unique stripe pattern. She works with many teams of brilliant people who do cool, amazing, and impactful things. 

She has received numerous awards for her research and mentoring, including University of Illinois Scholar, University of Illinois Chicago (UIC) Distinguished Researcher of the Year, NSF Faculty Early Career Development Program (CAREER), Association for Women in Science Chicago Innovator, and the UIC Mentor of the Year. Berger-Wolf has given hundreds of talks about her work, including at TEDx (featured on TED.com) and UN/UNESCO AI for the Planet. 

Taking research to conservation action and policy, Berger-Wolf is also a director and co-founder of the conservation software non-profit, Wild Me, home of the Wildbook project, which brings together AI, computer vision, crowdsourcing, and conservation. Wild Me enables the work of many conservation organizations and government agencies, such as the World Wildlife Fund, Bureau of Ocean Energy Management, National Oceanic and Atmospheric Administration, Kenya Wildlife Service, Grevy’s Zebra Trust, Giraffe Conservation, and many others. Wildbook has been recently chosen by the United Nations Educational, Scientific and Cultural Organization as one of the top AI 100 projects worldwide supporting the United Nations Sustainable Development Goals.   

Most of Berger-Wolf’s research has been supported by NSF. The graduate research fellowship and the CAREER awards helped launch her career. Her collaborative interdisciplinary research in computational approaches to understanding (animal) sociality was supported by several CISE and cross-cutting awards. Her passion for engaging kids in science and for educating the next generation of interdisciplinary scientists was supported by Broadening Participation in Computing and educational grants. NSF also supported the initial stages of research that became the foundation of Wild Me which is now powering conservation worldwide.  

NSF's large-scale AI and data science programs, such as Transdisciplinary Research in Principles of Data Science, National AI Research Institutes, and Harnessing the Data Revolution, are designed to foster the exploration of data science and AI, from foundations to applications. Berger-Wolf’s vision for collaborative interdisciplinary research in computational ecology and for turning images into the source of information about life on the planet has recently got a big boost from NSF. She and a team of interdisciplinary scientists have been awarded a $15-million grant to establish a new Harnessing Data Revolution Data-Intensive Research in Science and Engineering Institute, founding a new field of study: Imageomics. It will use AI and other computational approaches to extract biological information, particularly traits, from images, leading to new understanding of evolution and ecology.  

Tanya Berger-Wolf is a great example of how NSF supports scientists in their early and late stages of their research careers.