MFIA Clinic Advances New Project on Algorithmic Transparency
The Media Freedom and Information Access Clinic (MFIA) launched two new projects on algorithmic transparency this past semester, with the goal of developing model legislation to promote transparency around the use of algorithms by government agencies and disclosure of their potential biases.
The Clinic is working with the Connecticut Council on Freedom of Information, Yale Law School Sterling Professor of Law Robert Post ’77, and Information Society Project Visiting Fellow Alicia Solow-Niederman to research how Connecticut agencies currently evaluate the algorithms they use for potential problems such as ineffectiveness or bias, and to assess the level of public transparency about these algorithms under existing open records laws. MFIA Clinic students Chloe Francis ’23, Sarah Lamsifer ’21, and Karen Sung ’23, are working on the project.
During the spring 2021 term, the Clinic first conducted broad research into the state of algorithmic transparency and accountability and discovered that Connecticut agencies do not disclose much about how they use algorithms and how those algorithms might affect the public, such as by inadvertently perpetuating bias. The Clinic then decided to do a deep dive into two agencies: the Connecticut Department of Children and Families and the Department of Education. The Department of Children and Families uses an algorithm that assesses children’s risk levels to determine when government intervention is appropriate. The Department of Education conducts lotteries for child placements into schools. It is unclear what factors these lotteries consider, but they have been reported to lead to students missing out on their preferred schools, even when those schools had space for them.
To learn more, the Clinic worked with Solow-Niederman, an expert on artificial intelligence and government accountability, to draft Freedom of Information requests to the two agencies. According to Stephen Stich ’17, the MFIA Local Journalism Fellow supervising the project, “While algorithms certainly can help Connecticut agencies make better decisions, there’s a lot we don’t know about them, like how error-prone they are or whether they perpetuate bias. More fundamentally, we don’t even know if the agencies know the answers to these questions. The government should tell the public how it makes decisions, and things are no different when those decisions are at least partly automated.”
The Clinic plans to develop model legislation to improve algorithmic accountability based on what it learns through the FOIA requests.
The Media Freedom and Information Access Clinic (MFIA) is a law student clinic dedicated to increasing government transparency, defending the essential work of news gatherers, and protecting freedom of expression by providing pro bono legal services, pursuing impact litigation and developing policy initiatives.
By Leah Ferentinos