Automation, AI and Nudges: Improving Digital Interfaces to Reduce Administrative Burdens in the Safety Net

Session Submission Type: Panel

Abstract

This panel brings together a collection of papers that examines how digital interfaces can be reconfigured to reduce administrative burdens. As more and more of citizen-state interactions occur in virtual spaces, the design of digital systems can reduce burdens, or in some cases, make burdens greater due to algorithmic bias or digital divide issues. Each paper in this panel considers how changes to digital system in combination with the use of administrative data can reduce frictions in the social safety net. This is a topic with a great deal of promise, and risk, making rigorous empirical testing that sorts out observable effects from hyperbole all the more essential. 

The panel examines a variety of experiments in different policy settings (SNAP, housing, Medicaid, and TANF) and a variety of interventions, ranges from nudges to automation, and considering the potential for AI. In Kim et al., a relatively simple reminder nudge built into a SNAP application app has a large positive effect in pushing SNAP applicants to complete previously initiated applications. Both Cholli and Wu, and Unrath, look at the potential for automatic enrollment. Giannella et al. engage in a rigorous test of the efficacy and risk of Large Language Models as a tool to help caseworkers make determinations on potential applications. All of the papers employ experimental designs to provide causal estimates, with three of the four relying on field experiments. Because of the nascent nature of the use of LLMs, an experimental design was employed, using actual caseworkers but asking them to use AI tools with hypothetical rather than actual cases.

The panel includes a truly interdisciplinary group of scholars, including public policy, political science, economics, public administration, computer science and sociology scholars. The topics have clear policy relevance.

Policy Area

• Poverty and Income Policy

Secondary Policy Area

• Public and Nonprofit Management and Finance

Chair

• Donald Moynihan, University of Michigan

Discussants

• Jeremy Barofsky, Georgetown University

• Katherine Richard, Georgetown University

Organizer

• Donald Moynihan, University of Michigan

Individual Presentations

• Multiple Program Participation: Incidence and Impacts from Reducing Administrative Burdens - Non-Presenting Co-Author: Derek Wu, University of Virginia; Presenting Author: Neil A Cholli, Cornell University

• Can AI Reduce Caseworker Burdens and Improve Caseworker Performance? - Presenting Author: Zhaowen Guo, Georgetown University; Non-Presenting Co-Author: Eric Giannella, Georgetown University

• Automatic Enrollment for the People: Experimental Evidence on Default Enrollment in a Safety Net Program - Presenting Author: Matthew Unrath, University of Southern California

• A Simple Messaging Intervention Increases Completion of SNAP Applications - Presenting Author: Jae Yeon Kim, Harvard University; Non-Presenting Co-Author: Eric Giannella, Georgetown University; Non-Presenting Co-Author: Zhaowen Guo, Georgetown University; Non-Presenting Co-Author: Pamela Herd, Georgetown University; Non-Presenting Co-Author: Sebastian Jilke, Georgetown University; Non-Presenting Co-Author: Donald Moynihan, University of Michigan