Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

AI Trust Reshaping Administrative Burdens: Understanding Trust-Burden Dynamics in LLM-Assisted Benefits Systems

Jeongwon Jo, He Zhang, Jie Cai, Nitesh Goyal

TL;DR

This study investigates how LLM-assisted benefits systems reshape administrative burdens for SNAP applicants, revealing that trust in AI—across benevolence, integrity, and competence—mediates learning, psychological, and compliance costs. Through 10 in-depth interviews and a SNAP-LLM prompt-engineering workflow, the authors show that while AI can reduce some burdens (e.g., stigma, wait times, and information gaps), it also introduces new burdens related to data security, transparency, and perceived loss of human accompaniment. The paper proposes a design agenda for empathetic, transparent, and hybrid human–AI benefits administration, including evidence-based information disclosures and source-attribution mechanisms to calibrate appropriate trust levels. The findings advance a novel trust–burden framework for AI-enabled public services and offer practical implications for deploying LLMs in benefits programs while mitigating unintended harms. Overall, successful real-world adoption will hinge on balancing automation benefits with transparent accountability and preserving meaningful human advocacy where needed.

Abstract

Supplemental Nutrition Assistance Program (SNAP) is an essential benefit support system provided by the US administration to 41 million federally determined low-income applicants. Through interviews with such applicants across a diverse set of experiences with the SNAP system, our findings reveal that new AI technologies like LLMs can alleviate traditional burdens but also introduce new burdens. We introduce new types of learning, compliance, and psychological costs that transform the administrative burden on applicants. We also identify how trust in AI across three dimensions--competence, integrity, and benevolence--is perceived to reduce administrative burdens, which may stem from unintended and untoward overt trust in the system. We discuss calibrating appropriate levels of user trust in LLM-based administrative systems, mitigating newly introduced burdens. In particular, our findings suggest that evidence-based information disclosure is necessary in benefits administration and propose directions for future research on trust-burden dynamics in AI-assisted administration systems.

AI Trust Reshaping Administrative Burdens: Understanding Trust-Burden Dynamics in LLM-Assisted Benefits Systems

TL;DR

This study investigates how LLM-assisted benefits systems reshape administrative burdens for SNAP applicants, revealing that trust in AI—across benevolence, integrity, and competence—mediates learning, psychological, and compliance costs. Through 10 in-depth interviews and a SNAP-LLM prompt-engineering workflow, the authors show that while AI can reduce some burdens (e.g., stigma, wait times, and information gaps), it also introduces new burdens related to data security, transparency, and perceived loss of human accompaniment. The paper proposes a design agenda for empathetic, transparent, and hybrid human–AI benefits administration, including evidence-based information disclosures and source-attribution mechanisms to calibrate appropriate trust levels. The findings advance a novel trust–burden framework for AI-enabled public services and offer practical implications for deploying LLMs in benefits programs while mitigating unintended harms. Overall, successful real-world adoption will hinge on balancing automation benefits with transparent accountability and preserving meaningful human advocacy where needed.

Abstract

Supplemental Nutrition Assistance Program (SNAP) is an essential benefit support system provided by the US administration to 41 million federally determined low-income applicants. Through interviews with such applicants across a diverse set of experiences with the SNAP system, our findings reveal that new AI technologies like LLMs can alleviate traditional burdens but also introduce new burdens. We introduce new types of learning, compliance, and psychological costs that transform the administrative burden on applicants. We also identify how trust in AI across three dimensions--competence, integrity, and benevolence--is perceived to reduce administrative burdens, which may stem from unintended and untoward overt trust in the system. We discuss calibrating appropriate levels of user trust in LLM-based administrative systems, mitigating newly introduced burdens. In particular, our findings suggest that evidence-based information disclosure is necessary in benefits administration and propose directions for future research on trust-burden dynamics in AI-assisted administration systems.

Paper Structure

This paper contains 51 sections, 1 figure, 1 table.

Table of Contents

  1. Introduction
  2. Background: The Journey to SNAP Benefits
  3. Related Work
  4. Administrative Burdens in Public Services
  5. Information and Communication Technologies and Administrative Burdens
  6. Trust in AI
  7. Method
  8. Prompt Engineering
  9. Refining with Experts
  10. Interviews & Participants
  11. Interview Structure
  12. Data Analysis
  13. Findings
  14. Psychological Costs
  15. Stigma and Emotional Barriers in SNAP Navigation
  16. ...and 36 more sections

Figures (1)

  • Figure 1: SNAP-LLM's User Interface and Configuration Components. On the left side of the image is the interface. In the middle are the configuration components, including official policy documents and external resources from the Family and Social Services Administration, instructions for system behavior and limitations, and usage scenarios to illustrate common use cases. On the right side is the description of how interviews were conducted.