From Voices to Validity: Leveraging Large Language Models (LLMs) for Textual Analysis of Policy Stakeholder Interviews

Name: From Voices to Validity: Leveraging Large Language Models (LLMs) for Textual Analysis of Policy Stakeholder Interviews
Published: 2025-08-07
License: https://creativecommons.org/licenses/by-nc/4.0

Principal Investigator(s): View help for Principal Investigator(s) Min Sun, University of Washington

Version: View help for Version V1

American Educational Research Association Data Repository

Name	File Type	Size	Last Modified
Qualitative_analysis_using_openai.ipynb	text/plain	69.2 KB	08/07/2025 03:11:PM
README.md	text/x-web-markdown	2 KB	08/07/2025 03:17:PM
STM_topic_model.Rmd	text/plain	5.5 KB	08/07/2025 03:11:PM
codebook-2.csv	text/csv	10.7 KB	08/07/2025 03:11:PM

Project Citation:

Sun, Min. From Voices to Validity: Leveraging Large Language Models (LLMs) for Textual Analysis of Policy Stakeholder Interviews. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor], 2025-08-07. https://doi.org/10.3886/E237139V1

Project Description

Summary: During this six-month project, the team conducted a comprehensive landscape and trend analysis of Washington State’s K–12 public school systems to identify existing assets and gaps. In collaboration with state education leaders, practitioners, and other stakeholders, the team identified high-leverage questions and priority areas for improving outcomes for low-income students and students of color.The project also developed recommendations for strengthening state and local district capacity to engage in evidence-based continuous improvement. Findings were communicated to a diverse audience of policymakers and practitioners to inform the policy feedback loop, highlighting implications for policies and actions that can advance racial and economic equity in Washington’s public education system. Finally, the team identified key areas for further inquiry to guide ongoing improvement efforts.This study is part of a larger project focused on text analysis using interview data, applying qualitative and computational methods to examine stakeholder perspectives and identify actionable themes.

Scope of Project

Subject Terms: educational policy; nlp; human-ai interactive

Geographic Coverage: Washington

Time Period(s): 2018 – 2022

Collection Date(s): 9/2021 – 3/2022

Data Type(s): text

Methodology

Sampling: The data for this study comprise 24 interviews with diverse educational policy stakeholders. Our purposeful sampling strategy was designed to maximize representation based on the following criteria (Maxwell, 2004; Patton, 1990): (a) the level within public school systems, including classrooms, schools, districts, and the state; (b) geographic locations within the state, including urban, suburb, and tribal schools; (c) roles of interviewees, spanning system actors at various levels of educational systems and three branches of the government at the state level, as well as non-system actors such as community organization leaders, advocates, lobbyists, teacher union representatives, and philanthropic organizational leaders; and (d) characteristics of students and local communities in terms of race/ethnicity, socioeconomic status, language, and homeless populations.

Data Source: The interviewees were categorized into three primary professional groups: policymakers and administrators (state legislators, state-level policymakers, and school district administrators), educators (teachers and those in coaching or mentoring roles), and non-profit sector participants along with advocates (including teacher union representatives, policy advocates, and community leaders). Interviews were conducted virtually via Zoom, each lasting between 45 to 60 minutes.

Collection Mode(s): computer-assisted personal interview (CAPI)

Unit(s) of Observation: Individuals

Related Publications

Download this project

Published Versions

V1 [2025-08-07]

Export Metadata

Dublin Core

DDI 2.5

Report a Problem

Found a serious problem with the data, such as disclosure risk or copyrighted content? Let us know.

This material is distributed exactly as it arrived from the data depositor. ICPSR has not checked or processed this material. Users should consult the investigator(s) if further information is desired.