AI evidence extraction

Investigating the use of generative AI policies among ASPPH member schools and programs of public health

Ashish Joshi, Brian McGoldrick, Nidhi Mittal, Shongkour Roy, Zebunnesa Zeba, Michael Arthur Ofori et al. · Frontiers in Public Health · 2026

AI-generated evidence extraction, verified across multiple analytical personas. Not a substitute for the peer-reviewed original.

9/10

Relevance

0/4

Quality (LMQS)

Evidence

Citations

0.00

FWCI

This is an AI analysis. Read the peer-reviewed original at the publisher: https://doi.org/10.3389/fpubh.2026.1796810

Methodology & findings

Study design

Mixed-methods descriptive analysis incorporating: (1) systematic document collection and review from ASPPH member websites (October 2024–January 2025); (2) policy/guideline classification using consensus coding and predefined criteria from University of Wisconsin-Madison Policy Library; (3) content analysis of 7 focus areas; (4) thematic analysis (inductive approach with codebook development); (5) text mining using TF-IDF analysis; and (6) inter-rater reliability assessment using Krippendorf's Alpha..

Sample

N = 155, 15 groups

Primary method

Krippendorf's Alpha (Kα) for inter-rater reliability assessment; Percentage of agreement calculation (base agreement not accounting for chance); Descriptive statistics for institutional characteristics and document distributions; Content analysis (categorical coding of predefined variables); Thematic analysis (inductive approach with codebook development); Term Frequency (TF) and Inverse Document Frequency (IDF) text mining analysis; TF-IDF analysis for identifying frequently occurring terms and patterns; R statistical software (Hayes' KA macro for Krippendorf's Alpha calculation)

Main result

The study found that of 155 ASPPH member institutions reviewed, only 18 had formally adopted AI policies meeting stringent criteria. The analysis revealed that "'AI Integrity/Misconduct Guidance' was defined as language that specifies how AI use may violate academic integrity and/or be considered academic misconduct. Fourteen out of eighteen universities reviewed included academic integrity guidelines related to AI use in their institutional policies." Additionally, "the most common applications discussed include teaching (89.96%) and learning (82.54%)," with ChatGPT being "the most often-discussed tool, with 77.78% of member schools explicitly mentioning it."

Reports effect sizes and confidence intervals.

Research paradigm

Qualitative-quantitative mixed methods; descriptive and exploratory

Author conclusions

"This is the only known study that characterizes AI use guidance in schools and programs of public health. The findings of the collective analyses can be used to inform stakeholders of the current state of AI usage not only in schools and programs of public health but universities at large." The authors conclude that "There is a need to develop consistent, comprehensive policies on generative AI use across ASPPH member schools. However, existing program policies concerning AI rarely address it in context of community engagement." They note that "Although we demonstrate how current policies focus on academic integrity and ethical use, more schools are reviewing their programs to address AI integration in crucial career-building skills that will equip public health students for future career growth."

Risk of bias

Selection bias: Only publicly available documents were analyzed; internal policies were excluded; Document availability bias: 23 documents had no AI guidance; 2 institutions had documents not publicly accessible; Temporal bias: Analysis reflects guidance available through March 1, 2025 only; Rater bias: Low-to-moderate inter-rater reliability (Kα=0.514) suggests potential inconsistency in policy classification; Classification bias: Researchers applied predefined criteria that may not capture nuanced institutional contexts; Language bias: Analysis limited to English-language documents; Measurement bias: Reliance on standardized policy criteria may not capture all institutional AI governance approaches; Inter-rater reliability concerns: Kα = 0.514 indicates low to moderate agreement, though authors note this may be explained by the mathematics of Krippendorf's Alpha with low expected disagreement; Document interpretation bias: Three reviewers coded independently with potential for subjective interpretation despite consensus procedures; Selection bias: Only publicly available policies included; internal policies not captured; Classification bias: Moderate inter-rater reliability (Kα = 0.514) suggests potential inconsistency in policy/guideline distinction; Temporal bias: Document collection period (October 2024 to January 2025) captures only a snapshot; policies continue to evolve; Publication bias: Exclusion of 130 documents that did not meet policy criteria may underrepresent emerging guidance; Observer bias: Three reviewers' interpretations of policy criteria, despite calibration exercise

Limitations

"First and foremost, the results and discussion reflect available AI use guidance until 3/1/2025
It is likely that some ASPPH member schools have since developed or enhanced their policies or guidelines to the point that they would have warranted inclusion into our final corpus of policies." Additionally, "many schools circulate AI policy internally and do not formally publish AI guidelines on their websites." Furthermore, "the study analyzes publicly available institutional documents and therefore examines formal policy language rather than institutional practice or enforcement
Policies represent official statements of institutional expectations and governance structures but may not fully capture how AI use is implemented or regulated in practice." The inter-rater reliability analysis showed "low to moderate agreement" with Krippendorf's Alpha of 0.514, and "it is possible that stakeholders outside of the research team might have different opinions on whether a certain school's guidance should be considered a guideline or policy."

Open questions raised

Limited research on AI governance specifically within public health education
Gap between formal policy guidance and actual institutional practice/enforcement
Lack of clarity on intellectual property rights and ownership in the age of AI
Insufficient guidance on how institutions define and address AI misuse
Minimal attention to community engagement and AI use
Absence of comprehensive frameworks addressing the intersection of privacy and AI ethics beyond caution/restriction

Extracted from: pdfAgreement 61%

Explore related topics

Institutional Policy AI Governance Mixed Methods