Date of Award

5-7-2026

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Information Science

First Advisor

John R. Talburt

Abstract

The rapid adoption of Generative Artificial Intelligence (GenAI) and Large Language Models (LLMs) has intensified longstanding data quality concerns while exposing important gaps in current Data Quality Management (DQM) practice. This dissertation examines whether widely used DQM frameworks sufficiently address semantic data quality dimensions that are critical for trustworthy GenAI and LLM applications. Guided by Information Quality (IQ) theory, particularly the Wang and Strong conceptualization of data quality, the study evaluates selected DQM frameworks used in practice to determine the extent to which they emphasize semantic dimensions such as accuracy, value-added, relevancy, objectivity, reputation, believability, security, and appropriate amount of data. The study also investigates the persistent misconception that data validation alone is equivalent to data accuracy, especially when validation is not performed against an authoritative data source or system of record. A qualitative, multi-phase research design was used. First, a comparative document analysis was conducted on nine widely referenced DQM frameworks to identify the degree to which theoretical data quality dimensions are represented in practice. Second, a two-round Delphi study was conducted to validate the framework analysis and obtain expert consensus on the implications of the findings. In Round 1, fifteen expert panelists responded to twenty-three structured statements and seven open-ended questions concerning the adequacy of current DQM practice in the context of GenAI and LLM adoption. In Round 2, fourteen experts ranked the relative importance of semantic dimensions and evaluated practical measures and approaches for incorporating them into DQM activities. Descriptive statistics were used to summarize Round 2 rankings, while thematic analysis was used to synthesize qualitative feedback from both rounds. The findings indicate that current DQM practice continues to place heavier emphasis on syntactic dimensions such as validity, completeness, consistency, timeliness, and uniqueness than on semantic dimensions required for context-sensitive, trustworthy AI outputs. Delphi panelists strongly agreed that DQM practices must evolve in response to GenAI and LLM adoption, and they consistently identified accuracy, relevance, trustworthiness, and context-sensitive evaluation as high priorities. The findings further suggest that human review, cross-source validation, and authoritative reference data remain indispensable, even where automation is used. This study contributes by documenting the gap between IQ theory and IQ practice, validating that gap through expert consensus, and offering recommendations for strengthening DQM and data governance practices without proposing a wholly new framework. The dissertation concludes that organizations can improve trust, transparency, and risk management in GenAI and LLM implementations by increasing the priority of semantic data quality dimensions within existing DQM programs.

Download

Share

COinS