Date of Award

Summer 7-24-2024

Degree Type

Dissertation

Degree Name

Doctor of Business Administration (DBA)

Department

Computer Information Systems

First Advisor

Dr. Likoebe Maruping

Second Advisor

Dr. Po-An Hsieh

Third Advisor

Dr. Balasubramaniam Ramesh

Abstract

Generative Artificial Intelligence (GAI) augments human-machine collaboration and shifts enterprise Information Technology capabilities in the areas of innovation, automation, and engagement. More specifically, Large Language Models (LLM) have disrupted the software development field in the form of assistive tools supporting tasks from coding design and analysis to testing, deployment, and maintenance. Although there is extensive literature on employee technology acceptance, there has been little empirical investigation into developers’ post-adoptive intentions to explore use of GAI LLM tools under various conditions. The present research addresses this gap by integrating Net Valence Theory with elements of the Theory of Reasoned Action. An online survey instrument was used to measure individual perceptions pertaining to model output quality, model technology trust, model usage risk, and model usage benefit. The artifact-based beliefs and utility-based beliefs are hypothesized as predictors of intent to further explore using GAI LLM tools. The model is tested by conducting a serial and parallel multiple mediator analysis on data collected across 220 professionals. Results of the analysis confirm direct effects of quality, trust, and benefit as having a significant positive influence on exploration intent. In addition, sequential mediation between trust and benefit shows indirect effects of quality on exploration intent. Interestingly, the perception of risk was not significantly lowered by quality or trust, and high risk did not negatively influence exploration intention. Academic contributions expand theoretical knowledge on GAI LLM use in software development and validate a new contextualized framework. The findings help to explain cognitive tradeoffs that developers face in tool experimentation decision-making. It is the first study to further an understanding of the paradoxical phenomenon of broadening technology use when expected reward and risk are both heightened. The evidence has practical implications to better inform GAI LLM operationalization, regulation, and investment strategy. Insights may identify opportunities to define AI pair programming job value, enhance technology literacy programs, and fortify responsible data practices.

DOI

https://doi.org/10.57709/37422072

File Upload Confirmation

1

Download

Share

COinS