What is it about?
AI assistants based on Large Language Models (LLMs), such as ChatGPT, are now part of the daily work of many software engineers. Still, software development remains a human activity, so it is important to understand why some professionals adopt these tools and others do not. In this study, we looked at the factors that lead software engineers to adopt LLMs, with a focus on their cultural values—the beliefs and preferences people develop in the culture where they live and work. We combined a well-known model of technology acceptance (UTAUT2) with Hofstede's cultural dimensions, and we collected survey data from 188 software engineers. We then used a statistical technique (PLS-SEM) to test how these factors relate to each other.
Featured Image
Photo by Annie Spratt on Unsplash
Why is it important?
We found that two factors drive the use of LLMs the most: performance expectancy—the belief that LLMs help people work better and faster—and habit. Cultural values did not significantly change these relationships. For companies, this means that the same approach can work in culturally diverse teams: show clearly how LLMs improve performance and efficiency, offer training programs that demonstrate their benefits, create an environment that supports regular use, and keep tracking and sharing the improvements over time.
Perspectives
This paper connects two lines of my research: the role of national culture in software engineering, which was the focus of my PhD, and the adoption of generative AI. We expected cultural values to change how developers adopt LLMs, but the data told a different story—habit and perceived usefulness mattered much more. I find this kind of result useful, because it tells companies that the same adoption strategy can work across cultures.
Prof. Stefano Lambiase
Aalborg Universitet
Read the Original
This page is a summary of: Investigating the Role of Cultural Values in Adopting Large Language Models for Software Engineering, ACM Transactions on Software Engineering and Methodology, December 2025, ACM (Association for Computing Machinery),
DOI: 10.1145/3725529.
You can read the full text:
Contributors
The following have contributed to this page
