Developer Confidence, Competence, and Self-Efficacy in the Age of AI

Abstract

The emergence and incorporation of AI-based coding assistant tools in the software development process mark an unprecedented paradigmatic shift in the profession of software developers with significant implications for the essence and personality of problem-solving, learning, and software developers' profession. The current chapter undertakes an in-depth exploration of the intricate relationship between artificial intelligence assistants and the confidence, competency, and self-efficacy of software developers, relying on an exhaustive examination of the empirical literature on the subject. The current chapter draws on systematic reviews, large-scale surveys, and experimental studies that examined the impact of AI assistants, including GitHub Copilot, ChatGPT, and LLM-based coding assistants, on the productivity and learning of software developers. The Analysis of the findings presents a paradox: the ability of these AI assistants to improve the rate of task completion and cognitive ease in the performance of routine programming tasks contrasts with the risks of deskilling, over-reliance, and the erosion of problem-solving abilities, especially among novice programmers. In the chapter, the psychological facets of human-AI collaboration are investigated, including the role of self-efficacy beliefs as a mediator of tool adoption and use, and the impact of fear of failure. The new models of deskilling and upskilling are synthesized, and guidelines for the maintenance of developer competence in the face of AI assistants are proposed, which has important implications for software engineering education and developer skill building.

Ähnliche Arbeiten