Trust and collaboration in AI-assisted code review for software engineering teams
Guo, Jikang (2026)
Kandidaatintyö
2026
School of Engineering Science, Tietotekniikka
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi-fe2026052655382
https://urn.fi/URN:NBN:fi-fe2026052655382
Tiivistelmä
This thesis investigates the formation of trust and the support for cooperation in AI-assisted code review within software engineering groups. With the widespread application of tools like ChatGPT, Gemini and GitHub Copilot in software development, AI systems can assist in code generation, problem identification, testing, explanation and review. However, the suggestions from AI may be incorrect, hard to understand or inconsistent with the specific requirements of the project. Thus, much consideration is needed for AI-assisted code review since technical judgment is closely related to communication, knowledge exchange and duty division.
The study focuses on three aspects: factors influencing developers’ trust in AI-generated review suggestions, the effects of different developer-AI interaction modes, and the role of AI in team-level collaboration. A mixed-methods approach was used, including a questionnaire survey with 150 valid responses, 18 semi-structured interviews, and a case analysis of selected AI-assisted development and review-support tools. The findings show that developers do not simply accept or reject AI suggestions. Instead, they form conditional and calibrated trust based on technical accuracy, explainability, task risk, verification possibilities, and team norms. The thesis argues that AI should be regarded as a supportive assistant rather than an independent reviewer.
The study further shows that balanced trust requires both individual-level verification practices and team-level rules for responsibility, transparency, and final decision-making.
The study focuses on three aspects: factors influencing developers’ trust in AI-generated review suggestions, the effects of different developer-AI interaction modes, and the role of AI in team-level collaboration. A mixed-methods approach was used, including a questionnaire survey with 150 valid responses, 18 semi-structured interviews, and a case analysis of selected AI-assisted development and review-support tools. The findings show that developers do not simply accept or reject AI suggestions. Instead, they form conditional and calibrated trust based on technical accuracy, explainability, task risk, verification possibilities, and team norms. The thesis argues that AI should be regarded as a supportive assistant rather than an independent reviewer.
The study further shows that balanced trust requires both individual-level verification practices and team-level rules for responsibility, transparency, and final decision-making.