Software nowadays often has codebases with millions of lines of code. This presents significant challenges for software engineers in understanding, changing, and quality-assuring these large codebases as well as associated requirements specifications, design documents, and other documentation.
To support them in these challenges, software engineers have developed artificial helpers, known as software bots or agents, to automate tasks. While bots were once static with limited interactions, the rise of large language models (LLMs) has transformed them, enabling natural interactions between engineers and bots, allowing them to become virtual team members. However, this AI teaming raises new questions and challenges: How much do engineers accept and trust these new team members? How often and by whom should interactions be initiated? Are there individual differences in preferences and impacts on productivity? Answering these questions is crucial for designing the next generation of software bots to enhance productivity.
This project brings together the expertise of software engineering and psychology researchers to run a series of case studies with students and practitioners to get empirical answers to these questions. The outcomes will inform the design and creation of a technical framework that increases trust and productivity in the interactions between software engineers and bots. It will integrate trust measurement, interaction optimization, and personalization components into a cohesive system. The results will lay the foundation for a new level of productivity in software engineering collaboration.