Mapping a Landscape of Developer Assisting Software Bots

Licentiate thesis, 2021

Bots in software development have gained traction in research and in practice. However, there is no consensus on what properties and characteristics that define a bot. The term is used to describe a plethora of different tools with different usages, benefits and challenges. In this thesis we focus on bots for software developments (DevBots) with the goal to aid researchers in future studies involving DevBots. We aim to assist with the scoping and planning of such studies regarding what tools and related work to include or exclude from them. We do so by synthesising the different definitions of DevBots, combining views from literature and practitioners.

To achieve this goal, quantitative and qualitative research methods are used including literature review and semi-structured interviews. We have created a faceted taxonomy for DevBots which categorises DevBots by their most prominent properties. In addition we investigated what delineated DevBots from plain old development tools. Our analysis shows that achieving one single definition is not possible. Instead we identify and name three personas, i.e., practitioner archetypes with different expectations and motivations. The chat bot persona (Charlie) mostly sees DevBots as information integration tools with a natural language interface, while for the autonomous bot persona (Alex) a DevBot is a tool that autonomously handles repetitive tasks. Lastly, for the smart bot persona (Sam), the defining feature of bots is its degree of ``smartness''.

We have identified a process in the form of a flowchart, which researchers can use to test whether their tool is considered a DevBot by any of our personas. We have concluded that this definition is not congruent with contemporary definitions as only 10 of 54 investigated tools from a large dataset were considered DevBots by our process. Finally we have shown how the definitions and process can be used in practice by using them in the scoping and planning phase of two recently conducted studies.