Open-source Community Sustainability using Agent-based Models
2023 / 03
Abstract
Abstract: Collaborative open-source projects require constant maintenance and involvement, but these factors are usually evaluated after the fact, leading to technical debt and project failures. We attempt to resolve this state of affairs by exploring the use of agent-based models to model contributor behavior over long periods of time. Our approach draws from a diversity of approaches, including active inference, reinforcement learning, and cybernetic personality theory. The agent-based approach allows us to model various attributes of collective behavioral patterns generated in the course of open-source community interactions. These models allow for the ability to audit open-source sustainability, in addition to long-term contingency planning for open-source communities. This research program has led to the creation of an auditing tool that serves as a template for future open-source development and research. Introduction How do we make open-source projects more sustainable in terms of community involvement and long-term success? The use of data analysis of metrics from specific platforms, either alone or in combination, captures only a portion of information about long-term involvement in open-source projects (Trujillo et.al, 2021). However, it does not provide us with alternate scenarios of community evolution. We propose that agent-based models (ABMs) can be used to understand future trends given the current state of an open-source community. We present three distinct models (active inference, reinforcement learning, and Big 5 personality theory) that approach the multifaceted nature of open-source communities.