PLOS ONE has a paper titled "Scientific discovery in a model-centric framework: Reproducibility, innovation, and epistemic diversity" in which the authors used agent-based modeling to explore the idea of reproducibility of scientific results. To do so they created a framework of scientists' pursuit of a scientific truth, removed incentives and questionable research practices, and used model comparison rather than statistical hypothesis. They decided to use an agent-based model to analyze the replication process describing it as, "a forward-in-time, simulation-based, individual-level implementation of the scientific process where agents represent scientists." These are simple agent-scientists with no memory of their past decisions and no means to contact each other directly. They recognize that the simplicity of these agents is also one of the main limitations of their framework. They employ three types of scientist agents with different strategies and one replicator agent who repeats a previous experiment using new data. The authors describe the model in several pages and present their results. They conclude the link between reproducibility and the convergence to a scientific truth is not straightforward. It was possible to reproduce results that were never true and this could steer everyone in the wrong direction, so reproducibility and discovery of truth are not equivalent concepts. They also found that heterogeneous scientist populations better optimized the discovery process. This suggests cognitive diversity and interdisciplinary approaches will be more successful. You can read my take on the reproduction of agent-based models in an earlier blog posting.