The Importance of Replication In Uncertain Epistemic Landscapes

Abstract

Scientific research is mainly done in groups of scientists working in different parts of a specific research domain. This division of work is called the division of cognitive labour. The division of labour is originally interested in scientists that do different tasks in science. This thesis focuses on different attitudes of scientists and how they cooperate. We present an agent-based model of scientific research in which scientists are divided to explore and exploit unknown areas on an uncertain scientific landscape. The model is a complexification of the models from Muldoon \& Weisberg, and Thoma. Scientists aim to find and correctly identify the best approaches for a research problem. However, outcomes of scientific research can be erroneous. Therefore, replication is needed to verify what approaches are best. Failing to replicate is not a direct indicator of false results, but one should pay caution when one cannot reproduce outcomes of scientific research. Three distinct agent types for searching approaches are considered. These types have their unique search rules. The three agents are called: followers, mavericks and replicators. Followers look out for successfully done research and incrementally expand on this knowledge. Mavericks find areas that are not explored already. Lastly, replicators help to correctly identify the best approaches by replicating experiments. As a result, they take away uncertainty. A balance between exploring the landscape and replicating approaches is needed. The model shows that mainly having followers with a low proportion of replicators working on a scientific problem is beneficial. Mavericks are only helpful when scientists are inflexible.