Well, we're all familiar with expert systems.  We know that an expert system can come close to, but is not as good as, the actual expert that it's modeled on.  Why is this?  We know that the knowledge base and rules do not capture the entire expertise of an expert.  In addition, an expert also relies on his or own experience and tacit knowledge.

When the expert is a scientist doing scientific research, the experience and tacit knowledge are also shaped by (1) common examples that his or her science requires its students to learn, (2) common examples that are published in journals, and (3) common examples that are discussed at conferences – where these common, shared examples are what Kuhn calls “exemplars.”  These exemplars "fill in" the gaps where evidence and scientific method alone are insufficient to tell a scientist how to do his or her research.

Exemplars can be shared and successfully used in applying theories only when the sociology of the research group is already in place – so it's important also to develop not only research methods and theories, but also the research culture and behavioral side of researchers.