A Referral Approach to Finding Medical Informatics Reviewers

We proposed and investigated a referral approach to finding experts in distributed networked environments. An expert finding task, namely, finding ¿good¿ medical informatics reviewers in peer-review processes, was used in an agentbased simulation environment to evaluate the effectiveness, efficiency, and scalability of the proposed model. Experiments on a coauthorship network of 181 peers showed that the model based on topical relevance cues was able to identify experts within a very short referral chain and largely outperformed random walks. Applying heuristics to rewire the connections and to constrain the use of remote neighbors showed large improvement on efficiency and provided a better network underpinning for finding shortcuts to experts. The study showed results consistent with previous research on navigation in small worlds and focused on IR applications with richer dimensions of complexity.