Q&A: A system for the Capture, Organization and Reuse of Expertise

It is a time-consuming and difficult task for an individual, a group, or an organization to systematically express and organize their expertise so it can be captured and reused. Yet the expertise of individuals within an organization is perhaps its most valuable resource. Q&A attempts to address this tension by providing an environment in which textual representations of expertise are captured as a byproduct of using the system as a semiautomatic questionanswering intermediary. Q&A mediates interactions between an expert and a question-asking user. It uses its experience referring questions to expert users to answer new questions by retrieving previously answered ones. If a user’s question is not found within the collection of previously answered questions, Q&A suggests the set of experts who are most likely to be able to answer the question. The system then gives the user the option of passing a question along to one or more of these experts. When an expert answers a user’s question, the resulting questionanswer pair is captured and indexed under a topic of the expert’s choice for later use, and the answer is sent to the user. Unlike previous work on question-answering systems of this sort, Q&A does not assume a fixed hierarchy of topics. Rather, experts build the hierarchy themselves, as their corpus of questions grows. One of the main contributions of this work is a set of techniques for managing the emerging organization of textual representations of expertise over time by mediating the negotiation of shared representations among multiple experts. MOTIVATION: FOUR PROBLEMS IN KNOWLEDGE MANAGEMENT The “war stories” that represent the experience and expertise of individuals are invaluable resources to novices in similar situations. Access to expert advice might allow a novice to adopt past solutions or avoid common pitfalls. Expert advice on how to solve a problem can mean the difference between arriving at a solution quickly and failing altogether. At the same time, once the availability of expert advice is made known to a large group of interested individuals and the expert receives more and more requests for the same kind of information, providing expertise becomes burdensome at best, and intolerable at worst. In such a situation, there is a direct conflict between the goals of the two parties involved. Specifically, the knowledge-seeking individual’s goal to have constant and uncompromised access to expert advice is in conflict with the expert’s goal to use his time for other things. We call this the problem of expert/novice goal conflict. In light of this, we would like to provide an environment that attempts resolve the above conflict by reaching a compromise in which both the expert and the information-seeking user are satisfied most of the time. Such a solution would maximize the availability of expert knowledge while minimizing the amount of effort and time required by the expert to provide advice to his colleagues. Unfortunately, when an expert is unavailable, so is access to their experience. Moreover, having the expert simply “write it all down” so that other people could benefit from the expert’s knowledge in their absence could take years and would require considerable motivation. We call this the knowledge acquisition bottleneck [1]. It is a timeconsuming and difficult task for an individual, a group, or an organization to systematically express and organize their expertise. A more realistic solution would entail incrementally capturing knowledge over time, as a byproduct of some other activity. Moreover, although knowledge may be available to an information-seeking individual, it may not be accessible. For example, consider an individual with a question to ask, and assume he does not know of anyone capable of answering it. Even though the answer to his question may be available to him (from his willing colleague in the office down the hall), the question will remain unanswered. We call this the resource discovery problem. In light of this second problem, we would like to provide an environment in which resources that could satisfy a request would be provided automatically. Such an environment would need to have an up-to-date model of the kinds of requests a resource might satisfy. Ideally, this model would not require manual construction. Budzik, J. and Hammond, K. J. (1999). Q&A: A System for the Capture, Organiation and Reuse of Expertise. In Proceedings of the Sixty-second Annual Meeting of the American Society for Information Science. Learned Information, Inc., Medford, NJ, 1999.