Examines a statistical model in which the users of an online system continually update their estimated probability of success, and quit or continue the search according to the expected utility of each action. The implications for search strategies are discussed. (Author/EM)

Illustrates how various key abstractions of information finding, such as document relevance, a desired number of relevant documents, discouragement, exhaustion, and satisfaction can be modeled using the Petri Net framework. Shows that this model leads naturally to a new approach to collection of user data, and to analysis of transaction logs.…