Recalling information from memory can be thought of as a search process. Thus, an important aspect of recall is deciding when to stop searching. This is true whether remembering things that are fairly mundane, such as recalling the titles of one's favorite books, to remembering things that are crucial, such as a physician generating potential diagnostic hypotheses. Although the importance of criteria for search termination is recognized in most contemporary models of memory and decision making, very little is known about the processes that underlie these decisions. The research conducted by Dr. Dougherty and colleagues tests the applicability of a cost-benefit framework for describing both memory search and search termination decisions. Within this framework, costs are incurred by the act of searching (which takes time and effort), and gains are realized through successful outcomes. The notion that search termination decisions flow from a system that regulates costs and benefits raises two broad questions. First, in what ways do the costs and benefits of retrieval influence both search termination decisions and the time course of retrieval? Second, what role does monitoring the search process (a form of "metacognition") play? The working hypothesis is that appropriate search termination decisions enable one to maximize the utility of any particular recall attempt and that metacognitive monitoring plays an important role in this process.
Understanding the fundamental nature of search termination decisions has important implications for understanding the basic functioning of human memory and the role of memory in decision making. This is important for a variety of real-world contexts. For example, in time-critical triage situations, medical personnel must quickly retrieve potential diagnostic hypotheses from memory to render a treatment decision. Premature search termination may lead to an incorrect diagnosis whereas protracted search may compromise the care of other time-critical patients. This research may help improve information search efficiency and decision making quality.
Overview A common feature of many real-world tasks is the necessity of deciding when to terminate a search process. For example, in the task of luggage screening, a screener must decide when a bag has been sufficiently searched in order to declare it safe. Implicit in this task, is a decision to terminate the search of the x-ray. In a quite a different type of task, intelligence analysts search for possible explanations, or hypotheses, to explain newly acquired intelligence. In contrast to the luggage-screening task, which involves external information search, the generation of explanations involves internal information search – the retrieval of information from long-term memory. The overarching goal of the work supported under this grant is to understand the fundamental basis of search termination decisions. This program of research was initiated within the context of a larger theoretical question: How people search for and utilize information in decision-making contexts? Addressing this question, which is relevant for everything from luggage screening and medical diagnosis to intelligence analysis and artificial intelligence, requires a firm basic understanding of a variety of cognitive processes that are properly thought of as pre-decision processes. Although search termination decisions have received scant attention in the psychological literature until recently, this process is fundamental to any search task in which search the decision maker controls termination. Moreover, any realistic model of decision-making or artificial intelligent system must "know" the appropriate time-point to terminate search. Major Findings The primary findings from the research can be broken down into a set of experimental findings that describe the behavior of participants and computational modeling exercises that characterize the processes in terms of a cognitive model. A cognitive model is a computer program that instantiates our beliefs about how people search and terminate memory. The experimental findings revealed that peoples search termination decisions are remarkably lawful; the behavioral pattern appears to hold for both children and adults; the default rule that governs search termination essentially maximizes expected utility; and that the search termination process impacts other aspects of the retrieval process. Broader Impacts Though our work focused on stopping decisions in memory search, such decisions have been studied in other domains, such as convergent decision-making, problem solving, and information search tasks. It is interesting to note that search termination rule identified in our work is similar to stopping rules identified in the animal foraging literature, free recall tasks, as well as some visual search tasks. The fact that similar rules appear to be used across task domains suggests a common underlying mechanism. How this mechanism fits within the broader cognitive system, and how it can be accurately represented in cognitive models, is a topic of current research that is instrumental in refining our understanding of diagnostic reasoning in professional domains such as medical diagnosis, which will in turn lead to the creation of decision support systems that will enhance diagnostic accuracy. Throughout the course of this work, human resources were enhanced through the training of women and minorities. Three major contributors to this work were a female-minority undergraduate, a female graduate student, and a female post-doctoral student. All three of these students have moved on towards science related positions in academia or government. In addition, multiple other undergraduate students worked on the research and benefitted from the training they received.
