Search Results

The process of cognitive search invokes a purposeful and iterative process by which an organism considers information of a potentially diverse nature and selects a particular option that best matches the appropriate criteria. This chapter focuses on the neurobiological basis of such a goal-directed search by parsing the process into its main components, suggested here as initiation, identification of search space, deliberation, action selection, and evaluation and search termination. Unexpected uncertainty is suggested as a key trigger for the onset of the search process. Current data posit that this is represented in the anterior cingulate, parietal, and inferior frontal cortices, suggesting these areas could be particularly important in search initiation. A change in motivational state, likely signaled by a wide range of brain regions including the amygdala, can also play a role at this stage. The neural structures which represent the set of to-be-searched options may vary depending on the search domain (e.g., spatial, visual, linguistic). During deliberation, predictions regarding the consequences of selecting these options are generated and compared, implicating areas of frontal cortex as well as the hippocampus and striatum, which are known to play a role in different aspects of outcome evaluation. Action planning and selection likely involve an interplay between the prefrontal cortex and basal ganglia, whereas search termination could involve the specific neural networks implicated in response inhibition. The influence exerted over the search process by the major ascending neuromodulators (dopamine, norepinephrine/noradrenaline, serotonin, and acetylcholine) is also considered, and a particularly critical role suggested for dopamine and noradrenaline, given their ability to influence cognitive flexibility and arousal. Finally, pathologies of search processes are discussed, both with respect to brain damage and psychiatric illness.

This chapter discusses commonalities and differences in the cognitive mechanisms underlying different search tasks, such as spatial search, visual search, memory retrieval, action search, problem solving, and decision making. Three key issues relevant across all types of search are distinguished: (a) the initiation of search, (b) the maintenance and adaptive modification of the search process, and (c) the termination of search. As to search initiation, research is summarized concerning the effect of the number of cues on difficulty for executing search, and which factors structure the cue hierarchy. Discussion follows on how knowledge about metacognitive processes in memory might be used for better understanding the processes in maintenance of search, and heuristic principles for stopping search, possibly shared across different search tasks, are identified. Finally, consideration is given to how search processes might change as a function of experience and aging.

Human goal-directed action emerges from the interaction between stimulus-driven sensorimotor online systems and slower-working control systems that relate highly processed perceptual information to the construction of goal-related action plans. This distribution of labor requires the acquisition of enduring action representations; that is, of memory traces which capture the main characteristics of successful actions and their consequences. It is argued here that these traces provide the building blocks for off-line prospective action planning, which renders the search through stored action representations an essential part of action control. Hence, action planning requires cognitive search (through possible options) and might have led to the evolution of cognitive search routines that humans have learned to employ for other purposes as well, such as searching for perceptual events and through memory. Thus, what is commonly considered to represent different types of search operations may all have evolved from action planning and share the same characteristics. Evidence is discussed which suggests that all types of cognitive search—be it in searching for perceptual events, for suitable actions, or through memory—share the characteristic of following a fixed sequence of cognitive operations: divergent search followed by convergent search.

Generally speaking, two conditions make cognitive search possible: (a) symbolic structures must be present in the environment and (b) these structures must be detectable by a searcher, whose behavior changes based on the structures detected. In this chapter, information search on the Internet is used to illustrate how a theoretical framework of these two conditions can assist our understanding of cognitive search. Discussion begins with information foraging theory (IFT), which predicts how general symbolic structures may exist in an information environment and how the searcher may use these structures to search for information. A computational model called SNIF-ACT (developed based on IFT) is then presented and provides a good match to online information search for specific target information. Because a further component important to cognitive search is the ability to detect and learn useful structures in the environment, discussion follows on how IFT can be extended to explain search behavior that involves incremental learning of the search environment. Illustration is provided on how different forms of semantic structures may exist in the World Wide Web, and how human searchers can learn from these structures to improve their search. Finally, the SNIFACT model is extended to characterize directed and exploratory information foraging behavior in information environments.

Search is a common and crucial behavior for most organisms. It requires individuals to achieve an adaptive trade-off between exploration for new resources distributed in space or time and exploitation of those resources once they are found. Common to so many aspects of our lives, search behavior has been studied in a diverse range of scientific disciplines and paradigms: theoretical biologists study the characteristics of evolutionary search in high-dimensional spaces; behavioral ecologists analyze animals foraging for food; experimental psychologists investigate search in vision, memory, decision making, and problem solving; neuroscientists study the neural mechanisms of goal-directed behavior in humans and other animals; psychiatrists and clinical neuroscientists analyze aberrant volition such as drug-seeking behavior in addiction and attentional control in attention deficit hyperactivity disorder; computer scientists develop information search algorithms for mining large-scale databases and for individual navigation of the World Wide Web; social psychologists investigate how people seek and choose mates and friends; and political scientists study how groups look for solutions to problems. The need to integrate these insights further has led to the current book, which provides a cross-cutting perspective on the commonalities of cognitive search in different search domains.

Search can be defined as an attempt to arrive at a goal at an unknown location in the physical environment, as well as in time, memory, or any other space. Search is necessary because the quantity and quality of resources essential to survival and reproduction vary in space and time. In addition to exploration through actual body movement in their environment, animals search their external information space through selective allocation of attention and their internal information space to retrieve relevant items from memory. This chapter integrates data on search in three distinct domains—physical movement, attention to external information, and locating items in memory—to highlight the remarkable similarities between these three domains. First, resources in all three domains are typically distributed in patches. Second, in each of the three domains, animals typically keep searching in patches where they have recently found resources and leave areas when none are found or where they have already depleted the resources. Third, the neurobiological mechanisms modulating the exploration for and exploitation of resources in all three domains involve dopamine as well as, in many vertebrates, regions of the prefrontal cortex and striatum. It is suggested that, throughout evolution, animals co-opted existing strategies and mechanisms used to search their physical space for exploring and exploiting internal and external information spaces. The cross-disciplinary integration of theory and data about search can be used to guide future research on the mechanisms underlying cognitive search.

At a basic level, cognitive search involves several parameters: Under what circumstances should a search be initiated, and how should the goal be specified? What are the criteria by which the search is judged a success or failure? How are corrective actions implemented when search strategies are judged insufficient? Studies of cognitive control have the potential to address each of these questions. In this chapter, a number of issues related to executive control of search are discussed, including the way in which hierarchical search goals are monitored and updated. A new theory of cognitive control is proposed to begin to answer these questions, and open questions that remain are highlighted for future enquiry.

This chapter reports the discussion of a group of mostly behavioral biologists, who attempt to put research on search from their own discipline into a framework that might help identify parallels with cognitive search. Essential components of search are a functional goal, uncertainty about goal location, the adaptive varying of position, and often a stopping rule. The chapter considers a diversity of cases where search is in domains other than spatial and lists other important dimensions in which search problems differ. One dimension examined in detail is social interactions between searchers and searchers, targets and targets, and targets and searchers. The producer-scrounger game is presented as an example; despite the extensive empirical and theoretical work on the equilibrium between the strategies, it is largely an open problem how animals decide when to adopt each strategy, and thus how real equilibria are attained. Another dimension that explains some of the diversity of search behavior is the modality of the information utilized (e.g., visual, auditory, olfactory). The chapter concludes by highlighting further parallels between search in the external environment and cognitive search. These suggest some novel avenues of research.

This chapter examines the role that networks play in facilitating or inhibiting search for solutions to problems at both the individual and collective levels. At the individual level, search in networks enables individuals to transport themselves to a very different location in the solution space than they could likely reach through isolated experimental or cognitive search. Research on networks suggests that (a) ties to diverse others provide a wider menu of choices and insights for individuals, and (b) strong ties will be relatively more useful for complex information, and weak ties for simple information. At the collective level, these conclusions become less clear. The key question is how the collective operates to coordinate within the group versus beyond it so as to balance experimentation and convergence toward a solution. Collective coordination of search and collective evaluation of potential solutions may significantly influence the optimal network structure for collective problem-solving search.

All animals, including humans, search for a variety of different things in their natural environment, from food to mates to a suitable place to live. Most types of search can be represented as stopping problems of varying complexity, in which the animal has to decide when to stop searching and accept the current option. All forms of search take time, and in solving a stopping problem the animal has to trade off this time cost against the expected benefits of continuing to search. This chapter discusses two main approaches to predicting search behavior: the optimality approach and the heuristics approach. The optimality approach identifies the best possible solution to a search problem and thereby sets an upper bound to what natural selection can achieve. The heuristics approach considers simple decision algorithms, or “rules of thumb,” which animals may use to implement efficient search behavior. Although few studies have tried to integrate these functional and mechanistic perspectives, they are likely to provide complementary insights. Often, the form of an optimal strategy suggests which kinds of heuristics might be expected to evolve. Stopping problems may be simple, repeated, or embedded in other stopping problems. For example, if searchers assess the value of each encountered option by examining a series of cues, the assessment process can be considered as another stopping problem. When the searcher is uncertain about the environment it is in, its previous experiences during search can strongly influence the optimal behavior. Where a limited number of items can be accepted, as in mate search, a key constraint is whether the searcher can return to previously encountered items. Some search problems are complicated by the fact that the encountered items are themselves searching. The chapter concludes with a discussion of some open questions for future research.

The importance of understanding human memory search is hard to exaggerate: we build and live our lives based on what we remember. This chapter explores the characteristics of memory search, with special emphasis on the use of retrieval cues. We introduce the dependent measures that are obtained during memory search, such as accuracy and search time, and discuss how these have contributed to our understanding of human memory search. The three phases of memory search (initiation, progression, and termination) are discussed in relation to the strategies employed by the human retriever. Finally, the experimental paradigms used in the memory literature are compared to examples of animal foraging behavior to identify points of contact for developing a general cross-domain understanding of search processes.

Directing our behavior adequately to current goals requires a trade-off between cognitive flexibility and cognitive stability. In this chapter, empirical data and theories are reviewed which show that this trade-off depends on optimal modulation of frontostriatal circuitry by the major ascending neuromodulatory systems of dopamine, noradrenaline, and acetylcholine. Highlighted are the roles of dopamine in (a) the prefrontal cortex in the stabilization of goal-relevant representations and (b) in the basal ganglia in the flexible updating of those representations. The cognitive neurochemistry of cognitive flexibility is, however, complex, with different forms of flexibility implicating subcortical and/or cortical dopamine, noradrenaline, and/or acetylcholine. The review concludes with a number of open questions raised by attempts to reconcile the different, complementary theories about the neurochemistry of the flexibility-stability trade-off.

An essential facet of adaptive and versatile behavior is the ability to prioritize actions in response to dynamically changing circumstances, in particular when circumstances require the coordination of a planned course of action vis-à-vis instantaneous urges and extraneously triggered reactions. This chapter focuses on one aspect of cognitive search: the exploration of internal and external milieu for motivationally salient events (stimuli that are novel, ambiguous, infrequent, deviant, or unexpected, or register as a risk for undesirable outcomes or a risk for the exhaustion of resources) which may require appropriate adaptive action. A neurocognitive framework is described for understanding how cognitive control and cognitive search are modulated by motivationally salient events. This framework emphasizes the integration of a salience network in the brain with other large-scale neural networks, neurotransmitter systems, and homeostatic (autonomic nervous system) functioning. The anterior insula cortex and anterior cingulate cortex are core nodes of a salience network that monitors for motivationally salient stimuli. This framework helps to amalgamate findings from disparate literatures into a common conjecture and highlights the role of motivational salience in modulating cognitive search and cognitive control. The salience network transforms salience signals into an orienting response which serves to recruit the necessary physiological arousal and to engage task-relevant networks (involving attentional, working-memory, and adaptive action selection processes) while disengaging task-negative networks. Using representative examples as instructive points in case, it is argued that this integrative systems-neuroscience framework provides a parsimonious account of salience processing, and may provide novel insights into the neural basis of individual differences among healthy as well as pathological populations.

One oft-envisioned function of search is planning actions (e.g., by exploring routes through a cognitive map). Yet, among the most prominent and quantitatively successful neuroscentific theories of the brain’s systems for action choice is the temporal-difference account of the phasic dopamine response. Surprisingly, this theory envisions that action sequences are learned without any search at all, but instead wholly through a process of reinforcement and chaining. This chapter considers recent proposals that a related family of algorithms, called model-based reinforcement learning, may provide a similarly quantitative account for action choice by cognitive search. It reviews behavioral phenomena demonstrating the insufficiency of temporal-difference-like mechanisms alone, then details the many questions that arise in considering how model-based action valuation might be implemented in the brain and in what respects it differs from other ideas about search for planning.

Deliberation entails the sequential, serial search through possible options. This means that deliberation requires a mechanism to represent the structure of the world, from which predictions can be generated concerning these options and the expectations of the consequences of taking those options. Deliberation requires a mechanism to move mentally through those predictions as well as a mechanism to evaluate and compare those predictions. Neural signals for each of these factors have been found in the rat.

This chapter offers a selective review of behavioral and ecological perspectives on search behavior. Basic results from foraging theory are presented and their relationship to search is discussed. Techniques for the statistical description of searching motion are outlined, with a focus on the correlated random walk and the so-called Lévy flights—a technique that holds considerable promise. The problems of search in groups are reviewed at several levels. Both cooperative search (as conducted, e.g., by members of a social insect colony) and group movements of extremely selfish animals are considered. Finally, a review is provided of the producer-scrounger game, which considers the interactions within groups when some individuals parasitize the search behavior of others. The implications of these ideas are discussed and potential future directions for future enquiry are highlighted.

This chapter considers the range of visual search tasks, from those involving very briefly presented stimuli to those involving search processes that extend over many days. Most of the discussion centers on “classic” visual search tasks, as studied in the lab. Here, observers look for targets in displays of varying numbers of distractor items. The efficiency of these search tasks is driven by how effectively attention can be guided toward target items. Guidance, in classic search, is based on preattentive processing of a limited set of attributes (e.g., color, size). Thus, if the target is known to be red, attention can be guided to red items. If it is known to be big and red, both features can guide attention. Some of the rules of the human visual search engine are described and consideration is given to how these rules apply or change when moving from “classic” search tasks to real-world search tasks. Connections to other search literatures, including foraging and memory search, are highlighted.

Since Alan Turing, computer scientists have been interested in understanding natural intelligence by reproducing it in machine form. The field of artificial intelligence is characterized, to a large extent, by search algorithms. As search is a computational process, this too has been well studied as part of theoretical computer science, leading to famous results on the computational hardness of problems. This chapter provides an overview of why most search problems are known to be hard and why general search strategies are impossible. It then discusses various heuristic approaches to computational search. The fundamental message intended is that any intelligent system of sufficient complexity, using search to guide its behavior, should be expected to find solutions that are good enough, rather than the best. In other words, it is argued that natural and artificial brains should satisfice rather than optimize.

Considerable evidence suggests that the behavioral mechanisms for instrumental action selection are mediated by two distinct learning processes: a goal-directed process whereby actions are selected with reference to the incentive value and causal relationship between actions and associated outcomes, and a more reflexive habitual process in which actions are elicited by antecedent stimuli without any consideration of the associated outcome. This chapter reviews evidence from experiments in both rodents and humans which suggests that the behavioral dichotomy between these two modes of action selection are also reflected at the neural level, involving at least partly dissociable regions: a circuit involving the medial prefrontal cortex and dorsomedial striatum is implicated in goal-directed learning, whereas a region of posterior lateral dorsal striatum is implicated in habitual learning. Building on the arguments put forward by Winstanley et al. (this volume), it can be concluded that the specific neural circuits identified as contributing to goal-directed learning, but not those involved in habit learning, are a constituent element of the neural systems underlying cognitive search.