Keith James Holyoak - US grants

This award will support the activities of an interdisciplinary team of researchers in terms of maintenance of the Conneciton Machine of the UCLA Cognitive Science Research Program. The research resulting from this support will be in the following six areas: (a) language processing, which includes reasoning, planning, inference and search, (b) evolutionary studies in the newly emerging paradigm of artificial life, (c) simulation studies of neural networks - both synthetic and neurally plausible - for modeling memory, learning, language and vision, (d) parallel algorithms for biological image processing, for example, the manipulation of MRI images, (e) research in the fundamental mathematics and logic of parallel computation, such as the development of parallel programming language semantics, and (f) the use of parallel algorithms for scientific computing. This award will support the activities of an interdisciplinary team of researchers in terms of maintenance of the Conneciton Machine of the UCLA Cognitive Science Research Program. The research resulting from this award will be in cognitive sciences and in the development of parallel computation.

9310614 HOLYOAK This research will investigate the role of analogical similarity in guiding retrieval of previous experiences stored in memory. Analogical similarity depends on overlap in the underlying structure of situations, such as a common theme shared by two stories or a common goal structure underlying two different problems. Such "deep" similarities can be distinguished from more superficial types of overlap, such as similar characters and events in stories or similar objects involved in problems. Previous research has shown that access to memories is strongly influenced by such surface similarities, but has generally failed to find evidence that deeper types of similarity can allow one situation to trigger reminding of another. This project will use a variety of more sensitive measures to determine whether and when human memory is guided by deep as well as superficial similarities between complex situations. The influences of different types of similarity will be investigated by experiments in which college students process a series of stories in a context that does not make it apparent that a memory test will follow. After a delay ranging from a few minutes to a week, the subjects will be cued with other stories that differ in their similarity relations to the target stories in memory. In some cases a cue will be related to multiple target stories in memory. It is expected that the retrieval competition triggered by having multiple related targets in memory (a situation more in keeping with everyday experience that is the paradigm used in previous studies) will greatly improve the sensitivity of reminding to analogical similarity. Other experiments will use more indirect measures to determine which types of similarity influence the way in which information is processed spontaneously. These studies will determine whether reading a story that is similar at a deep level, a surface level, or both, will render a subsequent story more comprehens ible, or decrease the time required to read and understand it. These experiments will provide knowledge important for several related topics. Few current models of memory retrieval can account for an influence of deep similarity on reminding. If the project reveals that analogical similarity indeed influences memory access, this evidence would place important constraints on any adequate theory of the way complex knowledge is represented in and accessed from human memory. The results will also provide guidance for efforts to devise more psychologically realistic computer-based models of memory retrieval. In addition, sensitivity to analogical similarity can provide a basis for flexible learning and transfer in problem solving. The results are therefore likely to have important implications for training and education. ***

9511504 HUMMEL This research will develop a computer system to model how people learn schemas from examples. Schemas are knowledge structures that describe general situations, events, rules, or relationships. For example, a Restaurant Schema might specify the kinds of events that take place in restaurants, and the relationships between the actors in those events (e.g., chefs, waiters, and customers, etc.); a Family Relation Schema might specify the kinds of relationships that define various family members, such as sisters, brothers, parents, uncles, etc. A schema is very general, in that it refers to many specific situations, and also highly structured, in that it specifies the relationships between objects, events, or even between other relationships. Psychologists and computer scientists have long recognized the utility of schemas as a basis for reasoning: If one can recognize an object or situation as an instance of a general schema, then one can use the schema to reason about the specific object or situation. For example, the Family Relation Schema tells us to expect that Mom's brother, Bob, is our uncle, and that if Bob has children, then they will be our cousins. Due to the schema, it is not necessary to learn the family relations for every individual separately. Although the generality and structured nature of schemas make them extremely useful forms of knowledge, they also make them very difficult to model: Traditional symbolic approaches to computer modeling are good at representing structured information, but they have difficulty flexibly matching specific instances to general facts; by contrast, connectionist models are good at flexibly matching instances to general categories, but they have great difficulty representing structured information (such as rules and relationships). Perhaps for this reason, no one has ever developed a formal and general model of schema-based learning and reasoning. Hummel and Holyoak have developed a computer mo del of analogy that combines a connectionist architecture with the capacity to represent and learn relational structures. The model's capacity to represent structured information in a flexible, general manner makes it an ideal vehicle for simulating human schema learning and use. The research will apply the model's approach to the representation of structure to the problems of representing, learning, and using schemas. The result will be a model that can (a) learn general schemas from specific examples, (b) match new instances to learned schemas, and (c) use the schemas to make inductive inferences about the new examples. The model will be the first formal theory of human schema-based reasoning, and will contribute substantially to our understanding or human reasoning is a wide variety of domains. It will also be the first working computer system to use schemas to reason flexibly about general knowledge domains. ***

This project will extend and test a theory and computer model of human reasoning. People routinely reason using schemas, knowledge structures that describe general situations, events, rules, or relationships. For example, a Restaurant schema might specify the kinds of events that take place in restaurants, and the relationships between the actors in those events (e.g., chefs, waiters, customers); a Family Relation schema might specify the kinds of relationships that define various family members, such as sisters, brothers, parents, and uncles. A schema is both general, in that it refers to many specific situations, and highly structured, in that it specifies the relationships between objects, events, or even between other relationships. Psychologists and computer scientists have long recognized the utility of schemas as a basis for reasoning: If one can recognize an object or situation as an instance of a general schema, then one can use the schema to reason about that object or situation. Although the generality and structured nature of schemas make them extremely useful, these features also make schemas very difficult to model. Traditional symbolic approaches to computer modeling are good at representing structured information, but they have difficulty in flexibly matching specific instances to general facts; by contrast, connectionist models are good at flexibly matching instances to general categories, but they have great difficulty representing structured information (such as rules and relationships). We have recently developed a theory of human schema-based learning and reasoning, and implemented this theory as a working computer model. The model, called LISA (Learning and Inference with Schemas and Analogies), combines the flexibility of a connectionist architecture with the capacity to represent and learn relational structures. The resulting system can (a) learn general schemas from specific examples, (b) match new instances to learned schemas, and (c) use the schemas to make inductive inferences about the new examples. LISA makes several novel predictions about human learning and reasoning. Part of the NSF-supported research will be a series of experiments to test these predictions. Another part of the research will extend the LISA model to account for aspects of human story and event comprehension (e.g., If someone tells us `There are four plates per tray,` how do we know that this means each tray will hold, or contain, four plates?), and aspects of spatial reasoning (e.g., If we are told that Bill is taller than Charles and Abe is taller than Bill, how do we figure out that Abe is taller than Charles?). Importantly, the extended LISA model will account for these seemingly different capacities (as well as other capacities, such as reasoning by analogy) in terms of the same basic mechanisms as it uses to account for our ability to reason using schemas. The knowledge gained from the project will contribute to the development of methods for enhancing human learning and reasoning abilities, and also to the development of artificially intelligent systems.

Abstract
SES-0080375
Holyoak, Keith J.
University of California - Los Angeles

Theories of human reasoning are heavily influenced by a number of assumptions derived from formal accounts of deductive logic. A central assumption of these models is that the flow of inferences is unidirectional. The syllogistic, unidirectional nature of the reasoning processes rules out "reverse" inferences in which a person's conclusions might lead to a change in the evaluation of the premises and evidence. Violations of this assumption are viewed as signs of the frailty of human reasoning. In this project, we explore an alternative conception of reasoning and decision-making, one based on a theoretical paradigm called constraint satisfaction mechanisms. Our preliminary research has shown that reasoning entails bi-directional influences among the participating pieces of evidence, premises and conclusions, and that the changes occur mostly without awareness.

This project is a novel attempt to introduce this emerging model of cognition to decision making, particularly in the legal domain. First, we intend to apply it to the issue of evidence integration in the fact-finding phase. We will examine whether the process of evidence integration leads to changes in the evaluation of the individual pieces of evidence. In particular, we will examine cognitive effects on judgments of defendants' mental states, a determination that is crucial to adjudication in tort and criminal law (mens rea). Other studies will examine why decision makers encounter problems with ignoring inadmissible evidence.

The making of a decision can be a difficult and daunting task, as is often manifested both before and after the decision is made, yet at the time of making the decision people generally feel confident, even overconfident. In the second part of the project we intend to explore this relationship among pre-decisional conflict, confident decisions, and post-decisional regret. This issue is especially pertinent to the legal domain, since a central feature of legal culture is that decisions are taken very seriously. Contract law, for example, is based on a concept of reliance, and a failure to fulfill an obligation is considered a breach. There is even less flexibility when it comes to adjudication, wherein decisions announced by judges and jurors are treated as virtually immutable. Given the weightiness of the issues and the closeness of the vying positions often involved in legal transactions and disputes, it is important that we gain a better understanding of these seemingly paradoxical phenomena of conflict, confidence and regret.

model design /development

The objective of this proposal is to study the cognitive processes that enable effective decision making in a category of tasks that require discrete choices in the face of variables that are numerous, contradictory, ambiguous and incommensurate. Decisions of this kind are encountered in everyday life when we choose a job, purchase a house, or condemn a person to prison. The proposed theoretical framework, termed "coherence-based decision making," posits that throughout the decision making process, the mental representation of the task undergoes gradual change, ultimately shifting towards a state of coherence with either one of the decision alternatives. At the culmination of the process, the decision maker's mental model is skewed towards a state of coherence, with one alternative dominating its rival; the decision follows easily and confidently from this perception of the task.
In this proposal the Investigators plan to buttress and extend their previous work on this theory. The fourteen studies included in this proposal are intended to examine theoretical questions and to provide prescriptions with respect to improving people's ability to respond appropriately to incoming information; better understanding failures to ignore information; reducing the effects of group polarization; improving the process of group deliberation; reducing self-serving biases; taking into consideration the influence of need states on biased judgments; and more. The project is intended also to provide specific prescriptions for decision making in a number of legal contexts, including suggesting ways to better structure legal tasks; better handle exposure to impermissible evidence; reduce jury polarization and improve jury deliberation; provide means for gauging and reducing pre-trial optimistic overconfidence; and provide critical insight into policy questions regarding the regulation of gambling.

Attention; Brain; CRISP; Cognition; Cognitive; Computer Retrieval of Information on Scientific Projects Database; Encephalon; Encephalons; Event; Exhibits; Functional Magnetic Resonance Imaging; Funding; Grant; Human; Human, General; Institution; Investigators; Light; MRI, Functional; Magnetic Resonance Imaging, Functional; Man (Taxonomy); Man, Modern; NIH; National Institutes of Health; National Institutes of Health (U.S.); Nervous; Nervous System, Brain; Participant; Photoradiation; Prefrontal Cortex; Problem Solving; Process; Range; Research; Research Personnel; Research Resources; Researchers; Resources; Scanning; Source; United States National Institutes of Health; Work; design; designing; experiment; experimental research; experimental study; fMRI; neural; neuroimaging; novel; relating to nervous system; research study

Despite recent advances in artificial intelligence, humans remain unmatched in their ability to think creatively. Intelligent machines can use massive data to learn to identify patterns that are similar to learned examples, but people can use very small amounts of data to discover deep similarities between situations that are superficially very different (e.g., engineers have devised a cooling system for buildings using principles adapted from termite mounds). This type of creative thinking depends on analogy: the ability to find and exploit resemblances based on relations among entities, rather than solely on superficial appearances. The present investigation aims to show how relations can be learned from examples (in the form of either texts or pictures) and then used to reason by analogy. The work integrates recent advances in machine learning with more human-like learning mechanisms. Improved analogy models will increase the power of computer-based information retrieval, allowing both text and pictures to serve as retrieval cues to search large databases for items that are analogous in relational structure. The large analogy datasets generated for the project will be made publically available. More flexible search engines will help to automate creative tasks such as engineering design. Identifying the computational basis for relation learning and analogical reasoning will guide development of artificial intelligence systems by providing more efficient learning mechanisms. The research team is integrating research and education activities by using this project as a training opportunity in interdisciplinary research, encompassing psychology, statistics, computer science and mathematics.

The research will integrate advanced computational approaches with behavioral experiments on human relation learning and analogical reasoning, using both texts and pictures as inputs. The work is guided by cognitive theory on learning and reasoning, and exploits recent advances in the field of machine vision. The project includes the creation and validation of multiple databases of analogy problems. Experiments will be performed to establish human performance levels in a variety of tasks. Computational models will be developed by synergizing big-data learning through deep networks with small-data learning through Bayesian modeling. Models will be evaluated by comparison with human benchmarks. By addressing issues that arise in reasoning from natural inputs such as texts and pictures, the models to be developed will generalize to situations that people encounter in their daily life.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

The ability to reason about the relations between sets of concepts?relational reasoning?gives rise to abstract thought, and has fueled some of humanity?s greatest achievements in science and technology. Although prior research has identified where in the brain relational reasoning takes place, this project pushes the research field by addressing how the brain represents abstract relations. Specifically, the project aims to address three key questions: (1) Can the brain represent an abstract idea independently of the concrete entities that comprise the content of the idea? (2) Do people represent concepts in an abstract manner only when explicitly required to do so, or are abstract relations also retrieved spontaneously? (3) What neural markers reliably predict differences in reasoning capacity between individuals? That is, do individuals whose brains represent abstract relations more readily also tend to have stronger reasoning skills, and/or to perceive meaningful connections that others miss? This project will identify the computational basis for abstract thought and reasoning, thereby creating an opportunity to refine artificial intelligence systems by providing them with more efficient learning mechanisms. This work will inform future research examining how children, and adults as lifelong learners, form representations of abstract concepts.

This project integrates recent advances in multivariate fMRI, computational modeling, and behavioral methodology to discover the neurocognitive mechanisms underlying the representation of abstract relations. Research will systematically examine the neural bases of this representation, as well as the influence of task context and individual differences. First, behavioral priming and neural similarity measures, alongside metrics from a computational model of relational reasoning, will characterize the overlap in representation between pairs of concepts that are only abstractly related. Second, manipulation of task demands will determine whether the magnitude, location, and stability of neural representations vary with explicit cognitive instructions. Finally, development of a novel 'neural score' metric will determine neural markers of individual differences in relational reasoning.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.