Dedre Gentner - US grants

This research is concerned with the development and testing of psychological theories of similarity structure and similarity judgments. Similarity is a concept fundamental to virtually all aspects of cognition, theoretical and applied. For example, theories of memory rely on the observation that people often are reminded of similar events, and theories of learning predict transfer on the basis of similarity between learning and test conditions. In everyday life, the errors people make are often based on the use of inappropriate similarity matches, and a good deal of learning and transfer of training appears to be governed by what people find similar. Thus, an adequate theory of similarity is essential for theoretical advances in understanding cognition and in their application. This research stems from the fundamental insight that similarity of two things depends not only on the similarity of features or parts, but also on how those parts are put together, the relations among the parts. Previous research has demonstrated the advantages of formal mathematical models in understanding the role of attributes in similarity. The current research will apply and extend these previous formalisms to incorporate the contribution of relations to similarity, in addition to the role of attributes. Separating structure from processing principles is crucial for understanding a variety of experimental results, such as (1) that the similarity of object A to object B may be different from the similarity of B to A, (2) that the similarity of A to B may depend on what other objects are present, and (3) that adding the same property to objects A and B may change which one of the them is judged to be more similar to object C. Preliminary research indicates that these observations depend importantly on whether similarity is based on attributes, relations, or mixtures of attributes and relations. Most of the studies involve presenting people with material (e.g., geometric forms, descriptions of objects) and asking them to make similarity judgments. In other studies, reaction times and errors in various tasks will provide an indirect measure of similarity. Computer simulations of similarity and analogy will model the experimental results. The research will also begin the study of the role of structural similarity in learning and concept formation.

The goal of this research is to develop and test theories of human similarity processing. Similarity provides the basis for much of concept learning, transfer of knowledge from one situation to another, and generalization. Conversely, similarity can also be a source of human error, for instance when confusions between similar-looking materials or devices occur. This research will develop a new approach to encompass a range of human similarity processes, from causal and functional similarities to perceptual similarities. This approach departs from prior approaches in emphasizing the psychological processes that underlie similarity. The first set of studies will investigate interconnectivity, i.e., whether interactions between features can make some aspects of objects more salient in similarity judgments or even change the psychological interpretation of their properties. The second issue is the effect of alignment in judgments of the similarity of two figures. These studies will investigate how the seeming similarity between two figures changes as a function of how easy it is to place components of object representation in correspondence. A third set of studies will investigate the effect of similarity in constructive induction. These studies are based on the hypothesis that when people judge similarity, they sometimes align non- identical features in a way which captures only their abstract commonalities. For example, two engines, one missing a fan belt and the other with a broken generator, may be judged similar because they have dysfunctional parts. These research results will inform us about the range and flexibility of similarity processing. The next set of studies will address the dynamic process of similarity computation. In these studies, people will be asked to perform rapid similarity matches or will view similarity pairs for varying periods; this will allow comparison of fast and slow similarity judgments. If fast judgments have dynamics different from slow, interfaces and devices to which humans must respond quickly will have to be designed accordingly. The final set of studies will investigate the role of similarity in other cognitive processes. Some experiments will examine the role of similarity in category-based induction. For example, when people are told a new fact about some entity, (e.g., a tiger) they typically assume that the fact applies to other similar entities, (e.g., a lion). This research will investigate whether the similarity is symmetrical or directional and what aspects of the similarity affect further inferences that can be made. Other experiments will investigate whether the process of retrieval from memory can change the subsequent representation of the items. If items stored in memory are changed by the process that retrieves them, it suggests how gradual learning can occur and how certain similarity-based distortions can come about.

This project is being funded through the Learning and Intelligent Systems (LIS) Initiative. A group of nine senior investigators will study spatial competence, and its emergence over time, at the cognitive, computational, and neural levels. Topics to be studied include how people form spatial representations; how people communicate about spatial information using external symbol systems such as maps, diagrams, graphs, and linguistic descriptions; the role of the educational input received in American schools in supporting spatial learning; the optimal computational model of spatial learning; and, evidence of neural plasticity for spatial learning, based on both neuroanatomical study and neuropsychological evaluation. The common purpose of this group of related research endeavors is to examine the nature of environmentally-sensitive growth in spatial competence and how spatial learning can be maximized in the American population. Innovations for educational practice and educational software resulting from our research will be evaluated with the help of collaborating teachers.

The focus of the program is on the acquisition of spatial representation and reasoning. The researchers believe that in order to achieve their long-term goal of devising optimal education programs in spatial competence, they must first come to understand the nature of spatial learning and determine the types of experiences that lead to higher achievement in the domain, about which relatively little is known. Their previous research indicates that the development of spatial cognition is highly sensitive to input and thus is malleable. The current project brings together researchers from across cognitive science. They will explore three, presumably interrelated, aspects of spatial intelligence: First, one part of the team will investigate how spatial expertise can be enhanced directly by training the mental processes that operate on spatial representations. Second, another part of the team will investigate how spatial reasoning can be enhanced through maps and language. Third, the remaining part of the team will investigate how spatial reasoning can be enhanced by applying measurement and graphs to quantitative information.

Proposal 0628941
"SGER: Comparison and Explanation in Learning and Development"
PI: Kenneth D. Forbus
Northwestern University


ABSTRACT

This project will investigate the roles of comparison and explanation in conceptual change, in both children and adults. Comparison seems to play a major role in human learning: People compare multiple situations to help construct models, compare models with data to see how well they explain it, and compare models to each other to determine what data might distinguish between them. A key problem is deciding when to accept, revise, or abandon a model, based on the successes and failures of its explanations. This problem will be investigated by a combination of theoretical work, computer simulation, and psychological experimentation. The computer simulation will focus on how comparison processes are used in the construction, challenging, and testing of causal models, and on how learners should abandon, revise, or continue with its current models as a consequence. The phenomena modeled will be drawn both from the developmental literature (e.g., learning what kinds of things can float, and what quantities remain constant under physical transformations (i.e., conservation phenomena)) and from new studies of adults learning causal models. By adding to the science base needed to create software systems that can learn more like people do, this project will have Broader Impacts through improvement of intelligent tutoring systems, learning environments, and decision-support systems, and in general on the way intelligent systems are built.

The Spatial Intelligence and Learning Center (SILC) was established in the fall of 2006 as one of three second-cohort Science of Learning Centers. SILC's purpose is to develop the new science of spatial learning and to use this knowledge to transform STEM educational practice. Spatial learning is the acquisition of spatial knowledge and skills, and the use of spatial knowledge and skills to facilitate learning in both spatial and non-spatial domains. It provides the foundation for a wide range of reasoning skills in STEM-based activities, from solving mathematical problems to engineering new products to understanding graphical depictions of complex systems. Previous research shows that spatial skills are a strong predictor of entry into STEM disciplines in college and into STEM careers, that substantial improvement of spatial learning is possible, and that this improvement matters to STEM success. SILC has brought together researchers from multiple lines of work on spatial cognition and education and from a variety of traditional disciplines (e. g., cognitive science, psychology, artificial intelligence, linguistics, education, STEM disciplines), integrating them to achieve new insights. SILC researchers are developing a set of powerful tools for spatial learning, honing them into effective, deployable educational techniques and practices for STEM learning, including advanced technology (e.g., intelligent educational software), effective curriculum units (e.g., in elementary school mathematics), engaging activities (e.g., in children's museums), and spatial assessment instruments (e.g., testing children's spatial skills, testing adults' STEM-relevant spatial skills). Several of the insights, tools and products from SILC's initial funding period already hold transformational potential for spatial learning. Research and translational activities in the second and last funding period, from 2011-2016, will continue the investment in the science of spatial learning, in order to allow the fulfillment of this promise.

Earth Systems Science (40)

This project is creating new learning materials and strategies to foster the ability of students to reason by analogy about global climate change and is assessing student ability to use analogies to make and evaluate claims about global climate change. The PIs are testing the hypothesis that we can dramatically improve the climate change literacy of non-science majors through explicit instruction about analogical reasoning in a general education-science course that focuses on Global Change. The project is focusing on analogical reasoning because 1) analogical reasoning is central in higher-level learning, 2) scientists routinely use analogies to generate hypotheses and to solve research problems, and 3) scientists commonly use analogies to convey information about complex systems responsible for global climate change. The PIs are testing their hypothesis using matched experimental and control classes taught by the same instructor. They are assessing the ability of students to generate causal explanations of processes that drive global change through a mixed methods approach. The PIs are developing new instructional materials and assessment instruments based on the understanding of analogical reasoning by cognitive scientists. In particular, the PIs are helping students to understand analogies and are also improving their analogical reasoning by focusing on five dimensions of analogical reasoning suggested by decades of research: retrieval, mapping, evaluation, abstraction and re-representation. By analyzing the efficacy of their approach, the PIs are developing a new, detailed model of how best to teach analogical reasoning specifically within the context of improving climate change literacy.

Learning by analogy is a powerful and efficient method for acquiring new information. The goal of this research program is to trace the origins of analogical ability in human infants, using the simplest and most basic relation--that of sameness and difference between two things. The proposed research will identify the scope of infants' ability to detect same-different relations. Certainly infants do not have the same higher-order relational abilities of children and adults. This research will 1) identify the earliest evidence of relational learning processes in the first months of life; 2) trace its developmental trajectory over the first year; and 3) characterize the interplay between language and early relational ability.

The research addresses the key question of when children begin to think abstractly and how this ability can be fostered. Studies of early relational learning processes will provide insight into the conceptual capacities that are critical to higher-order learning in childhood and beyond. Further, the studies of interactions with language may lead to better understanding of the role of cultural and linguistic experience in conceptual development. The patterns revealed by these experiments will give parents and educators the requisite knowledge to support relational learning in STEM disciplines. They may also enable practitioners to identify patterns that deviate from typical development and suggest targeted interventions for children with cognitive or linguistic impairments.

Analogical ability is the ability to make relational comparisons between objects, events, or ideas, and to see common relational patterns across them. It is a cornerstone of higher reasoning ability, and is essential for learning mathematics and science. This project investigates the nature of this ability and how it develops in infants, tracing its development over the first year of life. Delineating the conditions that promote relational learning in young infants, will lead to insights into how best to promote relational learning in children and in adults who show lags in abstract learning. One result of the proposed studies will be a set of methods and tools that can be used by teachers and caregivers to support relational learning. For example, this research can serve as a springboard for developing targeted interventions for young children diagnosed with language delay, as well as those diagnosed with autistic spectrum disorders. Another result will be a better understanding of how to build artificial intelligence systems that learn more like people, with far less data than today's systems require.

The starting point for this proposal is a recent demonstration that 7- and 9-month-old infants can form abstract same and different relations, and apply them to new objects. The preliminary studies suggest that even 3-month-old infants are capable of relational learning; however, they are highly vulnerable to distraction by the interestingness of the objects in the pairs. The new research will use four series of experiments to trace infants' ability to learn abstract relations. The first will examine the processes that promote relational learning in 3-month-olds. The second will investigate the conditions that support spontaneous comparison and learning in 7- and 9-month-olds. The third series will test how language influences relational learning- specifically, whether naming relations can improve learning, and whether naming objects can impede relational learning. The fourth series of experiments will investigate the generalizability of these effects by testing a variety of abstract relations. Computational modeling of the learning patterns found in our studies will provide complementary insights on these processes. Taken together, these studies will reveal information critical to understanding analogical processes and the origin and evolution of higher-order cognition.

The goal of this project is to develop and test methods for training powerful thinking skills. Specifically, the research will explore techniques for improving students' analogical reasoning ability. The history of science has shown analogy to be a powerful tool in scientific discovery. And there is abundant evidence from laboratory and classroom studies that analogical comparison can improve students' ability to learn, especially in mathematics and science. For example, reading that sound waves are analogous to water waves can give students insight into how sound behaves. Yet not all students benefit equally from analogical teaching. The first goal of the research is to discover how best to equip students with methods for engaging productively with analogical comparisons of all sorts—in short, to become better thinkers and learners. The second goal is to test whether acquiring these skills will improve students' ability to learn in a new scientific arena: evolution. Overall, this project will provide new insights into analogical reasoning and learning as a fundamental ability and into how it can be trained.<br/><br/>The project will develop an analogical training program for undergraduate students targeting three intersecting facets of analogical thinking: (1) strategies for processing analogical mappings, including aligning two examples, drawing inferences from one example to another, and noting key differences; (2) metacognitive insight into the nature and uses of analogical thinking; and (3) ability to recognize opportunities for comparison. In Phase 1, iterative design and experimental methods will be combined to develop and test training components. In Phase 2, a complete training program for improving analogical reasoning and supporting science content learning (specifically, about evolution) will be tested. A group that has received analogical training will be compared with a control group on their ability to learn from analogical comparisons in evolution and in other topics, using a pre-post design. They will be assessed in both short-term (immediately after instruction) and long-term (6-7 weeks later) learning. A pilot study will explore adapting the analogical training program for eighth graders in a classroom setting. <br/><br/>This project is funded by the EHR Core Research (ECR) program, which supports fundamental research on STEM learning and learning environments, broadening participation in STEM, and STEM workforce development.<br/><br/>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.