Robert Lee Goldstone - US grants

The proposed experiments and model explore how people learn new concepts. The central thesis is that concept learning often changes how objects are perceived. That is, concepts and categories may be constructed on the basis of the perceptual features of objects, but these concepts may in turn influence what we see as the perceptual features. The aim of the proposed work is to specify and provide a formal account of such interactions between concept learning and perception. The influences of concept learning on subsequent perceptual judgments are investigated in a first series of experiments. The possibility that concept learning causes new object descriptions (based on perceptual features) to be created is considered in the second series of experiments. A neutral network computer model of concept learning is proposed. In this model, the concepts to be acquired alter the perceptual features used for categorization. Rather than assuming that fixed perceptual features are combined to determine categorization rules, this model allows for a mutual and simultaneous influence between concepts and perceptions. The research will have two tangible benefits. The first benefit will be to further our understanding of how people naturally learn. Unlike most laboratory studies, concept learning in the real world does typically involve alterations in what is perceived. For example, specialized experts learn to improve their perception and come to be able to classify better, for example, to classify some patients as suffering from cancer, some tumors as malignant, or some oil fields as promising. If we can understand how experts learn to improve their perception with experience, we will better be able to effectively train new personnel. The second tangible benefit will be to aid the construction of automatic devices for human-like categorization. By understanding the mechanisms that people use to create new categories, we will be in a better position to develop procedures that also can categorize objec ts in the world in an intelligent manner.

The Cognitive and Information Sciences (CIS) program offers a model undergraduate curriculum designed to expose students to the study of intelligent systems. It is intended as a model for the development of similar programs at other institutions as they begin to move into this emerging field. The core of the newly proposed major concerns formal theories of mind and information that integrate computer science, psychology, philosophy, neuroscience, and mathematics. The curriculum gives students training in the basic skills of this new field (writing, mathematics, experimentation, and computer programming), but also fosters the development of expertise in particular sub-fields such as human cognition, learning and instruction, logic, and computation. Newly designed courses will expose students to both basic scientific theories and to applications of these theories to education, system design, and database retrieval. The program will be assessed internally by pre/post exams within courses and exit interviews upon graduation, and externally by quarterly site visits by expert consultants and a comprehensive third-year review by consultants from Indiana University's School of Education. The tangible products of this proposal will include general information on the CIS degree and its requirements and materials for specific courses in the form of syllabi, lecture notes, slides, bibliographies, assignments, and exams. Most important will be the course software and laboratory software developed for all four of the core CIS courses, and several others; this courseware will emphasize interactive, hands on learning through doing, and will be in a form easily usable by other programs that are beginning or being planned in this area. These products will be disseminated through our existing technical report series and newsletter, through electronic media, web sites, and contacts at other universities, and will be provided upon request of other institutions starting such programs.

People show a remarkable ability to learn new concepts. Children learn to classify some animals as dogs, some foods as candy, and some people as relatives, Specialized experts learn to classify some mushrooms as poisonous, some tumors as malignant, and some wines as Bordeauxs. The proposed experiments and model explore how people learn new concepts. The central thesis is that concept learning often changes how objects are organized into features. We may build our concepts from the perceptual features of objects, but the concepts that we build in turn influence what we see as the features. The aim of the proposal is to provide a formal account of the interactions between conceptual and perceptual learning. The first series of experiments explores the mechanisms by which concept learning alters descriptions of the objects to be categorized. Particular emphasis is given to selective attention, unitization (integrating originally separate sources of information), and dimensionalization (isolating originally fused sources of information). The second series of experiments uses established and new operational definitions of features to quantify the influence of concept learning on object organization. A neural network model of concept learning is proposed. In this model, the concepts to be acquired alter the perceptual features used for categorization. Rather than assuming that fixed perceptual features are combined to determine categorization rules, this model allows for a mutual and simultaneous influence between concepts and perception. Medical professionals are often required to learn new concepts (e.g., malignant tumor, eczema, and Parkinson's disease). Many of these concepts have a strongly perceptual basis. An understanding of how perceptual concepts are learned, and how perceptual adaptation supports concept learning, could help to more effectively train medical professionals. More generally, the experimental results provide a framework for understanding expert/novice differences, for applying results on neural plasticity to behavior, for establishing training regimes for improving perceptual abilities, and for refining educational procedures that involve teaching concepts with a strong perceptual component.

Interactive computer simulations are increasingly being used to teach students
scientific principles, but relatively little systematic research has done on
the factors that make a simulation pedagogically effective. This research
explores how to design a computer simulation so that students learn the
scientific principle underlying it, and transfer this principle to new
domains. Principles from the interdisciplinary science of complex adaptive
systems are chosen because these mathematical formalisms are applicable across
a wide range of domains. Laboratory experiments with students will explore how
active exploration of one simulation benefits or hinders understanding of a
subsequently presented simulation based on the same principle. Experiments
will explore the roles of concreteness and idealization, simulation similarity,
and individual differences in abstract transfer. The scientific goal of the
inquiry is to gain an understanding of how perceptual experience can lead to
abstract conceptual understanding, and how conceptual understanding can change
perceptual experience. The practical goal is to translate this understanding
into general educational principles for integrating computer simulations into
classroom activities.

When people draw on information from their individual experiences and local surroundings and use it to make a decision or judgment on a particular issue, some individuals may obtain information permitting them to make a correct decision. In contrast, others may obtain biased or inadequate samples of information. In situations like this, pooling informational resources among a number of people may help the group as a whole arrive at a better decision. This proposal outlines two projects examining how individuals can draw upon information that flows through social networks of friends and acquaintances, and use it in making their individual decisions. The process is dynamic, for as each individual draws on others' input to make decisions, his or her decisions in turn provide additional information to others in the network. Each project considers the relationships of variables at several levels, from individual agents' decision strategies (e.g., how much weight to give socially provided versus individually obtained information), to inter-agent interactions in which agents exchange information, to the overall structures of social networks. The first project considers multiple agents who are sharing information and influencing each others' choices on a particular decision (such as the choice of consumer products, or the preference for a particular social policy). The second project considers multiple agents who are assessing each other as potential partners or mates, and examines how these decisions are influenced by the spread of information about the agents themselves through the social network. Both projects will use multi-agent simulation techniques and empirical studies with human participants. The projects will have several types of broader impacts. They will increase interdisciplinary interchange, by applying novel multi-agent simulation techniques in combination with social psychological theories and empirical findings, to deepen our understanding of the role of information flow through social networks in the decision-making process. The first project may help develop ways for people to more efficiently draw on the information and experiences of others as they make their own individual decisions (e.g., purchasing consumer products) and thereby improve the outcomes of those decisions. Finally, the second project may ultimately help enhance understanding of how people select others as partners or mates, and perhaps improve the quality of those decisions - with a potential impact on the high divorce rate in society and on the multiple social problems to which it contributes.

This Integrative Graduate Education and Research Traineeship (IGERT) award supports a training program on the dynamics of brain-body-environment interaction in behavior and cognition at Indiana University. The purpose of the training program is to create a new kind of scientist with expertise in both the experimental and theoretical tools necessary to analyze intelligence as an emergent property of a complex dynamic system. The training program includes new courses, a professional development seminar, a colloquium series that provides opportunities for extended interactions between students and top researchers, research internships, and opportunities for international collaboration. The program also includes a detailed assessment plan and a summer program for undergraduates from underrepresented groups run in partnership with Indiana University Northwest, approximately 80% of whose students come from minority, first-generation college, female or low socioeconomic categories. Broader impacts of this program include recruiting new students from underrepresented groups into cognitive science and providing graduate students with the opportunity to participate in cutting-edge multidisciplinary research. All materials produced by this program will be made freely available on the web. More generally, this program will foster new kinds of discourse between the various disciplines that make up cognitive science. Finally, by placing cognition within its proper embodied and situated context, the proposed training program may impact how society fosters and measures cognitive ability. IGERT is an NSF-wide program intended to meet the challenges of educating U.S. Ph.D. scientists and engineers with the interdisciplinary background, deep knowledge in a chosen discipline, and the technical, professional, and personal skills needed for the career demands of the future. The program is intended to catalyze a cultural change in graduate education by establishing innovative new models for graduate education and training in a fertile environment for collaborative research that transcends traditional disciplinary boundaries.

This project concerns the role of perceptually concrete materials in teaching generalizable scientific concepts, with a particular emphasis on computer simulations. While concrete content appears to facilitate the learning of specific pieces of information, previous research has shown that it may impair students' ability to apply that information to new situations. The proposed studies explore the relation between the superficial, concrete details through which a phenomenon is presented, and the abstraction of deeper scientific principles underlying the phenomenon. The specific content to be learned concerns the principles of complex systems in a range of STEM domains such as civil engineering, biology, economics, and mechanics.

The research methods involve classroom-based and laboratory experiments incorporating computer simulations of scientific principles. By observing how interaction with one simulation affects students' understanding of subsequently presented information, the investigators can assess the degree to which the underlying scientific principle has been successfully abstracted. The proposed research is divided into two broad lines. First, the investigators explore the role of comparison of multiple cases in students' ability to acquire generalizable principles from concrete examples. Second, they examine the degree to which physically and spatially plausible mechanical systems may serve as models for understanding complex or unintuitive causal systems. In both lines, the investigators explore the ways in which learning is affected by factors such as the degree and type of concrete detail depicted, the relative overlap of concrete and structural information between cases, and the concrete and structural variability with the test cases. Studies will include students in 8th grade science classes.

Most impactful innovations involve people interacting with each other and with the tools that they build explicitly to help them think and/or perform better. The investigators will organize two workshops to explore the ways in which an individual’s thinking is augmented and shaped by other people as well as by tools. Coordination among people, artificial intelligence systems, decision algorithms, and learning technologies can give rise to adaptive problem solving far beyond the capabilities of single individuals. Workshop topics will include how to: facilitate the spread of useful information while reducing misinformation, combine judgments from many individuals (humans and other animals) for improved decision making, create policies and infrastructures that allow groups to effectively solve their shared problems, best combine the strengths of human and computer problem solving to create systems more advanced than either component on their own, and use computers to make human learning more efficient and flexible. The workshops will feature panelist presentations organized into themes and monthly group discussions. A broader impact of the workshop is to provide a venue for students anywhere in the world to learn about exciting developments in augmented intelligence. For students coming from smaller universities with few resources for research, the workshop would provide exposure to state-of-the-art research and opportunities to interact with leading researchers in this emerging field. The topic of augmented intelligence represents an exciting area of research that can inform many applications including: the optimization of team coordination, the design of intelligent tutoring systems, computer support for human decision making, and the organization of social media platforms.

This two-part, temporally and geographically distributed workshop will integrate perspectives from cognitive scientists, computer scientists, education researchers, neuroscientists, and biologists to lay the foundations of a shared conceptual framework for a new science of augmented intelligence, premised on the observation that people very rarely solve problems or develop understandings on their own. Instead, people recruit other people and technologies to help them, and to such a large extent that the apt unit for understanding cognition is typically not an individual person, but rather a larger system that incorporates multiple people and the tools that they created to enhance their capabilities. The workshop participants will grapple with theoretical and applied questions involving the nature of minds, how adaptive systems come into existence that can robustly solve a wide range of problems, and how humans and machines can complement each other’s strengths during learning and knowledge creation. The workshop seeks to stimulate new research directions at the confluence of theoretical, empirical, and technological advances that hold promise of developing an integrative science of augmented intelligence, drawling from multiple disciplinary perspectives. Deliverable products from the workshop will include online public videos of participants’ lectures, archives of conversations, and special issues of journals devoted to augmented intelligence.

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.