Philip J. Kellman - US grants

Most objects perceived visually are partly occluded by other objects; yet human observers accurately detect the unity and boundaries of objects under most circumstances. There are other phenomena in which boundaries are perceived in the the absence of local visual information, such as illusory contours and some perceived transparency phenomena. In all of these cases, perceived object boundaries are interpolated between physically- specified edges. This project will address the various cases of boundary interpolation, testing and refining a recently proposed unifying theory, and more generally attempting to specify the conditions under which interpolation occurs. Specific aims include tests of edge interpolation in both two and three- dimensional displays, development of methods for generation of random displays, and development of a variety of measures for assessing boundary perception, including both perceptual report and non-verbal measures. This work will have implications both for understanding human perception of objects and for object and boundary detection in computer vision systems.

People are able to perceive objects and boundaries successfully, despite the fact that the visual input on which that perception is based is fragmentary across both space and time. For instance, despite the fact that objects are occluded a great deal by other objects and that patterns projected on the retina change continually whenever objects or observers move, people can ordinarily detect complete boundaries and shapes. There are other cases in which people perceive boundaries in the absence of local visual information, known as illusory contours and transparency phenomena. This project will address the various cases of boundary interpolation; it will test and refine a proposed unifying theory and extend that theory from static, two-dimensional displays to the three-dimensional case and to cases in which boundaries are obtained from information conveyed by motion. The project will also investigate convergent methods to assess boundary perception, including both perceptual report and objective performance measures. This research, in addition to its implications for understanding human perception of objects, will aid in the construction of object and boundary detection methods in computer vision systems.

This project is being funded through the Learning and Intelligent Systems (LIS) Initiative. The focus is on learning as it occurs in complex environments, where the data have rich and potentially confusing structures. Nine investigators in five different disciplines - biology, chemistry, linguistics, psychology, and high school teaching of mathematics and science - will mount a collaborative, multi-level experimental and theoretical analysis of the mind's learning structures. The work integrates research on formal analyses of learnability, the evolution of complex natural and artificial adapative systems, the genetics of memory, the mind's ability to keep track of language learning data, perceptual learning of complex displays like equations and molecular models, and the creation of integrative math and science modules for use with interactive learning technologies. The unifying theme running through all of the projects, and across every level of analysis, is the interaction between the structure of the brain's learning mechanisms, and the structure of the data that support learning. Two related leitmotives cut across the planned work. First, the project itself is conceived of as a complex, interdisciplinary learning environment for people ranging from high school students and science teachers in the Los Angeles community, to senior faculty at UCLA. Second, the research efforts interact with and inform advancements in the rapidly evolving technologies for learning, instruction, genetic screening, and the development of artificial systems.

DESCRIPTION: This project comprises a broad-based investigation into visual object and surface perception. Much of human thought and behavior is organized around the perception and representation of objects. Among the fundamental, unsolved problems of object perception is unit formation: How do we obtain descriptions of connected objects in the three-dimensional world despite interruptions - across space and time-of their contours and surfaces in their projections to the eyes? Answering these questions of unit formation and related aspects of object perception is the goal of the research. Psychophysical experiments using objective performance tasks with normal, human adult observers are used to probe the information (stimulus relationships) involved, the representations formed and the processing characteristics of human object perception. The data are used to construct geometric and information-processing models of the contour and surface relationships that lead to perception of connected objects, as well as to inform models of the neural activity underlying perception. Building on prior work, the current project extends the empirical and theoretical efforts into the less well-studied domains of three-dimensional relationships in object formation and dynamic (motion-carried) information. Particularly exciting is the suggestion in recent work that common principles may describe unit formation in two and three dimensions, and dynamic vision. These efforts will lead toward a better understanding of fundamental cognitive and behavioral processes - those that achieve representations of objects, surfaces and scenes. The results will have manifold implications for understanding normal and impaired human function, for constructing artificial (including robotic) vision systems for probing the neural mechanisms of perception.

Perceptual learning is defined as experience-induced changes in the way information is extracted by these researchers who intend to investigate how perceptual learning occurs and whether it has implications for mathematics and science learning. It is the process by which learners differentiate relevant structure from irrelevant variation. The methods of research involve experimental investigations of conditions affecting perceptual learning, effects of structure discovery, and structure mapping variants of learning procedures on transfer. Experiments will employ objective measures of learning such as speed, accuracy, and transfer to novel problems. They will test and apply new learning technology such as automated sequencing algorithms that use a learner's speed and accuracy to assess learning and to sequence events for optimal efficiency.

The panel reviewers noted that an investigation that might clarify perceptual learning might help supplement or even challenge prevailing constructivist theories of learning. This way of thinking about learning could be an important discovery for mathematics and science learning.

DESCRIPTION (provided by applicant): Using Perceptual and Adaptive Learning to Advance Chemistry Education This proposal addresses broad Challenge Area 12) Science, Technology, Engineering and Mathematics Education and specific Challenge Topic 12-OD-101: Efficacy of educational approaches toward promoting STEM competencies. The work will refine and test the efficacy of Chemistry learning modules that incorporate advanced perceptual learning and adaptive learning technologies, both of which have great, but largely untapped, potential to improve STEM learning. The target learning group will be community college students taking basic chemistry courses. This student population is a large and important group, representing a considerable pool of talent for a broad spectrum of health and science-related occupations. The project seeks to address two major problems that limit student learning. One involves students'abilities to grasp and process fluently crucial structures and relations. Recent work suggests that this problem can be addressed with dramatic success by perceptual learning principles embedded in learning technology. A second problem is that learning does not adapt to the needs of the individual learner. These limitations can be overcome by the use of adaptive learning technology in which learning events are sequenced based on learner accuracy and speed. The learning interventions to be tested take the form of Perceptual Learning Modules (PLMs). PLMs accelerate, through short, interactive trials, learners'abilities to quickly and accurately process key structures and relationships in complex domains. PLMs are web-based;they deliver learning and assessment via standard browsers over the Internet;and they can be easily scaled-up for broader use. The technology will incorporate recently patented adaptive learning algorithms that use a continuous stream of performance data to optimize principles of learning and memory, producing improved efficiency of learning and assessment. The project team, consisting of experts in cognitive science, learning technology, curriculum development and evaluation, chemistry and chemistry education, will refine two previously developed Chemistry PLMs and develop one new one using existing adaptive learning software, allowing a broad test of these technologies. Both laboratory pilot studies and full-scale efficacy studies in community college Chemistry courses will be carried out. Efficacy studies will include randomization by participant;treatment groups will be compared to control groups who spend comparable amounts of time using conventional materials;and conditions that test efficacy of specific components of perceptual/adaptive learning will be included. The proposed efficacy studies are expected to yield robust learning gains, indicating the potential of perceptual learning to produce pattern recognition and fluency, and the potential of adaptive learning technologies to fit individual learning needs and produce mastery. Results will have implications for Chemistry education and learning in other STEM fields, particularly for students who struggle with traditional instructional methods. Public Health Relevance: The proposed project will test the efficacy of web-delivered perceptual and adaptive learning technologies in Chemistry learning by community college students, a large and important group that includes a majority of all students enrolled in colleges, many of whom belong to groups traditionally under-represented in STEM fields. Results obtained in this project will inform educational practice;successful outcomes could lead to widespread effective use of perceptual and adaptive learning technologies to remove barriers and improve learning efficiency for large numbers of students in STEM curricula leading to a full-spectrum of health-related occupations.

The purpose of this project is to increase our understanding of how to optimize the features of learning technology systems, which have the potential to be applied widely across domains, settings, and age groups. The team of researchers from the University of California, Los Angeles, and the University of Pennsylvania focuses on learning technology that integrates (1) principles of perceptual learning that accelerate learners' abilities to recognize key structures and relations in science and math domains, and (2) adaptive learning algorithms that use a constant stream of performance data to adapt the learning process to each individual. The collaborating partners include two K-12 schools serving diverse populations and two community colleges.

The researchers will investigate the role of response time data as a novel input into both spacing and the setting of learning criteria in adaptive and perceptual learning systems. Specific hypotheses will be tested in a series of randomized controlled experiments. Students will complete pretests, posttests, and delayed posttests to evaluate gains and long-term durability of learning. The researchers aim (1) to test adaptive sequencing that utilizes learner response times (along with accuracy) to guide spacing in learning; (2) to improve learning systems based on understanding the role of response time in setting learning criteria; (3) to develop integrative adaptive and perceptual learning systems that incorporate best practices for the use of combined speed and accuracy data; and (4) to demonstrate the feasibility and effectiveness of such systems across STEM learning domains and age groups by testing learning modules for elementary mathematics and high school and college chemistry.

This project is a major step in the merging of principles of cognitive science with learning technology in service of STEM learning.

Developmental mathematics remains a critical obstacle to college readiness, involving large majorities of community college students and disproportionately impacting groups of students who are underrepresented in STEM - minority, low-income, and first generation college students. This project aims to improve our understanding of important instructional principles emanating from cognitive science that are related to success in community college mathematics. The studies in this project aim both to advance our understanding of learning and to apply learning innovations to known challenges in authentic learning settings, thus testing and extending the generalizability of laboratory findings to consequential mathematics learning with diverse learners. The team will focus on two areas of research in the cognitive science of learning that have that have important implications for learning complex domains like mathematics and science. The first is perceptual learning, which accelerates learners' abilities to quickly and accurately recognize key structures, patterns, and relationships. The second area is the development of adaptive learning algorithms that utilize real-time performance data in conjunction with principles of learning to improve the effectiveness and efficiency of learning by tailoring the learning process to each individual student. This project is designed both to advance the scientific understanding of perceptual and adaptive learning and to assess their potential to improve learning in community college mathematics.

Several sets of experiments will investigate basic scientific questions regarding 1) whether and how there may be a unified account of the benefits of spacing of learning events that applies across different forms of learning; 2) how intermixing passive and interactive events may improve adaptive learning; and 3) how adaptive methods may be used to enhance the role and benefits of comparisons in perceptual learning. Basic principles will be established through research in controlled laboratory studies. The project will then extend these findings by conducting applied tests of these principles to optimize the effectiveness, efficiency, and durability of learning, and to maintain motivation, with community college students enrolled in remedial mathematics courses.

This project is supported by NSF's EHR Core Research (ECR) program. The ECR program emphasizes fundamental STEM education research that generates foundational knowledge in the field. Investments are made in critical areas that are essential, broad and enduring: STEM learning and STEM learning environments, broadening participation in STEM, and STEM workforce development. The program supports the accumulation of robust evidence to inform efforts to understand, build theory to explain, and suggest intervention and innovations to address persistent challenges in STEM interest, education, learning and participation.

Project Summary/Abstract Cancer screening from visual displays, as in dermatology and radiology, depends crucially on the expertise of medical practitioners, but current data indicate that even among experienced professionals there are significant and persistent error rates. While there have been impressive advances in the technologies of medical imaging, considerably less attention has been paid to the learning processes involved in the training of medical image interpretation. Research in perception and cognition indicates that the central process by which people become able to detect and classify complex and subtle patterns and structures in visual images is a process known as perceptual learning. Through perceptual learning mechanisms, with appropriate practice in a given domain, the brain progressively improves information extraction to optimize task performance. These mechanisms are largely unaffected by the traditional didactic instruction common in medical education; instead, they depend on interaction with large numbers of examples with task-relevant feedback. Recent work has shown that application of principles of perceptual learning can dramatically accelerate accuracy and fluency in medical learning domains. Evidence suggests that these training methods can be markedly enhanced, and customized for individual learners, by incorporating novel adaptive learning algorithms based on principles of learning and memory. The primary aim of this project is to investigate principles and mechanisms of perceptual and adaptive learning in the learning of multiple diagnostic categories in dermatologic screening and mammography, with the ultimate aim of improving training and proficiency in cancer image interpretation. Studies with novices in lab settings will establish basic principles and hypotheses, and selective studies with nurse melanographers, residents, and physicians will test validation with actual practitioners. Culminating studies of melanographers in actual dermatologic screening settings will compare practitioners who train with best-practices perceptual- adaptive learning modules (PALMs) to control participants. Specific studies will investigate the incorporation of signal detection concepts into adaptive perceptual learning systems; the role of comparisons in defining and differentiating perceptual categories; the relative benefits of passive and active learning episodes across learning phases; and the relationship between the stringency of mastery criteria and the degree to which resulting performance is accurate, fluent, generalizable, and long-lasting.

This project aims to serve the national interest by developing and testing innovative instructional technologies based on well-established principles of perceptual and adaptive learning to help undergraduates master neuroanatomy. During a time when there has been a sustained explosion in both neuroscience research and in undergraduate enrollment in neuroscience courses and majors, effective neuroanatomy education is critically important in providing a common foundation by which the brain and its functions can be understood. However, mastering neuroanatomy involves several different forms of learning and poses significant instructional challenges. Perceptual learning refers to improvements in how people pick up information as they gain experience and practice in a given domain. Prior research shows that interactive computer-based perceptual learning interventions reliably develop students’ ability to quickly and accurately recognize and distinguish complex structures, patterns, and relationships, such as those that are encountered in neuroanatomy courses. However, perceptual learning has largely been ignored in formal education because the learning conditions that promote it most effectively are rarely available in standard course materials and instructional formats. This project plans to combine perceptual and adaptive learning technologies in a novel format known as Perceptual Adaptive Courseware (PAC), in which short cycles of expository instruction are interwoven with web-based adaptive interactive learning, enabling students to advance and consolidate their learning. The approach may improve learning in neuroanatomy and also provide valuable information regarding the applications of these learning innovations to other STEM domains.

The project’s goal is to generate new knowledge by investigating whether enhancements based on this combination of perceptual and adaptive learning technologies produce improved instructional outcomes for undergraduates learning neuroanatomy. The PAC intervention will be iteratively developed, piloted, and tested across two cohorts of undergraduates enrolled in multiple sections of an upper-level course in behavioral neuroscience at a public university with a diverse student body. The project will implement a controlled experimental design comparing groups that experience the PAC intervention for a defined segment of the neuroanatomy curriculum to groups that experience instruction in normal lecture and lab formats for the same portion of the curriculum. Pre/post assessments will yield quantitative data on students’ learning of the targeted content. Students, teaching assistants, and instructors will also provide qualitative data to help design and optimize the user experience with respect to features such as navigation, pacing, graphics, and interactive interfaces. Perceptual Adaptive Courseware, which runs on laptops, desktop computers, and tablets using standard browsers, represents a model approach that can be readily extended to other STEM domains. Findings from the project will be shared with faculty who teach neuroanatomy or related disciplines through meetings and publications of the Society for Neuroscience and the Faculty for Undergraduate Neuroscience. Results will also be disseminated to audiences interested in applications of cognitive research to instruction and learning technology. The NSF IUSE: EHR Program supports research and development projects to improve the effectiveness of STEM education for all students. Through the Engaged Student Learning track, the program supports the creation, exploration, and implementation of promising practices and tools.

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.