A graduate of Indiana University (Ph.D., 2007) and Arizona State University (B.S., 1999), Roger Stanton is a published author and professor. Dr. Stanton has performed a number of in-depth studies on category learning. In the opinion of Roger Stanton, category learning is a fascinating area of study that requires a deliberate commitment to further research.
Categorization involves the process of placing items into groups known as “categories,” explains Roger Stanton. However, this definition merely scratches the surface of what categorization means with regard to the field of cognitive psychology. Roger Stanton says that categorization refers to how people assign events or objects to categories. Further, Roger Stanton describes categorization as a way to structure knowledge and organize it into select categories. Categorization also relates to the various inferences that people draw when determining to what category an event or object belongs, adds Roger Stanton.
Q: What is cognitive science?
Roger Stanton: Cognitive science is an interdisciplinary scientific study of the brain. Its origins can be traced back to the mid-1950s, a time when researchers in a number of fields began to formulate theories based on computational procedures and complex representations. In the 1970s, the Cognitive Science Society was developed and the industry journal Cognitive Science was first published. Since then, more than 70 colleges and universities in Europe, North America, Australia and Asia have established courses in the field of cognitive science.
Q: Is cognitive science regularly studied in the academic community?
Roger Stanton: Even though many institutions of higher learning present courses related to cognitive science, it’s really not a well-known research area. As a general area of study with various research aspects, cognitive science was a perfect subject for my own research projects. Overall, the mind can typically be characterized as a distinct set of complex associations, which are then represented as a complex and layered network. It’s truly fascinating to study.
Q: What are neural network models?
Roger Stanton: Neural network models, by their standard definition, are simplified models of our biological neurons. Neural network models allow cognitive scientists to simulate intelligent behavior.
Q: How were neural network models first developed by scientists?
Roger Stanton: In computer science, as well as a host of related fields, these models were inspired by the central nervous systems of animals, which are capable of pattern recognition and machine learning. They have been generally presented as systems of interconnected “neurons,” known for computing values of the entity by feeding the given information into the network.
While at St. Mary’s College of Maryland, Roger Stanton worked with St. Mary’s College students across a wide array of activities outside of the traditional classroom setting. He served as the statistics consultant for St. Mary’s Projects—which are senior honors thesis projects. Roger Stanton was the primary investigator in the Category and Concepts Lab at St. Mary’s College of Maryland, and he advised St. Mary’s College students as research assistants in the laboratory. As part of this research, St. Mary’s College research assistants co-authored journal articles and presented at professional conferences. Additionally, he served as the academic advisor for several clubs and honors societies at St. Mary’s College, including Psi Chi, the cycling club, and the jiu jitsu club.
This was Roger Stanton’s first publication. This publication follows up a previous article that presented evidence that a prototype model provided a better account of category learning data than did an exemplar model. In this particular study, a linear separable category structure is tested, and the prototype and exemplar models were compared on their ability to account for the data. The exemplar model provided a better account of the individual participants’ data. Importantly, a linearly separable category structure has several features that should be conducive to a prototype-based strategy, and thus this result is strong evidence that an exemplar model provides a superior account of human category learning performance.
This article also compares prototype and exemplar models. However, in this case a previous article showed that a prototype model actually provided a better account of learning performance in the well-known Medin & Schaefer (1978) 5/4 paradigm. The article shows that when the models are equated in terms of flexibility, the exemplar model does provide a better account of the learning data.
Exemplar models and decision bound models make remarkably similar predictions across a wide array of experimental settings. Although the models have been compared in terms of their overall quantitative fit to data, there is little evidence for a qualitative comparison. This article demonstrates that when probabilistic feedback is given to participants, the decision bound model predicts that this probabilistic feedback will not effects the participants response times. However, a random-walk exemplar model predicts that the probabilistic feedback will produce slowed response times. The results of the study confirmed the predictions of the random-walk exemplar model.