Education and Training
1 OVERVIEW
To carry out interdisciplinary research in Cognitive Science, IGERT students will need a diverse set of tools. Specifically, they will need to be familiar with mathematical and computational methods for modeling behavior, while also having a working understanding of experimental design and data collection in the study of cognitive phenomena in humans and animals, and an appreciation of what biological systems can teach us about the engineering of intelligent behavior in real-world environments.To provide IGERT students with these tools, The IGERT program at MSU balances training within each student's home department with interdisciplinary coursework and research beyond the home department. Each trainee takes two years of intensive coursework (some of it developed specifically for this IGERT program) and mentored research training, followed by 2-3 years of independent research mentored by an interdisciplinary guidance committee. Each student's course curriculum consists of a mix of department-based courses that he or she will need for professional training within their respective disciplines, integrative "core courses" in a sequential decision-making approach to cognitive science, and cross-disciplinary courses, selected as needed to fill in gaps in the student's knowledge of other fields.
2 INTERDISCIPLINARY TRAINING GOALS
Our program builds upon our experience training students in Zoology, Psychology, and Computer Science to do collaborative research. In that project, we were struck by the major differences in outlook separating students in different fields, even when they were interested in similar problems. One fundamental difference arises from the fact that biologists and psychologists must approach behavior in a primarily top-down fashion, through incremental improvement of initially vague hypotheses about the underlying processes, whereas computer scientists must work out how to build complex behavioral abilities from the ground up. We find that computer science students are often dismayed by how imprecise hypotheses are in behavioral biology and psychology, and by how tolerant researchers in these fields are of unexplained variation in the phenomenon under study. Students in biology and psychology, on the other hand, sometimes question the plausibility of models that ignore the messy details of behavior in favor of computational elegance. Communication between disciplines is further hampered by differences in background knowledge and vocabulary. Finally, there are major differences in the methods used for testing hypotheses, and in the ethical issues that arise in different fields.Our program is designed to overcome such differences in outlook, and to encourage students to see connections and parallels among fields. We are guided by the following core objectives:
- Establish a shared conceptual framework and a common vocabulary for discourse in this framework. This shared framework recognizes that many cognitive problems can be cast as problems of sequential decision-making, and as such can be analyzed using the conceptual and computational tools that have been developed for such problems. A shared framework is essential for communication across disciplines. It is established through a series of integrative courses, and through weekly research-group meetings that allow frequent opportunities for interaction.
- Foster strong disciplinary training. A deep knowledge of at least one discipline provides a solid foundation to which one can refer in exploring commonalities and contrasts with other disciplines. Students complete the requirements within one department, interact extensively with students and faculty in their home department, and participate in lab meetings held in connection with the individual research programs of the IGERT investigators.
- Foster exploration of concepts, goals, and research methods in other fields. Complementing the disciplinary training is extensive exposure to research in other fields. This is achieved through the interdisciplinary IGERT Core courses described below, through courses offered in other departments and through the Cognitive Science Program, and through participation in the Cognitive Science Seminar course. This combination of depth within a discipline and breadth of exposure to other disciplines is organized by reference to the shared conceptual framework mentioned above. Our goal is to provide students an intellectual "map and compass" that they can use to guide their own interdisciplinary research.
- Develop teamwork skills in pursuing shared goals. This objective recognizes that interdisciplinary research is often carried out by groups of people with complementary expertise. The shared conceptual framework plays a role in fostering a shared set of goals, but we also build opportunities for teamwork into our courses and mentored research experiences.
- Strong technical training. Making bridges across disciplines is easier if one has facility with a wide spectrum of scientific techniques, including mathematical modeling, statistical analysis, and computer programming. Trainees will differ in the skills they bring with them, so gaps in students' backgrounds are remedied through specialized courses and summer workshops.
- Develop ability to formulate and carry out a coherent research plan. Meeting this objective is clearly critical to the success of our students individually, and to the long-term success of the IGERT program. Our experience is that this objective will be met as a natural consequence of meeting the other objectives.
Now we explicate the training program that we have designed to achieve these objectives.
3 DEPARTMENTAL REQUIREMENTS
We expect that students concentrating within their own field of study will enter the program with different backgrounds and different career goals. For students to compete for jobs in their respective fields, they must be adequately prepared. While our aim is to give students a rich interdisciplinary perspective on sequential decision-making, we do this in a way that respects the need for appropriate professional training within each discipline. For example, computer science students studying artificial intelligence need to be thoroughly trained in machine-learning methods that do not necessarily involve sequential decision-making, and hence require courses that do not necessarily incorporate the interdisciplinary perspective that we adopt for the program as a whole. Similar discipline-specific constraints arise in the training of students in other cognitive science disciplines.In general, departments permit some flexibility in the courses that students can take to fulfill degree requirements. However, we expect that some of our IGERT students may take a somewhat heavier course load than traditional students.
4 INTEGRATIVE CORE COURSES
Here we describe a set of three courses that all IGERT trainees will take during their first three years. The goal of these courses is to foster interdisciplinary perspectives on intelligent behavior, and to help students develop the intellectual and technical tools needed for integrative research in sequential decision-making. We have designed our core-courses in a way that would make them of value to any student interested in cognitive science. In this way, we expect that the training program will be valuable in nurturing the growth of graduate training in cognitive science at MSU.IGERT Core 1: Nature and practice of cognitive science (Year 1, second semester). This course, which would be required not only of IGERT students but also of all Cognitive Science Program students at MSU, would be designed to create the intellectual bridges among disciplines. The approach would be to consider a series of cognitive domains (e.g., navigation, language, vision, social organization) from multiple viewpoints, and thus to lay out a broad map illustrating the diversity of approaches to the study of intelligent behavior. One major goal would be to examine how scientists in different subdisciplines frame scientific problems and test hypotheses, highlighting the special challenges that arise in each field.
It is in this course that we examine the ethical issues that arise in different fields. For example, we consider the ethics of experimentation on human and animal subjects, the social impact of technology, and standards of scientific integrity that apply in different fields. Dealing with ethical issues in this first course accomplishes two goals. First, it provides an additional way of helping students appreciate the differences among fields that make up Cognitive Science. Second, discussing ethical issues in the first year of a student's program helps place these issues in the foundation of the intellectual framework that we hope to provide for students.
This course is coordinated by a single instructor, with guest presentations by other cognitive scientists, including faculty from outside the core IGERT group. The format combines lectures with student-led discussion of readings.
IGERT Core 2: Sequential decision-making as a unifying theme in cognitive science (Year 2 or 3, first semester). This course examines common issues that cut across the study of sequential decision-making in animals, humans and machines. Whereas the previous course stresses the differences in how particular cognitive phenomena are analyzed in different fields, here the emphasis is on showing students that a wide variety of cognitive problems can be cast as sequential decision-making problems. For example, the study of language processing, navigation, and selective attention require very different methods, and are explored by different subdisciplines in the behavioral sciences. Viewing these different cognitive domains as sequential decision-making problems recognizes that they have certain abstract features in common. This course provides a detailed introduction to sequential decision-making as reflected in these various cognitive domains, and to the computational models by which sequential decision-making problems can be analyzed.
The sequential decision-making framework is potentially very broad, and so we do not restrict the focus to the domains making up our major research efforts. Following are examples of domains in which sequential models already are being used, or, in our view, could be profitably applied: economic decisions in the context of feeding or reproduction; dominance relationships in social groups; human judgments and decision-making; dynamic scene perception over time.
The first half of this course consists of lectures and tutorials by faculty from different disciplines, explaining the sequential-decision-making approach as it applies to their respective study systems, and introducing students to empirical and computational methods needed for the study of these systems. The second half of the course consists of student-led presentations based on computational and empirical student projects on topics cutting across different fields. The projects may involve data analysis, computer simulations, or modeling, and could use the methods introduced in the first part of the course. Students are expected to work in teams in carrying out and presenting their projects.
This course is designed to produce a high degree of interaction among students both in and out of the class. Furthermore, the presentations are very effective in helping students and faculty alike appreciate the differences that separate different disciplines in the approach to scientific problems, and to explore ways of surmounting these differences. Past classes have been a highly rewarding experience for all of us, and several of the class projects have developed into full-fledged lines of research that have been presented at conferences and/or published as journal articles and book chapters.
Seminar in cognitive science (all years). This seminar course, currently offered as Psychology 863, is open to students in the Interdepartmental Graduate Specialization in Cognitive Science, and is integrated with our Distinguished Speakers in Cognitive Science lecture series. Students in this course meet with program faculty and students the week before each lecture to discuss readings, and then meet as a group with the speaker during his or her visit. Thus they learn about state-of-the-art integrative research from the scientists who do the work. To strengthen the training impact of this course for the IGERT program, we ask speakers, in their meetings with the student group, to discuss issues about professional development, so that students can learn about the strategies for success. We also strive for racial and ethnic diversity among the speakers, to provide the widest possible range of role models. The current requirement is for students to enroll for one full year of this seminar (1 credit hour each semester), but we expect that IGERT students will participate in the seminar program even during semesters when they are not enrolled, just as current Cognitive Science Program students and faculty regularly participate in the functions associated with the Lecture Series.
D.5 CROSS-DISCIPLINARY COURSES
In addition to the core courses and coursework needed to meet departmental requirements, we expect that students would take 1-2 additional courses in other departments in their first and second years in order to remedy gaps in their training and prepare them for interdisciplinary research. For example, most students entering the Zoology and Computer Science departments will not have obtained much background in cognitive psychology (including psycholinguistics), and yet they will need to have a deeper familiarity with core concepts and theoretical perspectives in this field than we can give in the introductory cognitive science course. Similarly, students outside of Zoology will need to be familiar with concepts in evolution and behavioral ecology.Following is a list of courses that are likeliest to serve as cross-disciplinary courses. These courses are included among the requirements of students in the respective departments, but students within each discipline will likely take additional courses not on the list. Note that this list does not include courses in statistics, mathematics, and programming that individual students might need to take in support of their research training.
Computer Science
- CSE 802 Pattern Recognition
- CSE 803 Computer Vision
- CSE 841 Intro to Artificial Intelligence
- CSE 847 Machine Learning
- CSE 848 Evolutionary Computation
- CSE 941 Selected Topics in AI
Psychology
- PSY 801 Perception
- PSY 802 Low-level cognitive processes
- PSY 803 High-level cognitive processes
- PSY 814 Psycholinguistics
Zoology
- ZOL 822 Behavior: Animal Communication
- ZOL 822 Behavior: Social Behavior
- ZOL 849 Evolutionary Biology Modeling and Simulation
- ZOL 896 Population and Community Ecology
Other Departments
- LIN 834 Linguistics: Syntactic Theory
- FW 852 Fisheries and Wildlife: Systems