Brian Michael Slator

Brief Professional History: I studied Computer Science and Artificial Intelligence in the Computing Research Laboratory (CRL) at New Mexico State University under Dr. Yorick Wilks, Director. CRL was at the time rated among the top 5 AI research institutions in the world in Natural Language Processing. I was awarded a doctorate from the Computer Science Department at NMSU in 1988. I was a tenure-track faculty member at North Dakota State University for two years, 1988-1990, while also a summer research employee at CRL, and then was on the research faculty at The Institute for the Learning Sciences at Northwestern University for six years, 1990-1996. ILS was at one time the largest AI and Educational Media Research Center in the United States. I am currently, 1996-present, once again on the tenure-track teaching faculty at North Dakota State University.

My research interests follow two diverse but related tracks:

1. Computational Linguistics, in particular lexical semantics, where my doctoral studies originated, and where I was an instrumental part of two NSF research grants (1989-1994). In January of 1996, MIT Press published my (co-authored) definitive survey of the field, Electric Words: Dictionaries, Computers, and Meanings , and in the same month the Communications of the ACM published a survey article by the same authors; also in 1996, I co-edited a special theme issue of "IEEE Computer" devoted to Interactive Natural Language Processing which appeared in July; and
2. Educational Multimedia, currently my primary interest, in particular:
   • virtual role-playing environments where students are immersed in authentic game-play simulations that teach real-world strategies, and domain specific concepts and procedures; and
   • integrated performance support systems where users are provided job-aid systems that support work through an analysis of process and expert help and advice for just-in-time training in the workplace.

My appointment at Northwestern was in the Institute for the Learning Sciences (ILS) where I researched and developed multi-media applications in educational technology and case-based reasoning. In this work I designed and managed the development of a number of cutting edge systems for intelligent tutoring and job-aid style performance support. The application areas for these systems were mostly business oriented: tax planning, management consulting, retail sales, internal auditing and so forth. In doing this work, I amassed extensive experience in developing practical artificial intelligence applications. Moreover, the theoretical elements in these systems ranged through a variety of interesting conceptual arenas:

• knowledge acquisition from subject matter experts;
• knowledge representation and indexing theories;
• classification of questions and conversational categories;
• task analysis and process modeling;
• simulated environments (for role playing);
• user interface design and implementation;
• the appropriate use of digital video as a tool for teaching; and,
• design and implementation of tools for the construction of educational media, indexing and knowledge engineering.
• software agents for intelligent case-based and deductive tutoring
• agent-based simulations of virtual environments
• science education in virtual environments

These several research and development techniques have not been randomly employed, but rather are focused towards the design and implementation of a particular flavor of multi-media educational environment. These systems feature a strong model of task, either through a process flow analysis or a domain simulation, and give the student/user an authentic problem to solve within some particular domain --- putting the student in control of their experience. The systems monitor progress (not by modeling students, but by modeling tasks), and stand ready to provide help and expert advice at felicitous points in the interaction.

Short descriptions of some of those ILS research projects are available online, as well as an inventory of the NDSU research projects I am currently engaged in or planning.

Through the application of these ideas and the coalescence of these theoretical streams, I have developed a research agenda of my own which can be characterized as follows:

• teaching systems in general should promote active, not passive, learning
• tutoring systems should capitalize on the natural human ability to "role-play" as a way to teach through cases and exemplify the axiom "law school is not to learn the law, but to learn to think like a lawyer".
• given that the best way to learn is to "learn by doing", and the best kind of remediation is "failure based", learning environments should be constructed where the tasks are authentic, the "doing" is interesting and, most importantly fun, and where the failures are instructive;
• tutoring systems should be based on modeling tasks, not on modeling students;
• instruction and remediation should be organized around a model of mentoring and shared expertise.

As a consequence of these experiences and the application of these theories, I believe the best way to implement intelligent tutoring is through the development of a certain class of educational game that I will call "virtual role-playing environments". In these pedagogical applications, students are thrust into realistic (and fun) simulated environments and given an authentic problem-solving goal that is both immediate and integral to their survival in the simulated world (along with tools and resources to enable them to succeed). The system is designed to assist and tutor, however remediation and instruction are always at the discretion of the student, not the system, and the thrust of the material is towards "how to think like an X" strategies, rather than "how to do X" procedural advice.

Role playing games immerse the student in a simulated environment where their success at learning their role in the simulation is directly reflected in their success at the game and, crucially, this role-play also serves to communicate the thinking skills needed to understand the corresponding real world task. For example, a student will learn the principles of micro-economics if they are engaged in a world where they are expected to manage a retail establishment. But that world must be active and not simply reactive. The student, of course, must be able to "click on" the interface, but the successful game is one that "clicks on" the player in return. Experience shows that if the environment is stimulating and sufficiently engaging, players will throw themselves at it, terrier-like. This is the operational goal of any teaching system.

In order to meet pedagogical goals, the student's ability to understand and employ the strategies, and grasp the concepts and procedures of the domain must be directly reflected in their ability to keep their retail establishment afloat. For this to work, in turn, the environment must be carefully crafted, and should act and react as realistically as possible. The student's success or failure is then based on their ability to understand the competitive nature of the situation and their ability in turning a (simulated) profit. The system's success or failure is, in turn, based on its ability to engage the student.

A paper describing a particular instantiation of these ideas can be downloaded from:

Success in learning systems of this sort do not depend on mastering trivial tasks, or figuring out the tricks or vagaries of the interface, or guessing the answers to multiple choice questions as a grading scheme to evaluate student performance --- the student's success is a function of their success at the game, and thereby their success at grasping the underlying real-world strategies, and the domain's concepts and procedures. The system is crafted such that students cannot succeed unless they learn what is needed to prosper, and they will not fail if they do the right things. However, the system is not primarily designed to introduce and reinforce facts, it is designed to exemplify strategies and methods of problem solving, i.e. press home the lesson of "thinking like a retailer".

Performance support systems are in many ways similar, but are also different in important ways. In this pedagogical context, a user brings with them the goals of their task but also their training; they are already somewhat proficient at their jobs, and what they need is a task-based environment that supports their work in useful ways. This is accomplished through a user-centered task analysis, an infrastructure of forms and tools, and a system of context-sensitive help and advice that provides support and answers at the time needed.

For example, when an internal auditor is planning an engagement, they need forms to focus their thinking and they need access to relevant previous cases in order to inform their planning and sharpen their understanding of the task at hand. A useful performance support system will provide this material at the right time and through an interface that is transparent and felicitous, so that work can be accomplished in a timely and effective way. To manage this support, the job-aid system for internal auditing must have a representation of two key things:

1. the tasks expected of the auditor, and
2. the processes being audited.

This structured support, combined with case retrieval and expert advice, form the backbone of a special type of tutorial, job-aid system that coalesces case-based retrieval with structured task analysis for the purpose of performance support.

Teaching technologies have progressed far beyond the page-turning systems of years past, to an era of intelligent tutoring, expert retrieval, learning by doing, case-based instruction, and simulation-based environments for student experiences and remediation. These are fundable pursuits where practical results can be shown, and they address important social and educational imperatives.