Since the beginning of AI, mind games have been studied as relevant application fields. Nowadays, some programs are better than human players in most classical games. Their results highlight the efficiency of AI methods that are now quite standard. Such methods are very useful to Go programs, but they do not enable a strong Go program to be built. The problems related to Computer Go require new AI problem solving methods. Given the great number of problems and the diversity of possible solutions, Computer Go is an attractive research domain for AI. Prospective methods of programming the game of Go will probably be of interest in other domains as well. The goal of this paper is to present Computer Go by showing the links between existing studies on Computer Go and different AI related domains: evaluation function, heuristic search, machine learning, automatic knowledge generation, mathematical morphology and cognitive science. In addition, this paper describes both the practical aspects of Go programming, such as program optimization, and various theoretical aspects such as combinatorial game theory, mathematical morphology, and MonteCarlo methods. B. Bouzy T. Cazenave page 2 08/06/01 1.

After their beginnings in computer-aided instruction, automated tutors have re-emerged as intelligent tutoring systems. These intelligent tutors have obtained considerable success by using results from cognitive psychology and artificial intelligence to permit non-traditional instruction which is tailored to their individual students. The success of these automated tutors is due to their precise understanding and modeling of both the student and the domain being taught. A common measure of the robustness of an automated tutor is the size of the domain that it can understand. The schema-based Prolog tutor described in this dissertation is capable of recognizing a larger class of programs than existing Prolog tutors. By using powerful generalized transformations, our Prolog tutor can generate this class of programs from a very small set of normal form programs. Thus, our Prolog tutor recognizes a larger class of programs using fewer normal form programs than existing Prolog tutors. One o...

Go provides artificial intelligence (AI) and cognitive science researchers with an easily specified formal domain in which skills of human intelligence cannot be matched by currently known programming techniques. Go is a much more widely played game than chess (principally in Japan, Korea and China), yet it is not well known to AI and cognitive science researchers and our goal in this paper is to introduce some of the challenges of the game to the AI community in the form of a comparison with chess. Go has been called a possible “task par excellence for AI ” by Berliner [1] and we conclude that Go is a domain in which the development of new programming techniques is not only possible but is in fact necessary. Computer programs are challenging human performance in almost every level of gameplaying endeavour: the world backgammon champion is a neural network program [7]. The performance of chess programs- originally the drosophila of research into search techniques in AI- is at the Grandmaster level: in 1994, the highest rated chess player in the world, Kasparov,

Research on expertise in visual-spatial tasks such as chess has ignored the introspective reports of top-level practitioners, overemphasized pattern recognition as the sole mechanism underlying skilled performance, and neglected the role of mental imagery in the thinking process. In addressing these limitations I propose a new theory of expertise, the “mental cartoons hypothesis,” and illustrate its properties by application to chess, which has historically been the primary domain for psychological studies of human expertise. Experiment 1 uses a classic image-scanning paradigm to show that chess masters and novices differ substantially in their ability to visualize chess moves, even in semantically impoverished contexts, extending the range of novice-expert differences from pattern recognition and knowledge representation to mental imagery processing. Experiments 2 and 3 exploit original very-long-term memory recognition and recall tasks to show that the memory representations of famous chess positions held by chess masters

The study of expertise has a very long history that has been discussed in several other chapters in this handbook (Ericsson, Chapter 1.1; Amirault & Branson, Chapter 2.2). The focus of this chapter is on influential developments within cognitive science and cognitive psychology for over the last three decades. Our chapter consists of two parts. In

This paper presents a simple but very powerful technique to support user interface evaluation along with a prototype open environment --- I-Observe, the Interface OBServation, Evaluation, Recording, and Visualization Environment --- which supports a preliminary implementation of this technique. This technique operates by recording user interface sessions in multiple modalities, both as a trace of interesting events and through video images. It then provides tools to allow the user interface evaluator to combine these modalities, analyzing the event stream to search for patterns of interesting or important user actions, then using the recorded timestamps associated with these actions to present only the sections of the video recording of interest. This allows, for example, all places where the user invokes a help system or a particular command to be observed without requiring the evaluator to manually search the recording or sit through long sessions of unrelated interactions. By combin...

To account for the large demands of working memory during text comprehension and expert problem solving it is proposed that the traditional models of working memory involving temporary storage have to be extended to include a long-term working-memory portion. According to the proposed theoretical framework cognitive processes are viewed as a sequence of stable states representing end products of processing. In skilled activities these end products are stored in long-term memory and kept directly accessible by retrieval cues in short-term memory as proposed by skilled memory theory. These

This paper addresses Spatial Reasoning in the game of Go. Combinatorial complexity of the game of Go obliges the Computer Go community to study spatial representations of human players that are complex. These representations contain the concepts of grouping and fractioning. Spatial Reasoning in Go is qualitative. It creates objects and relationships between them in order to qualify them. This description of Spatial Reasoning in Go is a positive contribution to the SR community. Keywords Spatial Reasoning, Objects, Relationships, Game of Go. 1. Introduction Classical reasoning appears to be linked with Natural Language (NL) as human beings are obliged to use words to express their reasonings. NL has been growing between human beings in order to make them communicating, therefore the human body does not provide a communicating tool as efficient as NL is. But other kinds of reasoning exist : painting, drawing, calligraphy, spatial gesture, dance are good way to express spatial informat...

Although virtually any human activity can be viewed as the solving of a problem, throughout the hstory of the study of problem solving, most research has concerned tasks that take minutes or hours to per-form. Typically subjects make many observable actions during this period, and these actions are interpreted as the externally visible part of the solution process. Even if subjects are required to solve problems in their heads (for example, to mentally multiply 135 x 76), they are usually asked to talk aloud as they work, and the resulting verbal protocol is interpreted as a sequence of actions (see chapter 1). Thus the tasks studied are not only long tasks but also multistep tasks. The earliest experimental work on human problem solving was done

In this thesis, we try to analyze the feasibility of applying neural networks, especially the temporal difference (TD) learning model, in the game of Go. First, a framework of decomposition of computer Go into smaller problems is given. Previous applications of neural networks and TD learning in Backgammon and Go are then given. We derive