It is widely acknowledged that one-on-one human tutoring is one of the most effective ways to provide learning, however, the source of its effectiveness is still unclear. Tutor-centered, student-centered, and interaction hypotheses have been proposed as possible explanations of the effectiveness of human tutoring. Most research has addressed this question by analyzing tutorial sessions at the dialogue move or speech act level. The present paper adopts a different approach by focusing on word usage patterns in 50 naturalistic tutorial sessions between human students and expert tutors. Specifically, each unique word in the session was designated as a student initiative word, a tutor initiative word, or a shared-initiative word. Comparisons of the frequencies as well as the weights of the words assigned to each of these categories indicated that the student and tutor share initiative even though the tutor’s are considerably more verbose. The implications of the results for the development of an ITS that aspires to model expert tutors are discussed.

We revisit a framework for designing peer-based intelligent tutoring systems motivated by McCalla's ecological approach, where learning is facilitated by the previous experiences of peers with a corpus of learning objects. Prior research demonstrated the value of a proposed algorithm for modeling student learning and for selecting the most beneficial learning objects to present to new students. In this paper, we first adjust the validation of this approach to demonstrate its ability to cope with errors in assessing the learning of student peers. We then deepen the representation of learning objects to reflect the expected time to completion and demonstrate how this may lead to more effective selection of learning objects for students, and thus more effective learning. As part of our exploration of these new adjustments, we offer insights into how the size of learning object repositories may affect student learning, suggesting future extensions for the model and its validation.

Creating natural language tutorial dialogue systems that realize effective strategies is a central challenge for intelligent tutoring systems research. Traditional approaches generally require large development time, do not generalize well across domains, and do not match the flexibility and natural language sophistication of human tutors. A promising approach that may offer several benefits is data-driven system development, in which a dialogue policy is learned from corpora of human tutorial dialogue. To date these learning approaches typically focus on optimizing the tutor’s choice of act, and do not explicitly model the instances in which the tutor chose not to act. This paper reports on a hidden Markov modeling (HMM) approach within human textual tutorial dialogue that explicitly represents the tutors’ choices not to intervene. The results show that an HMM that models tutor non-interventions predicts tutor moves significantly better than a model that does not explicitly represent the non-interventions. The findings have implications for automatically modeling tutorial strategies and for learning dialogue policies from corpora.

Intelligent tutoring systems (ITS) is a multidisciplinary field of study that draws upon artificial intelligence, computer science, and cognitive science to create computerized tutoring systems that offer immediate feedback and individualized instruction. Broadly construed, most ITSs can be characterized as having two loops: an outer loop and an inner loop. The outer loop intelligently selects the next relevant task for the student to complete. The inner loop iterates over individual problem-solving steps and provides contextually relevant feedback and instructional guidance. The ultimate goal of an ITS is to promote deep learning that persists over time, transfers to new domains, and accelerates future learning.

Behavior-based and reactive control methods are popular choices for building fast and lightweight intelligent controllers for resource-constrained systems. Reactive methods are extremely useful in highly resource-constrained applications, but at a cost: they tend to be even more susceptible to certain types of failures than deliberative techniques. Without a planner to adapt to changes, even a small failure can result in incorrect behavior from the entire controller. In this paper, we propose extensions to behavior-based subsumption that can detect four types of failures.

A typical ARG consists of a Puppet Master who issues that have historically prevented ARGs from designs the game and informs players of the unfolding of mainstream adoption. A generic game engine runs on a the story. With the advent of smart-phones with GPS, geo-location enabled mobile device enables users to play ARGs progressively make use of the actual world as the any game modeled as a dependency graph of game content.

In a competitive game it is important to identify the opponent's strategy as quickly and accurately as possible so that an effective response can be planned. In this vein, this paper summarizes our work in exploring using first order inductive learning to learn rules for representing opponent strategies. Specifically, we use these learned rules to perform plan recognition and classify an opponent strategy as one of multiple learned strategies. Our experiments validate this novel approach in a simple real-time strategy game.

Games are an integral part of the human experience. Starting in our childhood and continuing throughout our lives they teach us about the world through the concepts of rules, strategies, and outcomes. They help prepare us for our future, provide entertainment, bring us together socially, and give us characters to root for -- making ordinary people heroes for a moment. Digital games build on centuries of play and interaction bringing to the modern age a unique and creative form. Fully integrated into modern life, the video game industry now rivals that of the motion picture and music industries and their products are fully integrated into our digital lifestyles. Computers with advanced graphics capabilities have contributed to the immersive interactive experience that attracts many to spend as much of their leisure time playing video games as watching television or listening to music.

Data mining has primarily focused on statistical properties of data alone and not necessarily on what could be done with the patterns. While there has been some work on measuring usefulness of patterns in decision making but not on using such measures for driving the mining process. We introduce a framework to mine clusters that support decision making. We use an extrinsic measure that evaluates patterns based on their utility in decision making. We show empirical validationof our approach on several test domains.

There are two primary types of graph-based data miners: frequent subgraph and compression-based miners. With frequent subgraph miners, the most interesting substructure is the largest one (or ones) that meet the minimum support. Whereas, compression-based graph miners discover those subgraphs that maximize the amount of compression that a particular substructure provides a graph. The algorithms associated with these two approaches are not only different, but they also may result in dramatic performance differences, as well as in the normative patterns being discovered. In order to compare these two types of graph-based approaches to knowledge discovery, in the following sections we will compare two publicly available applications: GASTON and SUBDUE.