Ebook: Artificial Intelligence in Education

The 13th International Conference on Artificial Intelligence in Education (AIED-2007) is being held July 9–13, 2007, in Los Angeles, California. AIED Conferences are organized by the International AIED Society on a biennial basis. The goal of the International Artificial Intelligence in Education (AIED) Society is to advance knowledge and promote research and development in the field of Artificial Intelligence in Education. AIED is an interdisciplinary community at the frontiers of the fields of computer science, education and psychology. It promotes rigorous research and development of interactive and adaptive learning environments for learners of all ages, across all domains. The society brings together a community of members in the field through the organization of the AIED Conferences, a Journal, and other activities of interest. The AIED conferences are the main International forum for reporting the best international research in the field of AI in Education. The conferences provide the opportunity for the exchange of information and ideas on related research, development and applications. Previous conferences have been held in Kobe, Japan in 1997; Le Mans, France in 1999; San Antonio, USA in 2001; Sydney, Australia in 2003 and Amsterdam, The Netherlands in 2005.

Each conference adopts a theme that reflects the interests of the community and its role at the cutting edge of research. The 2007 theme is: Building Technology Rich Learning Contexts that Work. The nature of technology has changed since AIED was conceptualised as a research community and Interactive Learning Environments were initially developed. Technology is smaller, more mobile, networked, pervasive and often ubiquitous as well as being provided by the standard desktop PC. This creates the potential for technology supported learning wherever and whenever learners need and want it. However, in order to take advantage of this potential for greater flexibility we need to understand and model learners and the contexts with which they interact in a manner that enables us to design, deploy and evaluate technology to most effectively support learning across multiple locations, subjects and times. The AIED community has much to contribute to this endeavour.

Here are some statistics: Overall, we received 192 submissions for full papers and posters. 60 of these (31%) were accepted and published as full papers, and a further 52 are included here as posters. Full papers each have been allotted 8 pages in the Proceedings; posters have been allotted 3 pages. The conference also includes 2 interactive events, 10 workshops, 5 tutorials, and 16 papers in Young Researcher's Track. Each of these has been allotted a one-page abstract in the Proceedings; the workshops, tutorials, and YRT papers also have their own Proceedings, provided at the conference itself. Also in the Proceedings are brief abstracts of the talks of the four invited speakers:

Roxana Moreno, University of New Mexico

Tak-Wai Chan, National Central University of Taiwan

Danaë Stanton Fraser, University of Bath in the United Kingdom

Gregory Abowd, Georgia Institute of Technology

For the first time in the AIED conference we instituted a meta-review process. Thanks to our Senior Program Committee members, this change went quite smoothly. We believe that not only was the quality of the reviewing better, but the reviewing process was more rewarding with reviewers able to see their colleagues views and, if needed, discuss differences of opinion. Thanks, also, to the reviewers who were recruited by the Senior Program Committee members to help out in this critical task.

We would like to thank the many people who helped make the conference possible. Firstly members of Lewis Johnson's Local Organizing Committee, Jihie Kim, Andre Valente, Carole Beal and Chad Lane. Our Sponsorhsip chair Art Graesser and Rebecca Campbell, our copy editor. Special thanks to Jim Greer, our AIED Society President, who made sure we stayed on schedule. The committees organizing the other events at the conference have helped to make the conference richer and broader. Young Researcher's Track, chaired by Judith Good and Kaska Porayska-Pomsta; Tutorials, chaired by Roger Azevedo and Carolyn Rosé; and Workshops chaired by Ivon Arroyo and Joe Beck.

For those who enjoyed the contributions in this Proceedings, we recommend considering joining the International Society for Artificial Intelligence in Education: http://www.iaied.org. We certainly hope that you all enjoy the AIED-2007 conference.

Kenneth R. Koedinger, Carnegie Mellon University, USA

Rosemary Luckin, University of London, UK

Program Committee Co-Chairs

Users typically navigate the state space of a task through the explicit manipulation of its constituent elements. This paper proposes the Principle of State Expansion (PoSE), a novel alternative to such explicit manipulation. Under the PoSE, users manipulate a collection of states and navigate through the state space by selecting states to expand. An abstract task is presented that contrasts the approaches and demonstrates the generality of the PoSE. This is followed by discussion of WordBird and X-Panda, two concrete implementations of software that follow the PoSE. In conclusion, the advantages over explicit manipulation are discussed especially with reference to the potential for sophisticated learner modelling.

Teaching problem-solving in formal domains aims at two purposes: 1) increasing the students' skill in addressing the problem in a goal-oriented way, and 2) increasing their competence in expressing themselves formally. In a dialog, suitable tutoring strategies addressing both issues may be quite delicate, especially when meeting formally inaccurate or even faulty statements which nevertheless are useful from the problem-solving perspective. Based on evidence from Wizard-of-Oz studies on human tutoring in mathematical theorem proving, we have developed a model for addressing formally inaccurate statements within a problem-solving context. Based on an error recognition and correction module, the model performs a conceptually-motivated error categorization and generates a suitable response.

In this paper we describe ASPL, Advanced Semantic Platform for Learning, designed using the Magpie framework with an aim to support students learning about the Semantic Web research area. We describe the evolution of ASPL and illustrate how we used the results from a formal evaluation of the initial system to re-design the user functionalities. The second version of ASPL semantically interprets the results provided by a non-semantic web mining tool and uses them to support various forms of semantics-assisted exploration, based on pedagogical strategies such as performing lateral steps and query filtering.

We have modeled changes in electroencephalography (EEG) - derived measures of cognitive workload, engagement, and distraction as individuals developed and refined their problem solving skills in science. Subjects performing a series of problem solving simulations showed decreases in the times needed to solve the problems; however, metrics of high cognitive workload and high engagement remained the same. When these indices were measured within the navigation, decision, and display events in the simulations, significant differences in workload and engagement were often observed. In addition, differences in these event categories were also often observed across a series of the tasks, and were variable across individuals. These preliminary studies suggest that the development of EEG-derived models of the dynamic changes in cognitive indices of workload, distraction and engagement may be an important tool for understanding the development of problem solving skills in secondary school students.

TuTalk supports the rapid development of dialogue agents for learning applications. It enables an experimenter to create a dialogue agent with either minimal or no programming and provides the infrastructure needed for testing hypotheses about dialogue. Our main goals in developing this tool were to provide 1) an authoring interface and language for setting up the domain knowledge and resources needed to support the agent and 2) a plug-and-play type of system that facilitates the integration of new modules and experimentation with different core modules. In this paper we describe the authoring tool and the usability studies that have shaped its design, the dialogue that is supported and features of the authoring language and their usage history.

Women's under-representation in fields such as engineering may result in part from female students' negative beliefs regarding these fields and their low self-efficacy for these fields. Empirical evidence indicates that computer-generated interface agents are effective in influencing students' interest, motivation, attitudes, and self-efficacy. Hence, in this experimental study, we investigated the potential of interface agents to serve as effective social models for changing attitudes regarding the utility of math and the hard sciences and self-efficacy for these fields. 113 middle-school students interacted with either a female or a male computer-generated interface agent or they did not interact with an interface agent. The findings from this study indicate that interface agents may be used effectively as social models for influencing middle school students' attitudes and beliefs about mathematics and the hard sciences and their mathematical ability. Nevertheless, the efficacy of the agent depended on the characteristics of the agent with the female agent tending to be the most effective regardless of the subject gender.

We describe the design and evaluation of an affective pedagogical agent persona for Intelligent Tutoring Systems. The goal of our research was to develop an agent embodying a persona of a caring mentor interested in the learner's progress. The agent's behaviour is guided by a set of rules that are triggered by the states of the session history. Four agents were integrated with EER-Tutor for a formative evaluation study. The mentor persona secured strong rapport with the users; the audible narration was seen as a strong feature of the agents.

The Tactical Language and Culture Training System (TLCTS) helps learners acquire basic communicative skills in foreign languages and cultures. Learners acquire communication skills through a combination of interactive lessons and serious games. Artificial intelligence plays multiple roles in this learning environment: to process the learner's speech, to interpret and evaluate learner actions, to control the response of non-player characters, to generate hints, and to assess the trainee's mastery of the skills. AI is also used to assist in the authoring process to assist in the generation and validation of lesson content. This paper gives an overview of the current system, and describes the experience to date in transitioning the system from research prototype into a training system that is in regular use by thousands of users in the United States and elsewhere.

The authors of topic map-based learning resources face major difficulties in constructing the underlying ontologies. In this paper we propose two approaches to address this problem. The first one is aimed at automatic construction of a “draft” topic map for the authors to start with. It is based on a set of heuristics for extracting semantic information from HTML documents and transforming it into a topic map format. The second one is aimed at providing help to authors during the topic map creating process by mining the Wikipedia knowledge base. It suggests “standard” names for the new topics (paired with URIs), along with lists of related topics in the considered domain. The proposed approaches are implemented in the educational topic maps editor TM4L.

This paper describes an adaptive system called VL-PATSy, an extension to an existing system (PATSy) that adds a mechanism for serving vicarious learning (VL) resources. Vicarious learning is the notion that people can and will learn through being given access to the learning experiences of others. The VL resources represent an innovative form of learner support for a complex cognitive task (clinical reasoning). The VL resources are delivered in response to system-detected reasoning impasse events or in response to student demand in timely and context-sensitive ways. A three tier XML-oriented, rule-based architecture was chosen as an efficient and lightweight implementation solution and a design pattern approach was utilised.

Previous research on the use of diagrams for argumentation instruction has highlighted, but not conclusively demonstrated, their potential benefits. We examine the relative benefits of using diagrams and diagramming tools to teach causal reasoning about public policy. Sixty-three Carnegie Mellon University students were asked to analyze short policy texts using either: 1) text only, 2) text and a pre-made, correct diagram representing the causal claims in the text, or 3) text and a diagramming tool with which to construct their own causal diagram. After a pretest and training, we tested student performance on a new policy text and found that students given a correct diagram (condition 2 above) significantly outperformed the other groups. Finally, we compared learning by testing students on a third policy problem in which we removed all diagram or tool aids and found that students who constructed their own diagrams (condition 3 above) learned the most. We describe these results and interpret them in a way that foreshadows work we now plan for a cognitive-tutor on causal diagram construction.

Previous research has highlighted the advantages of graphical argument representations. A number of tutoring systems have been built that support students in rendering arguments graphically, as they learn argumentation skills. The relative tutoring benefits of graphical argument representations have not been reliably shown, however. In this paper we present an evaluation of the LARGO system which enables law students graphically to represent examples of legal interpretation with hypotheticals they observe while reading texts of U.S. Supreme Court oral arguments. We hypothesized that LARGO's graphical representations and advice would help students to identify important elements of the arguments (i.e., proposed hypotheses, hypothetical challenges, and responses) and to reflect on their significance to the argument's merits better than a purely text-based alternative. In an experiment, we found some empirical support for this hypothesis.

Students' actions while working with a tuoring system were used to generate estimates of learning goals, specifically, the goal of learning by using multimedia help resources, and the goal of learning through independent problem solving. A Dynamic Bayesian Network (DBN) model was trained with interface action and inter-action interval latency data from 115 high school students, and then tested with action data from an independent sample of 135 students. Estimates of learning goals generated by the model predicted student performance on a post-test of math achievement, whereas pre-test performance did not.

MathGirls is a pedagogical-agent-based environment designed for high school girls learning introductory algebra. Since females are in general more interested in interactive computing and more positive about the social presence of pedagogical agents, the environment provides a girl-friendly social learning environment, where pedagogical agents encourage the girls to build constructive views of learning math. This study investigated the impact of agent presence on changes in the girls' math attitude, their math self-efficacy, and their learning; on the girls' choice of their agents; and, on their perceptions of agent affability. The results revealed that the girls with an agent developed a more positive attitude and increased self-efficacy significantly, compared to the girls without an agent. Both groups showed significant increase in their learning. The study also showed that the girls chose female agents significantly more than male agents as their learning partners and perceived peer-like agents as more affable than teacher-like agents. The finding confirms the instructional value of pedagogical agents functioning as a social cognitive tool [1].

The primary goal of this study was to investigate the role of feedback in an intelligent tutoring system (ITS) with natural language dialogue. One core component of tutorial dialogue is feedback, which carries the primary burden of informing students of their performance. AutoTutor is an ITS with tutorial dialogue that was developed at the University of Memphis. This article addresses the effectiveness of two types of feedback (content & progress) while college students interact with AutoTutor on conceptual physics. Content feedback provides qualitative information about the domain content and its accuracy as it is covered in a tutoring session. Progress feedback is a quantitative assessment of the student's advancement through the material being covered (i.e., how far the student has come and how much farther they have to go). A factorial design was used that manipulated the presence or absence of both feedback categories (content & progress). Each student interacted with one of four different versions of AutoTutor that varied the type of feedback. Data analyses showed significant effects of feedback on learning and motivational measures, supporting the notion that “content matters” and the adage “no pain, no gain.”