
Professor Rebecca Eynon
Professor of Education, the Internet and Society
Rebecca Eynon's research focuses on learning and the Internet, and the links between digital and social exclusion.
This one day workshop will bring together social scientists and computer scientists to discuss the possibilities for companions and pedagogical agents to enhance learning and education both now and in the future. The workshop will be based around six presentations allowing time for discussion throughout the day. Four broad themes will inform the debate:
The Motivational Environments research agenda is advancing Human-Computer Interaction methods, theories, technologies, architectures, and environments that incorporate wearable and ambient sensors, responsive spaces, robots and relational agents to promote creativity, teamwork, learning, and motivation. Affective Learning Companions, real-time multi-modal characters are beginning to sense and respond to learners’ affective cues. Game As Life is extending avant-garde ‘real-life games’ and blurring motivations in everyday digital-physical-social environments. Astronaut Robot Mission Simulators and Pet-Building activities are promoting project based teams and education through novel Human-Robot Interactions. Creativity research methodologies and support tools are being advanced through empirical investigations in research organizations and everyday life experience and through the development computer supported collaborative environments. This suite of technologies and activities is forming a transdisciplinary framework and context to study and foster life-long motivation, learning, teamwork, and creativity.
Learning companions and pedagogical agents are frequently developed within interdisciplinary research projects, with considerable demands placed upon their evaluation. This can result in a tiresome burden of instruments and activities for the user, particularly if the evaluation is provided as a discrete, separate activity. Here, an alternative approach to the evaluation of pedagogical agents is presented. This approach developed within the eCIRCUS project places the evaluation within the learning experience rather than as a separate event. The presentation will discuss the evaluation of ORIENT (Overcoming Refugee Integration with Empathic Novel Technology) an educational role play environment aimed at 13-14 year olds and focused on friendship dynamics and integration. The key results from this approach and its impact on the user experience will be outlined. Finally, the benefits and costs of using in-context evaluation for other learning companions and pedagogical agents will be discussed.
Natalie Person: From Auto Tutor to Guru: An Overview of Intelligent Tutoring Systems with Natural Dialogue and Emotion Detection
Over the past 12 years, researchers at the Institute for Intelligent Systems have been developing intelligent tutoring systems that simulate the teaching strategies of effective human tutors. Systems have been developed for various domains including computer science, physics, critical thinking, and most recently, biology. All of these systems include pedagogical agents that assist learners as they interact with multimedia interfaces. The ITSs that we have developed are specifically designed to simulate the pedagogical dialogue of human tutors and to engage students in conversations about various domain topics. Results from extensive outcome studies indicate that students do learn when interacting with our systems (effect sizes of 0.4 to 1.0). Our most recent system GURU simulates the dialogue of expert human tutors and incorporates a pedagogical agent that converses with the learner to help them learn biology. Guru is equipped with a camera, eyetracker, and posture sensor that are used to detect student engagement and emotion.
This talk presents recent research in a multi-year effort to optimize the design of dynamic support for collaborative design learning in the domain of thermodynamics, using an integration of the Virtual Math Teams (VMT) collaboration environment developed at Drexel University, with pedagogical agents developed at Carnegie Mellon University that are triggered dynamically in the midst of the collaboration by means of an automatic analysis of the collaborative discussion as it unfolds. Using an accompanying simulation environment developed at Northwestern University, students are able to implement design ideas using graphical interface widgets, and to explore the relationships between the settings of various parameters within the design. They share their design ideas with one another through the collaboration environment. The pedagogical agents support their reflection and learning in the midst of these design discussions. In the recent study Carolyn will describe, they specifically focused on issues related to ‘Green Design’ with Rankine cycle based power plants. In the collaborative design exercise, students worked in pairs to struggle with trade-offs between power output and environmental friendliness in the design of thier power plant. The experimental manipulation contrasted alternative strategies for integrating the pedagogical agents with the student discussions in terms of learning gains, perception of collaboration and of the collaboration environment, as well as dynamics of the conversations themselves.
The aim of the Learning Companion project at the University of Oxford is to develop a technical tool that helps adults to make more productive use of the web for their personal, self-directed learning, which they might undertake simply for their own satisfaction, or in order to improve their personal circumstances in some practical way. The tool, called the Learning Companion, encompasses a computer-based ‘conversational agent’ – in effect, an animated speaking character on a computer screen – designed to give practical support, guidance and focus to the independent learning activities of adults who currently lack the confidence or the opportunity to take part in organised learning, but who might be prepared to study topics of their own choice in their homes using the Internet. In contrast to earlier work in this area, the domain expertise (apart from generic IT skills) is not provided as part of the learning environment. Instead the learning companion encourages the user to engage with the variety of information sources and expertise on the Internet. Thus its educative role will be primarily in providing responsive and enabling interactions that lead learners towards taking control over their own learning. This presentation will report the findings from phase 1 of the research that aims to evaluate the feasibility of the Learning Companion by using a ‘Wizard of Oz’ methodology: an experimental procedure which elicits data of how human subjects engage in dialogue with computers. This presentation will discuss the kinds of interactions adult learners had with the tool, exploring how individuals engage in simulated human-machine conversations for educational purposes, how the Learning Companion is perceived by the target group and consider if and to what extent such a Learning Companion can facilitate learning despite not teaching any domain-specific content knowledge.
Professor of Education, the Internet and Society
Rebecca Eynon's research focuses on learning and the Internet, and the links between digital and social exclusion.
Former Visiting Fellow
Yorick Wilks has interests in artificial intelligence and the computer processing of language, knowledge and belief. His current research focuses on the possibility of software agents having identifiable personalities.
Assistant Professor of Human Computer Interaction, School of Computing and Informatics and the Arts, Media, and Engineering graduate program, Arizona State University
Department of Educational Studies, University of Oxford
Principal Lecturer for Reachout, Department of Computing, Engineering and Technology, University of Sunderland
Associate Professor of Psychology and Chair, Department of Psychology, Rhodes College
Assistant Professor in the School of Computer Science, Language Technologies Institute and the Human-Computer Interaction Institute, Carnegie Mellon University