Collaborative learning with affective artificial study companions in a virtual learning environment

This research has been carried out in conjunction with Chapeltown and Harehills Assisted Learning Computer School (CHALCS) and local schools. CHALCS is an 'out-of-hours' school in a deprived inner-city community where unemployment is high and many children are failing to meet their educational potential. As the name implies CHALCS provides students with access to computers to support their learning. CHALCS relies on many volunteer tutors and specialist tutors are in short supply. This is especially true for subjects such as Advanced Level Physics with low numbers of students. This research aimed to investigate the feasibility of providing online study skills support to pupils at CHALCS and a local school. Research suggests that collaborative learning that prompts students to explain and justify their understanding can encourage deeper learning. As a potentially effective way of motivating deeper learning from hypertext course notes in a Virtual Learning Environment (VLE), this research investigates the feasibility of designing an artificial Agent capable of collaborating with the learner to jointly construct summary notes. Hypertext course notes covering a portion of the Advanced Level Physics curriculum were designed and uploaded into a WebCT based VLE. A specialist tutor validated the content of the course notes before the ease of use of the VLE was tested with target students. A study was then conducted to develop a model of the kinds of help students required in writing summary notes from the course-notes. Based on the derived process model of summarisation and an analysis of the content structure of the course notes, strategies for summarising the text were devised. An Animated Pedagogical Agent was designed incorporating these strategies. Two versions of the agent with opposing 'Affectations' (giving the appearance of different characters) were evaluated with users. It was therefore possible to test which artificial 'character' students preferred. From the evaluation study some conclusions are made concerning the effect of the two opposite characterisations on student perceptions of the agent and the degree to which it was helpful as a learning companion. Some recommendations for future work are then made.