This is an archive of the original site and you may encounter broken links and/or functionality

0265 A latent semantic analysis-based service for providing personalised formative feedback on conceptual development within PBL Groups


13:00 - 14:00 on Wednesday, 8 September 2010 in Pos


0265 A latent semantic analysis-based service for providing personalised formative feedback on conceptual development within PBL Groups
Fridolin Wild, Alisdair Smithies, Gillian Armitt, Isobel Braidman


0265 A latent semantic analysis-based service for providing personalised formative feedback on conceptual development within PBL Groups
Fridolin Wild, Alisdair Smithies, Gillian Armitt, Isobel Braidman
Constructivist approaches stress the importance of collaborative knowledge co-construction, for example in problem-based learning (PBL) groups (Wood 2003). PBL learners find it difficult to identify the limitations of their own understanding and topic coverage, and can benefit from personalised formative feedback (Shute 2008). Our experience indicates that tutors find it difficult to determine learners' individual conceptual understanding in group situations to inform personalised feedback. Latent semantic analysis (LSA) uses statistical computations to analyse textual relationships. The LSA-based CONSPECT service visualises the conceptual understanding of individual learners, enabling 'on demand' probing and formative feedback on their conceptual development. The theoretical framework for the use of the service within PBL groups is Stahl’s model of knowledge building (Stahl 2006), which integrates personal and collaborative “knowing”. Through the EU-funded LTfLL project, a bespoke on-line service was developed based on LSA. Learners' texts on medical topics are compared semantically with the PubMed database as background corpus. Force direction, an optimised graph layout algorithm (Fruchterman and Reingold 1991), underpins a two dimensional visualisation of the LSA output (conceptogram), demonstrating the learner's conceptual understanding. This poster illustrates the conceptual monitoring information seen by the user, how data input is processed to achieve this, and validation results. A weblink to a short video of the software is provided. Learners can input texts on medical topics and view the resulting conceptogram. They can compare their own conceptograms with those of other (anonymised) users to identify and then improve upon areas of weakness. Learners can release their conceptograms to peers or tutors for further feedback. The service was piloted in a School of Medicine. Validation results, based on focus groups, interviews and a questionnaire, demonstrate the educational value of the software and the implications for managers of introducing this service. CONSPECT can facilitate more effective learning through identifying shortcomings in individuals' conceptual understanding and topic coverage, and can improve tutor effectiveness by informing personalised feedback in group contexts. Future developments include a 'teacher console' identifying learners in need of special attention and pinpointing topics that are generally not well understood.