Hypermedia Discourse: Semantic Structures for Sensemaking

Simon Buckingham Shum, Knowledge Media Institute, The Open University, http://kmi.open.ac.uk/sbs

Background
This workshop asks what role semantics structures have to play in teaching and learning. Since 1995, the Hypermedia Discourse research programme at the Open University’s Knowledge Media Institute has been theorising, developing and evaluating tools for learners, educators, researchers and professional analysts to model and visualize the structure of problems, discussions in meetings, and debates: http://kmi.open.ac.uk/projects/hyperdiscourse

Challenge 1: How to model contested domains?
The semantic web formalises consensus: that is what an ontology is. Where there is consensus, this is useful. When teaching a domain whose structure is well understood, where effective learning strategies can be codified, and where user behaviour can be classified, knowledge-based, adaptive applications clearly have a role to play. There are impressive results showing that intelligent tutoring environments can outperform human teachers in certain respects (eg. learning algebra).

Other modes of learning are, however, more exploratory. Even if the domain is well understood and there is one correct answer, for pedagogical reasons, one wants the students to articulate their own understanding. Particularly in the absence of consensus or a single correct answer, the focus of semantics has to shift to scaffolding the deliberations that stakeholders must engage in, as they seek to articulate and debate the significance of information elements. These may be very different depending on perspective.

Our work is focused on this part of the design space, developing tools to craft semantic structures that scaffold flexible information management, and the moves in dialogue and argumentation. There need be no consensus about the domain’s structure. But there is enough consensus on how to make contributions and connections between ideas – and if necessary express disagreements – in principled ways about the domain in question, using the visual languages that we provide.

We therefore see semantic structure, when visualized, as a way for representations to provoke reflection, as analysts engage in personal or collective sensemaking (after Weick) around wicked problems (after Rittel).

Challenge 2: Fostering fluency with semantic structures?
Central to our work is the challenge of negotiating the formalisation gulf: the representational gap between how we think normally, and the structure required by a semantic tool. We want to develop intellectual tools for structuring information that are usable without having to be an ontology engineer or information scientist.

We see such tools as new ways of reading and writing ideas: a new literacy. Through extensive field deployment of research prototypes, coupled with qualitative video analysis in experimental and field data, we are developing an account of the nature of fluency with such tools. This starts from the cognitive tasks involved in learning to use them as beginners, through to expert, almost artistic performance and improvisation with them in real time, highly pressured contexts such as NASA missions, or personnel rescue operations.

Semantic Tools: Example 1: Compendium
Our most mature platform is Compendium: http://compendium.open.ac.uk. This supports real time knowledge construction in meetings, or can be used for personal information management and reflection. This is a robust hypermedia mapping tool that is freely available, with a growing community of practice, and an annual workshop. It is now integrated with the Moodle virtual learning environment as part of the OU’s OpenLearn initiative, publishing OU resources for free on the web: http://www.open.ac.uk/openlearn. Guides and screencasts introducing it in this context can be found at: http://openlearn.open.ac.uk/course/view.php?id=2824 and http://compendium.open.ac.uk/openlearn/screencasts.html

Numerous examples of its applications in learning are now published including mapping the Iraq Debate, course assignments, history of psychology: http://openlearn.open.ac.uk/course/view.php?id=2824#section-4 and http://openlearn.open.ac.uk/blocks/compendium/browsemaps.php?context=22&id=2824

In the JISC Memetic project, Compendium was integrated with the Access Grid, providing an environment for semantically tagging replayable videoconferences: http://www.memetic-vre.net. This work is now being extended for work with choreographers in the AHRC-EPSRC-JISC e-Dance project: http://ahessc.ac.uk/e-dance.

Semantic Tools: Example 2: Cohere
Cohere http://cohereweb.net is a visual environment for making meaningful connections between ideas, and optionally tagging those ideas with websites. As a Web 2.0 application it provides a highly interactive user interface enabling users to search, browse and visualize webs of ideas, problems, solutions and arguments, creating their own idea types, and their own connection language. They can import their own bookmark feeds to start weaving connections between existing tags (e.g. from del.icio.us, connotea, citeulike or moodle). Maps can be addressed as URLs, tracked via RSS, and embedded in other websites.

Cohere is an example of a Pragmatic Web application http://www.pragmaticweb.info. In contrast to the semantic web (in which entities have fixed attributes and roles), the pragmatic web orientation is to see meaning as fundamentally context dependent, negotiated by stakeholders, drawing inspiration from semiotics (meaning is both denotative and connotative) and speech act theory (we act when we speak). Thus, in Cohere, Ideas can play different roles depending on the connection to other Ideas, and an Idea’s “meaning” is derived from its referents (websites and connections). As a Web 2.0 application, the ontology of Idea roles and connection types is not closed, but can be taken in whatever direction users wish (as in folksonomies). Yet the web of connections is not a spaghetti of undifferentiated ‘go to’ links, as we currently have on the web or in citation databases, but has semantics that can be leveraged by the software to help users filter views.

Cohere is currently in alpha testing (contact me for an account), shortly to be released as a beta. It is based on the prototypes developed in the EPSRC Scholarly Ontologies project (2001-04): http://kmi.open.ac.uk/projects/scholonto

These tools will be demonstrated live at the workshop.