This post gives an overview of our approach to this area. In brief, we’re developing a framework to guide systematic mapping of knowledge features onto corresponding representation types.
The diagram below shows an example from our work. It treats liking and disliking as two separate dimensions, rather than as opposites, and then uses a corresponding graphical representation.
We’ve drawn heavily on knowledge representation as practised in Artificial Intelligence, which we’ve found invaluable for core formalisms. Two concepts that we use often are net representations and hierarchical levels (for instance, using hierarchical levels to represent emergent properties at each level).
We also make heavy use of graph theory and facet theory. We usually treat facet theory as a variant of graph theory, rather than in the traditional information science sense. Treating it in this way provides more internal consistency in representation, and also resolves various theoretical issues about the semantic status of facets in the information science sense.
We’ve done various work on mapping set theory and measurement theory systematically onto visualisation; for instance, by showing statistical probability via the density of shading in a diagram.
We’ve also factored in the issue of probabilistic versus frequentist descriptions, with regard to human factors. This is explicitly inspired by Gigerenzer’s work on presenting statistical information via visual representations that play to human strengths in parallel processing, pattern matching and subitising, versus human weaknesses in serial processing and calculation.