The first term hopefully marks the end of a long struggle for les mots justes ever since Stephen Clark began to develop his Glasgow thesis on Literate Development. In that work, Stephen explored the use of symbolic links between representations of context and representations of design. Keith van Rijsbergen raised an important issue at each of Stephen's reviews (which fortunately took a back seat at his viva!): just what are these links? We promised to find out before Stephen submitted, but as with much research the best intentions didn't quite materialise.
Two years after Steve's graduation, I finally hit on a plausible answer to Keith's insistent query: the links are actually twin strands that form into common representations. Such representations can be grounded in contextual data and are thus Grounded Theories as understood within social science. Such representations can also be computed from design specifications, resulting in a true meeting of equals over the typically deep void between the real world and the designs that aim to fit it. Given representations of the same type and similar enough values, we can imagine a process akin to surface attraction where the tips of strands stick together without extra adhesion or fastening. That's representational attraction for you.
Where representational attraction doesn't happen (which I suspect will be most of the time, and where not, it will be short lived as the world moves on), the strands need to be knotted together in some way. In reality, human adaptation in the real world is what ties together the loose strands between designs and their context of use. The extent of knotting required for a rendezvous of strands depends on the mismatch between tips' representation types and values. In essence, one representation will contain more variety than its intended partners, and thus bring about the classic cybernetic imbalances at the heart of Ashby's Law and Beer's derived notion of Variety Engineering. The extent of fit (and the consequential required human adaptation) can then be measured as the union of all variety imbalances across the mesh of a grounding net.
The resulting big picture for HCI thus becomes the construction of meshes between representations of context and design. Strands at the edges of the mesh are spun (as Grounded Theories) from contextual data or extruded (as functions) from design specifications. Associated strands are then joined up with various degrees of success. This puts the meat in a my meta-method for HCI: SASEJ (pronounced sausage!)
Designs do thus not so much "fit" their contexts of use as "mesh" with them to some degree of success. "Mesh" here is both structure and property: structure as a mesh of joined strands spun and extruded from context and design; property as "meshing" --- of harmony between two entities. In an ideal design, the grounding mesh would be wholly formed from strands that mesh at their tips.
I will present examples of common representations for context and design, demonstrating why tasks and scenarios have had to be so central to HCI to date, but how other representations are actually more suitable for focused assessments of fit to context.
For more information contact: cs0gco@isis.sunderland.ac.uk