The skull beneath the skin: Entity-relationship models of information artifacts

Data modelling reveals the internal structure of an information system, abstracting away from details of the physical representation. We show that entity-relationship modelling, a well-tried example of a data-modelling technique, can be applied to both interactive and noninteractive information artifacts in the domain of HCI. By extending the conventional ER notation slightly (to give ERMIA, Entity-Relationship Modelling for Information Artifacts) it can be used to describe differences between different representations of the same information, differences between user's conceptual models of the same device, and the structure and update requirements of distributed information in a worksystem. It also yields symbolic-level estimates of Card, Pirolli and Mackinlay's index of ''cost-of-knowledge'' in an information structure, plus a novel index, the ''cost-of-update''; these symbolic estimates offer a useful complement to the highly detailed analyses of time costs obtainable from GOMS-like models. We conclude that, as a cheap, coarse-grained, and easy-to-learn modelling technique, ERMIA. usefully fills a gap in the range of available HCI analysis techniques. (C) 1996 Academic Press Limited