A system called ADAPTS (adaptive diagnostic and personalized technical support), illustrated in the context of technical manual data for a US Navy H-60 helicopter, is presented in this paper. ADAPTS demonstrates the use of interrelated and nontrivial user, domain, and task models to achieve tailored support for the performance of technicians at a variety of levels of skill and training.

In the illustrated case, the domain model represents the components of the helicopter; the task model represents technician activities the manual can support (for example, replacement of units); and the user model represents some 20 characteristics of the user, with respect to competency, confidence, and experience for various user actions. The user model is seeded with stereotypes for different technician backgrounds and is used to tailor information display. Tailoring includes default position (expanded or collapsed) of checklist items and instructional text passages, as well as tailored hyperlinks with cues for individual relevance of those links.

It is as exciting as it is rare to see an approach to devising adaptive systems that utilizes truly nontrivial modeling of the user, domain, and task. Through this, relevant adaptation is achieved. The eclectic set of 28 references is well distributed through the 1990s and earlier, and largely avoids major academic journals. The choice of references illustrates that serious model-driven adaptive human computer interaction has remained largely hidden, emerging as somewhat more mainstream only for the problem of fitting content to small-screen devices.

The authors have done a great deal of work to demonstrate how models can be used to adapt the user interface without being disruptive to the user (as in the case of dynamically chosen hyperlinks and color coding). It would still be interesting to explore further possibilities for more dramatic adaptation. Also, there is the outstanding question of whether this adaptation actually improves performance.