Measuring Interaction Design before Building the System: a Model-Based Approach

Early prototyping of user interfaces is an established good practice
in interactive system development. However, prototypes cover
only some usage scenarios, and questions dealing with number of
required steps, possible interaction paths or impact of possible user
errors can be answered only for the specific scenarios and only after
tedious manual inspection.
We present a tool (MIGTool) that transforms models of the behavior
of a user interface into a graph, upon which usage scenarios
can be easily specified, and used by MIGTool to compute possible
interaction paths. Metrics based on possible paths, with or without
user navigation errors, can then be computed. For example, when
analyzing four mail applications, we show that Gmail has 3 times
more shortest routes, has twice more routes that include a single
user error, has routes with 13% fewer steps, but has also optimal
routes with the smallest probability to be chosen.
Without MIGTool, this kind of analysis could only be done after
building some prototype of the system, and then only for specific
scenarios by manually tracing user actions and relative changes to
the screens. With MIGTool the exploration of suitability of a design
with respect to different scenarios, or comparison of different
design alternatives against a single scenario, can be done with just
a partial specification of the user interface behavior.
This is made possible by the ability to associate scenarios steps
to required user actions as defined in the model, by an efficient
strategy to identify complete execution traces that users can follow,
and by computing a range of diverse metrics on these results.