Methods to assess and ensure system usability are becoming increasingly important as market edge becomes less dependent on function and more dependent on ease of use, and as recognition increases that a user's failure to understand how an automated system works may jeapordise its safety. While ultimately only deployment of a system will proveits usability, a number of approaches to early analysis have been proposed that provide some ability to predict the usability and human-error proneness of the fielded system. The majority of these approaches are designed to be used by human factors specialists, require specific expertise that does not fall within the domain of software engineering and fall outside standard software development life cycles.

However, amongst this number, some rigorous mathematical methods have been proposed as solutions to the more general problem of ensuring quality of system designs but with limited success. This paper discusses their limitations both in terms of the broader software engineering agenda and in terms of their effectiveness for usability analysis, the opportunities that they offer and discusses what might be done to make them more acceptable and effective. The paper positions those methods that have been effective against less formal usability analysis methods.

| download |