La construction des échantillons dans la conception ergonomique de produits logiciels pour le grand public. Quel quantitatif pour les études qualitatives?

Abstract

The aim of this article is to facilitate the participation of direct users in the different steps of the design process, namely the investigation of needs, specifications and evaluations (of models, prototypes, and final versions). Such a study needs to limit the size of samples and take into account the variety of situations through a small number of tests. According to Nielsen, 80% of problems can be identified through an evaluation of six people if the test deals with one profile and by three or four people per group if the test deals with several groups. The design of software assistance for public transport users is used as an example for discussion. The study highlights that the analysis is concerned with more than just problems and incidents; it also deals with the evolution of strategies, and provides positive feedback. Moreover, the number of required testers does not only depend on the descriptive criteria of the population. It also depends on the types of products, the number of functions and the context of investigations: dynamic and realistic situations allow one to situate the action better and to output more results than static presentations or simulations. Thus, the idea of a sample is to be extended from the population towards the various aspects of the situations (such as contexts, tasks and tools). The test scenarios have to include each corresponding combination. For example, if N participants test T tasks within C different contexts, one gets N×T×C test situations. Thus, only a few people are required to test a vast number of situations and obtain a great deal of representative and relevant information (such as needs and evaluations).