Augmenter la validité des évaluations des applications graphiques interactives. Combiner les approches prédictives et empiriques

While human-computer interaction (HCI) researchers invent a lot of interaction techniques that could improve usability of graphical applications, most of our interfaces are still based on the WIMPmodel (Windows, Icons, Menus and Pointing). The difficulty of assessing the performance of an interaction technique in a real context of use can partly explain this gap. This article aims at improving the evaluation process used in HCI by following a methodology where predictive and empirical aspects are complementary to augment the external validity of lab experiments. This approach is supported by two tools: the Complexity of Interaction Sequences (CIS) model and the TouchStone platform. This evaluation methodology defines a design space with performance measures to allow interface developers to make informed choices. MOTS-CLÉS : interaction, évaluation, expérimentation, contexte d’utilisation, validité.