In this paper we propose to include two up-to-date separate concepts, namely social computing and usability metrics, in intelligent interaction agents to enhance a user-centred, adaptive human-computer interaction (HCI) on the Web. Social computing refers to the application of sociological understanding to the design of interactive systems. Usability metrics are software quality metrics with a long history of successful application in software engineering. We introduce preference metrics, which quantify the subjective evaluations and preferences of users, and performance metrics, which measure the actual use of working software, as suggested parameters that enable user interface adaptation. From all terms, a new user-centred and adaptive interaction multi-agent model and architecture is proposed in e-learning/e-teaching on the Web.