The importance of emotions in old age is evident as they are closely linked to adaptation processes, health status and cognition. Particularly related to this paper is the broadly studied relationship between negative emotions and negative health outcomes. In this sense, real-time physiological emotion detection systems can assist in developing solutions for minimising the effects of negative emotions on elderly people's health. However, the intrusiveness of continuous physical monitoring is an element that might produce negative reactions from the user and therefore reduce the acceptability of such systems. In this paper we present the results of a scenario-driven user-centred analysis involving a group of elder people in order to evaluate their response to emotion recognition-based assistive technology. The main conclusion of this study is that emotion detection systems appear to be of great benefit when they offer user-supervised ways to increase human support in situations involving strong affective negativity, as opposed to fully-automated computer interaction.