In this article, I assess an existing language acquisition architecture, which was deployed in linguistically unconstrained human–robot interaction, together with experimental design decisions with regard to their enactivist credentials. Despite initial scepticism with respect to enactivism’s applicability to the social domain, the introduction of the notion of participatory sense-making in the more recent enactive literature extends the framework’s reach to encompass this domain. With some exceptions, both our architecture and form of experimentation appear to be largely compatible with enactivist tenets. (...) I analyse the architecture and design decisions along the five enactivist core themes of autonomy, embodiment, emergence, sense-making, and experience, and discuss the role of affect due to its central role within our acquisition experiments. In conclusion, I join some enactivists in demanding that interaction is taken seriously as an irreducible and independent subject of scientific investigation, and go further by hypothesising its potential value to machine learning. (shrink)