Abstract : This thesis examines the underlying sociotechnical dynamics in the development of an online chat device via Internet –Internet Relay Chat (IRC). It argues that this device is constantly co-constructed by a set of human and non-human actants organized in sociotechnical networks called ‘IRC networks', and that there is no clear distinction between users and designers of the system. The research seeks to understand the modalities of, and the motives for this co-construction. The problematics stands at the intersection of two research fields: the sociology of innovation –in the context of science and technology studies (STS), and computer-mediated communication (CMC) studies. It draws on recent theoretical developments suggesting that the role of users in the construction of technical systems has been underrated, particularly in the case of digital artifacts, which seem to offer more ‘plasticity' in response to usage. While the notion of a user being not only an actor of, but also a contributor to technical devices is becoming widely accepted today, it is important to note that the nature of this contribution is generally limited to the contents. The case of IRC is interesting in that it shows how users can contribute to the design of the device itself. The research relies on a conceptual framework based on three theoretical approaches in STS: social construction of technology, actor-network theory and the “ecological” model of social worlds. These three approaches are combined to provide elements of a co-construction theory which, compared with classical innovation models, redistributes agency among the actors of the development of a sociotechnical device. Thus, even the roles of user and designer appear to be co-constructed in the process. The concepts of community of practice and “arena of technical skillfulness” are called upon in the analysis of the actors' social motives for their engagement in this co-construction. At the methodological level, the research is a case study focused on the genesis and the evolution of two major IRC networks: EFnet and Undernet. The specific object of study is a series of controversies that arose between 1990 and 2001 and which led to qualitative leaps in the evolution of these networks. The main methods mobilized are discourse analysis and online ethnography, combined with inquiry techniques used in STS for the study of sociotechnical controversies. Concretely, the inquiry protocol consisted of three components: a) online observation; b) content and discourse analysis of a corpus of documents available online and the archives of four mailing lists and two Usenet newsgroups; c) synchronous and asynchronous online interviews with twelve key actors of IRC development. Identifying “critical moments” in IRC development helped find controversies and structuring events. Their analysis led to isolate the notion of service as pivotal to understand the modalities of co-construction of the device. First understood in its usual sense (related to the notion of user as customer), services were gradually ‘translated' into increasingly elaborate programs, which almost became black boxes: official ‘robots' (or bots), also known as IRC services. On EFnet, this process has long been stifled by administrators. But users have developed their own strategies to fill the gaps of the original technical protocol, notably by creating their own bots in order to protect their channels. By contrast, promoters of the Undernet have ensured that their network would stand apart from EFnet by putting service to users and user involvement in the device at the core of their project. On the Undernet, the concept of channel service is a hybrid one: it is a service both in the organizational sense, and in the technical sense of ‘bot'. Such definitional ‘fuzziness' of the notion of service in IRC actors' discourse is an indication of its status as a boundary object around which different ‘philosophies' of online chat are articulated. These views sometimes diverge to the point of creating technical boundaries between IRC networks, and therefore, between different communities of chat practice.