2. Background: Actor-Network Theory

2. Background: Actor-Network Theory

This section gives the parts of ANT the framework depends on. Appendix A gives the full account: founders, roots, origins, later developments, uses and criticisms, with sources.

2.1 Origins

ANT was developed at the Centre de Sociologie de l'Innovation in Paris from about 1980, mainly by Michel Callon, Bruno Latour and John Law, with Madeleine Akrich as a fourth architect. Steve Woolgar co-wrote the laboratory study in which the approach began (Latour and Woolgar 1979). It drew on the sociology of scientific knowledge, ethnomethodology, the semiotics of A. J. Greimas, and Michel Serres's concept of translation.

The approach appeared as a set of statements in 1986. Callon's study of scallop restocking in St Brieuc Bay set out three principles (agnosticism, generalised symmetry and free association) and four moments of translation (problematisation, interessement, enrolment and mobilisation). The network he described failed: the scallops did not anchor and the fishermen took the stock. Several founding cases are about networks that failed, or had to be held together against forces pulling them apart. Between 1987 and 1992 the vocabulary this paper uses became established, including in a glossary by Akrich and Latour (1992).

2.2 Key ideas

ANT is often summarised as "networks of human and non-human actors". That summary misses the ideas that make it useful for engineering.

• Actant and actor. An actant is whatever makes a difference to a course of action, without assuming in advance whether it is a person. An actor is an actant given an identity in an account. In an agentic system, many things act on an answer before anyone names them: a default date, a status filter, a client that drops a channel.
• Mediator and intermediary. An intermediary carries meaning unchanged. A mediator transforms it (Latour 2005). The framework assumes every hop is a mediator until shown otherwise.
• Translation. Turning one actor's terms into another's always changes something, and can betray the original (Callon 1986). The framework calls such a change a loss.
• Inscription and script. Designers write a scenario of use into an object; de-scription reads it back from the object in use (Akrich 1992). A tool's name and description are inscriptions that script what a model will do.
• Obligatory passage point. A point made indispensable through problematisation, which other actors are persuaded or compelled to pass through (Callon 1986). Being technically central does not make a server one.
• Relational ontology and performativity. Actors get their capacities from their relations, and descriptions help bring about what they describe (Callon 1998). A tool cannot be understood from its code alone, and its description is tested for what it produces.
• Power, durability and irreversibility. Power is the effect of relations assembled and held in line (Callon and Latour 1981; Law 1986). Arrangements last when delegated to materials (Latour 1991), and translations become hard to reverse as others come to depend on them (Callon 1991). A rule computed in a server lasts in a way a prompt does not.
• Immutable mobile. An inscription stable enough across distance and context to be compared and combined (Latour 1986). Latour (2013) later stressed that there is no transport without transformation.

2.3 Method or theory

ANT's authors offered principles and rules of method, not a procedure. Latour's rules include following the actors and studying things in the making, before they become black boxes (Latour 1987). In 1999 he argued that ANT is better understood as a method for following how actors build their worlds than as a social theory. That is one position: others argue ANT carries commitments of its own (Whittle and Spicer 2008). For this paper the point is narrower. Since ANT's founders never offered a step-by-step procedure, the framework's procedure is an adaptation, and it must be judged by its cases.

2.4 How ANT has been used

ANT has been used in three modes:

• Explanation after the event. How a fact, technology or organisation came to hold together or fail. This is ANT's main use, in science and technology studies and, from the mid-1990s, in information systems (IS) research (Walsham 1997).
• Intervention. ANT ideas used inside a change while it happens, as in action research (Lewis and Townson 2004; Braa, Monteiro and Sahay 2004).
• Design. ANT used to inform designers or organise design, as in the analysis of standards (Hanseth and Monteiro 1997), the flexible-standards strategy (Braa and others 2007) and participatory design (Binder and others 2011).

Most published work is in the first mode. Section 3 considers how close the other two come to a method.

2.5 Criticisms

ANT has drawn sustained criticism. Extending symmetry to non-humans, critics argued, removes the ground for sociological explanation (Collins and Yearley 1992). Network accounts are told from the side of those who build networks, and miss those who do not fit them (Star 1991). Constructivist studies of technology are indifferent to whether a technology does good or harm (Winner 1993). ANT struggles to ground critique (Whittle and Spicer 2008), and does not address structures and inequalities that exist before a network is built (Elder-Vass 2008). Because ANT explains how a claim comes to be accepted without judging whether it is true, it can blur acceptance and correctness. And it gives the analyst great freedom and little procedure, so almost anything can be described in its terms after the event. Section 7 sets out what the framework does in response to each criticism, and what stays open.

2.6 What the framework borrows

The table gives each concept as ANT defines it, as the framework uses it, and the relation between the two. Where the framework's use is narrower or simpler, the paper does not attribute that use to ANT. Appendix A, part 1.11, gives the full mapping of 21 concepts.

The largest change is to the four moments. Callon used them to describe how a network was built and how it failed. The framework uses them as stages to work through. That is a change of purpose, so the framework's claims rest on its cases, not on ANT's authority.