Actor-network theory and the more-than-human: artefacts and non-humans act

Sociomaterial and more-than-human accounts insist that the explanation of an interaction must include its non-human participants. A door-closer holds a door a person would otherwise hold; a form's default settles one decision and imposes it on every later user. This ninth of ten units argues that the vocabulary these accounts supply is useful to a designer, and it is the last to give artefacts agency. It makes the question of what a component is doing, and for whom, askable. The vocabulary is bought at two costs: a symmetry permissive enough to justify almost any arrangement, and a method that describes networks without yielding design criteria of its own.

The two costs sit in different literatures: the method stays in the sociology of science, and HCI borrowed only the vocabulary.

The hinge: an artefact can carry an action

Latour (1992) begins from a seat belt he could not disobey. His car would not stay quiet until he buckled it. He describes an engineer linking the ignition to a sensor in the belt, so a rule he had once accepted became a property of the assembly of driver and car. The working example is smaller: a hinged door, and the automated door-closer he calls the groom. Its strike notice stood in for work a person would otherwise do. Latour names this move delegation or displacement, and calls the behaviour imposed back onto the user prescription. A design can then be read as a programme of action, with an anti-programme wherever users resist. The analyst then asks which participant carries which task.

What the method asks, and what it will not give

Latour (2005) gives the method. The social is what has to be assembled and explained, not a substance that explains anything. The analyst therefore traces associations and follows the actors, rather than invoking society or culture as causes. Law (1992) states the same commitments compactly, treating social relations as network effects and insisting that networks are materially heterogeneous. Agents, texts, devices and architectures are therefore analysed in the same terms. Callon (1999) defends the refusal to supply a general theory of the actor, arguing that this absence combined with the role given to non-humans is a strength to be preserved. He describes markets as products of disentanglement, framing, internalisation and externalisation. None of the three offers a criterion for design. The method tells an analyst how to describe a network, not what a good network looks like.

HCI took boundary objects and left the method behind

Star and Griesemer (1989) studied the Museum of Vertebrate Zoology at Berkeley between 1907 and 1939, where amateurs, professionals and administrators had to cooperate despite divergent viewpoints. They found two activities that made cooperation possible: standardising methods, and boundary objects, which are adaptable to different viewpoints while robust enough to keep their identity across them. They distinguish four types: repositories, ideal types, coincident boundaries and standardised forms. The concept answers a design question: what artefact can two communities with different interests use? Orlikowski and Scott (2008) carry the same move into information systems. They report that more than 95 per cent of the articles in four leading management journals take no account of technology. They set a sociomaterial position, in which the technical and the social are constitutively entangled, against two earlier streams.

Infrastructure decides what the visible interface can do

Star (1999) turns the approach on the background. Infrastructure, she argues, is relational and ecological, meaning different things to different groups and inseparable from the balance of action, tools and the built environment. It is also frequently mundane, consisting of plugs, standards and bureaucratic forms. It is hard to study because it is transparent in use and visible mainly when it breaks. The invisible work of maintaining conventions then drops out. Her examples are deliberately mundane. A staircase is infrastructure for one person and a barrier for a wheelchair user or an older person who cannot climb it. One structure serves some and obstructs others. A plaster sold as flesh coloured carries a skin colour that is not everyone's. The International Classification of Diseases fixes what can be written on a death certificate. The arguments that settled its categories vanish once the list is printed. Adding a single extra button to a workflow can defeat a whole group of users, which is the paradox in miniature. An interface is the visible part of a system whose invisible part decides what the visible part can do.

The network in circulation describes those already central

Star (1990) examines the model of heterogeneity set out by Latour and Callon, and reads it as a managerial model of actor-networks. She sets against it alternative accounts, among them the multiple membership and marginality she associates with women of colour. Her corrective is a theory of multiple membership, which follows standardising technologies and people who belong to several social worlds. Such people are, in her phrase, cyborgs, humans with machines. The design implication is that a standard an interface enforces presumes a single social world.

Contested question: is symmetry an analytic device or a moral blank cheque?

The disagreement is not whether artefacts matter, which every source here concedes. It is what the symmetry between human and non-human agency is for. Latour (1992) and Callon (1999) treat it as the device that makes non-humans visible at all. Callon argues that the absence of a theory of the actor is a strength to be preserved. Latour asks where the morality in his seat-belt case sits, in the driver or the machine, and declines to choose.

Star (1990) holds that the version in circulation is not neutral. Read as a managerial model, she argues, it describes the networks that already work, and lets members marginal to all of them fall out of the account. Her answer is multiple membership, not a stronger symmetry. Puig de la Bellacasa (2011) presses from the same direction, arguing that Latour's matters of concern lack a critical edge. She proposes matters of care, which require a speculative commitment to neglected things, so the description is partial by design rather than by accident.

What would settle it is a use of the vocabulary that can rank rather than report. Describe two designs for one task in actor-network terms, one that leaves the people affected able to act and one that takes that ability away. See whether the symmetrical description can say which is better without an undeclared norm. Latour's (1992) own case shows the difficulty, since it can say that driver and car obey the law but not whether that is an improvement.

Both sides of the dispute argue from concepts and cases rather than from tests. Latour (1992) and Callon (1999) reason from one seat belt and one market, and Star (1990) and Puig de la Bellacasa (2011) answer with arguments of the same kind. The only counting evidence in the unit is Orlikowski and Scott's (2008) survey of four journals. While the dispute is open, a designer should describe the arrangement symmetrically first. The description should then state the norm by which one arrangement is judged better, and record where that norm came from. The strongest objection is that the unit's second cost may be a demand the theory never accepted. Mol's (2002) question, which practice's object an interface enacts, is already a design criterion.

The more-than-human turn: more vocabulary, still no criterion

Coulton and Lindley (2019) argue that design contexts now hold complex interdependencies of human and non-human actants. Their example is the Internet of Things, where things transcend their physical forms and include algorithms, humans, data and business models, each importing independent but interdependent motivations. Their contributions are a review of HCI work beyond the anthropocentric perspective, a constellations metaphor, and a speculative design. A connected thing is no use offline. Nicenboim and colleagues (2023) trace decentring from HCI back to critical posthumanism, and distil five dimensions for more-than-human practice. They conclude that decentring is a material practice through which designers also make posthuman knowledge. Giaccardi, Redstrom and Nicenboim (2024) introduce a special issue on the same terrain and frame the open questions about the turn's scope. The programme is stated; the criterion is not.

Enacted objects: which practice does the interface carry?

Mol (2002) supplies the sharpest design vocabulary in the unit. Her fieldwork in a Dutch university hospital follows the diagnosis and treatment of atherosclerosis, and finds the condition appearing as many different things from one moment, place, apparatus or speciality to the next. The multiplicity is not fragmentation, she argues. The disease is made to cohere through tactics such as transporting forms and files and holding case conferences. Read as design theory, that says an object is enacted in practices. The question is then which practice's object an interface enacts, and which others it overrides. Frauenberger (2019) sets the wider frame, arguing that entanglement with digital technologies is challenging HCI's foundations. He reviews actor-network theory alongside post-phenomenology, object-oriented ontology and agential realism, and proposes four perspectives for Entanglement HCI. Actor-network theory is one of the four resources he reviews, and the wave it belongs to has no agreed periodisation.

Where the argument is heading

Nicenboim and colleagues (2023) make decentring a material practice rather than a stance. Giaccardi, Redstrom and Nicenboim (2024) name the trajectories they expect, among them redefining data, interfaces and responsibility. Frauenberger (2019) puts the tracing of accountabilities inside the research programme itself. Puig de la Bellacasa's (2011) matters of care is the live candidate for the design criterion the approach lacks. It is not yet a method. Whether the turn generates criteria of its own, or keeps supplying a vocabulary others must convert, is what the coming years will answer.

The sources this unit rests on

• Callon, M. (1999). Actor-Network Theory: The Market Test. The Sociological Review 47(1_suppl). doi:10.1111/j.1467-954X.1999.tb03488.x Defends the refusal of a general theory of the actor and reads markets as products of disentanglement, framing, internalisation and externalisation. The defence of symmetry the contested question turns on.
• Coulton, P. and Lindley, J. G. (2019). More-Than Human Centred Design: Considering Other Things. The Design Journal 22(4), 463-481. doi:10.1080/14606925.2019.1614320 Argues that design contexts hold complex interdependencies of human and non-human actants, using the Internet of Things, and adds a constellations metaphor and a speculative design. The more-than-human turn stated for designers.
• Frauenberger, C. (2019). Entanglement HCI The Next Wave? ACM Transactions on Computer-Human Interaction 27(1), 1-27. doi:10.1145/3364998 Argues that entanglement with digital technologies is challenging HCI's foundations, reviews actor-network theory alongside three other relational theories, and proposes four perspectives for Entanglement HCI. It places the approach inside an unsettled periodisation.
• Giaccardi, E., Redstrom, J. and Nicenboim, I. (2024). The making(s) of more-than-human design: introduction to the special issue on more-than-human design and HCI. Human-Computer Interaction 40(1-4), 1-16. doi:10.1080/07370024.2024.2353357 Introduces the special issue, arguing that human activity has altered the planet, with design playing a part, and that human-centred approaches may not meet the challenges ahead. The field's account of an unsettled programme.
• Latour, B. (1992). Where are the missing masses? The sociology of a few mundane artefacts. In Bijker and Law (eds), Shaping Technology/Building Society, MIT Press. https://www.bruno-latour.fr/node/258.html The unit's hinge: a door-closer and a seat-belt interlock read as delegated action, with prescription as the behaviour an artefact imposes on its user. The vocabulary, and a case that resists ranking designs.
• Latour, B. (2005). Reassembling the Social: An Introduction to Actor-Network-Theory. Oxford University Press. doi:10.1093/oso/9780199256044.001.0001 Argues that the social is what has to be assembled and explained rather than a substance that explains, and sets out tracing associations as the method. A procedure that yields description rather than design criteria.
• Law, J. (1992). Notes on the theory of the actor-network: Ordering, strategy, and heterogeneity. Systems Practice 5(4), 379-393. doi:10.1007/bf01059830 Treats social relations as network effects and insists that networks are materially heterogeneous, so agents, texts, devices and architectures are analysed in the same terms. The compact statement of that method.
• Mol, A. (2002). The Body Multiple: Ontology in Medical Practice. Duke University Press. doi:10.1215/9780822384151 Follows atherosclerosis through a Dutch university hospital and argues that the condition is enacted differently in different practices, made to cohere by tactics such as transporting forms and files and holding case conferences. The unit's sharpest design question, which practice an interface carries.
• Nicenboim, I., Oogjes, D., Biggs, H. and Nam, S. (2023). Decentering Through Design: Bridging Posthuman Theory with More-than-Human Design Practices. Human-Computer Interaction 40(1-4). doi:10.1080/07370024.2023.2283535 Traces decentring from HCI to critical posthumanism, distils five dimensions for more-than-human practice, and concludes that decentring is a material practice that also makes posthuman knowledge. The most concrete working-through of theory into design.
• Orlikowski, W. J. and Scott, S. V. (2008). Sociomateriality. Academy of Management Annals 2(1). doi:10.5465/19416520802211644 Reports that more than 95 per cent of the articles in four leading management journals take no account of technology, and sets a sociomaterial position, in which the technical and the social are constitutively entangled, against two earlier streams. The approach travelling into information systems without its method.
• Puig de la Bellacasa, M. (2011). Matters of care in technoscience: Assembling neglected things. Social Studies of Science 41(1), 85-106. doi:10.1177/0306312710380301 Argues that Latour's matters of concern lack a critical edge and proposes treating facts and socio-technical assemblages as matters of care, which demand a speculative commitment to neglected things. The unit's alternative to symmetry as neutral method.
• Star, S. L. and Griesemer, J. R. (1989). Institutional Ecology, Translations and Boundary Objects: Amateurs and Professionals in Berkeley's Museum of Vertebrate Zoology, 1907-39. Social Studies of Science 19(3), 387-420. doi:10.1177/030631289019003001 Shows how scientific work needing both divergent viewpoints and generalisable findings was carried by standardised methods and boundary objects, in four types. The concept HCI borrowed most, and evidence these ideas travel.
• Star, S. L. (1990). Power, Technology and the Phenomenology of Conventions: On being Allergic to Onions. The Sociological Review 38(1_suppl), 26-56. doi:10.1111/j.1467-954x.1990.tb03347.x Reads the actor-network model of heterogeneity as a managerial or entrepreneurial one and answers it with a theory of multiple membership drawn from marginality. Evidence that the approach can carry power.
• Star, S. L. (1999). The Ethnography of Infrastructure. doi:10.1177/00027649921955326 Sets out methodological questions for studying infrastructure with ethnographic tools, finding it relational, ecological and mundane, and names the paradox that it is transparent in use and visible mainly when it breaks. It grounds the claim that the background decides what the interface can do.