Like the Windmills in your Mind

506px-DonquixoteA Discussion of Agency

Due largely to the influence of Actor Network Theory, a lot has been written in recent years about agency. Much of which I find that I disagree with (Graves-Brown 2013a). At one level, the root of these discussions lies in the neglect of material things in anthropology, psychology and sociology, the response to which dates back, at least, to Peter Ucko’s article on penis sheaths (1969, see also Olsen 2010, Graves-Brown 2013a). Yet as in most dialectical processes (and this one might be seen as a variant of the mind:body debate), the effort to rehabilitate the importance of physical things in our world has led to a lurch in another unhelpful direction, in this case towards a kind of metaphysical animism where inanimate objects are invested with agency.

Origination

To state my own position from the outset, I would argue that an entity (be it human, non-human, living or non-living) is only an agent if it has the ability to originate action. Moreover, this means that an agent must also be responsible, causally and (potentially) morally, for its actions.

ANT

In reading the literature on Actor Network Theory, a number of ambiguities arise, which perhaps need to be flagged up early on. Firstly its clear that the translation “actor” is ambiguous (there is a French saying traduction c’est trahison), Latour often talks of actants, and its not clear if this is equivalent to the english language concept of an actor as a causal agent. This is not helped by the sleight of hand by which Latour critiques the opposition between humans and non-humans. In so doing he manages to lump all non-human organisms with inanimate objects together. Thereby equating rabbits with roller skates, he bolsters his claim for the falsity of the human/non-human dichotomy but elides the difference between living and non-living.

But abolishing false categories simply by dispensing with categorisation doesn’t really help. Yes our world is made up of a network of entities, all of which play a part in what happens, but for this to be the case, we need not assume that all entities are qualitatively the same.

Ghosts

Effectively, any discussion of agency is haunted by two ghosts, that remain in the wings of the debate, but hardly ever appear on the stage. First is, as noted above, the persistent mind:body dualism that we get from Descartes, which sets the mind above the physical world of both the body and, by implication, of things. Here the irony (or is it a paradox?) is that whilst attempting to eliminate the dualism of humans and non-humans, the result often appears to be that minds are extended out into the inanimate world (e.g. Malafouris 2013 and Graves-Brown forthcoming). Moreover, to be an agent implies that, in some sense, an entity has free will, that in any situation it could, as the philosophers say, have done otherwise. But then we are stuck with all those nasty questions about how a thing gets free will, a problem that is generally exacerbated by the reductionist and deterministic mindset that pervades most of science.

As I said at the beginning, it often seems as if, haunted by these two ghosts, we resort to a kind of vague animism where agency pervades reality like the aether. I think there is another way.

Causation

To understand agency, then, I think we need to start with causation, and here I suggest that there are at least three alternatives. The first is determinism, where everything that happens and all that we do (or think we do) is determined by molecules, atoms and sub-atomic particles. The more Newtonian clockwork version of this is somewhat complicated by quantum physics, which implies that more than one state can exist at the same time and hence that Schroedinger’s cat can be both alive and dead. One way out of this is a multiverse model in which the possible outcomes of each event give rise to a separate universe. But note that, as the philosophers have pointed out, indeterminacy doesn’t help – its just the obverse of determination. What follows from the determinist, and usually reductionist, account of causation is that consciousness and our belief in free will is an epiphenomenon – we think that we have free will but this is merely an illusion.

Perhaps the best answer to this is that determinists conflate free will with both the causeless cause, and an inflated sense of what free will actually means. To take the second first, as Dennett (1984) points out, to claim that we have free will is not to claim that we can fly simply by flapping our arms. Which leads back to the first, in that the claim for free will is not independence from processes of physical causation, but rather of the emergence of the capacity to originate action within the limits of physical constraint.

But I am getting ahead of myself. The second solution to the problem of causation is metaphysics. That is to say that there are things, spirits, gods, celebrities or whatever, that exist outside of the physical rules of the universe. Strangely, this is kind of what is implied by Cartesian mind:body dualism (with the mind somehow outside the physical universe), even though Cartesianism isn’t supposed to be metaphysical. One way round this is to follow Teilhard de Chardin in suggesting that there is a “Noosphere” – that thought is a kind of physical entity like electrons or Higgs bosons.

Personally, never having encountered anything convincingly metaphysical, I can’t buy into this. But you might think my own explanation, to which I now turn, is a bit weird in its own right. I would argue that free will, and by implication agency, is a product of emergence, a concept that dates back at least to Aristotle. Reductionism implies that all processes and entities can be explained by being decomposed into their component parts. Emergence suggests, quite literally, that the whole is greater than the sum of its parts. Or perhaps just different. If we take, as an example, one of the the simplest molecules in the universe, H2O, it is composed (under terrestrial conditions) of two highly volatile gases Hydrogen and Oxygen, but is itself a relatively inert chemical. Because of its bipolar structure, the water molecule is a highly effective solvent, and en masse has some interesting properties both in its liquid phase and sold phase (snowflakes). The properties of water cannot be reduced to those of either Hydrogen or Oxygen – they are the product, and only the product, of the combination of two atoms of Hydrogen and one of Oxygen (Hydrogen Peroxide H2O2 behaves quite differently). If we then consider the enormous number of complexifications that emerge in organic chemistry, proteins, DNA, cells, nervous systems, organisms and societies, it clear that any number of properties can emerge as the product of such systems. A notion that was recognised long ago by von Bertalanffy, one of the founders of systems theory who regarded living things as “open” systems.

To phrase my earlier claim, then, an agent is an entity that, in virtue of its emergent properties, is capable of originatingaction.

Cruise Missiles and Keys.

berlinWhat does this mean in practice? Latour’s (2000) Berliner key is designed so that apartment dwellers cannot remove their door key without relocking the door. But the key cannot chose not to admit someone if they are drunk, or wearing the wrong shoes or not wearing a tie, as could a human doorman. The key has been made for a fixed purpose and has no capacity to originate action. Moreover, the key, faced with an entity that can originate action, cannot oblige an actor to do anything. A drunk person (or for that matter one wearing the wrong shoes) could leave their key in the door, or just get multiple copies of the key and use a fresh one each time. The key sets up a range of constraints, but it cannot adapt to circumstance because it cannot originate action.

Although Descartes is supposed to have claimed that one could hear the clockwork inside other animals, it is obvious that what constitutes a thing is its capacity to originate action. Hannah Arendt (1958) argued that whilst death is universal, only living things can give birth. Indeed, origination is actually the basis of natural selection. Though creatures derive from the same stock, they a capable of different behaviours, actions, that are differentially successful (the determinist alternative being that every action is inscribed in our DNA). But to my mind, the best counter example to the animist attitude to (inanimate) things, is the fact that in specific cases, non-living systems can originate action. Whist evolution took millions of years to produce genuine actors, human technology is rapidly producing machines that may qualify as such (de Landa 1991; Hellström 2013).

800px-Tomahawk_Block_IV_cruise_missile_-cropA crude example of this is the cruise missile. Here I have to admit that, given that most of the data about Raytheon’s Tomahawk are classified, I can offer only some vague information. Basically, the cruise missile is a sophisticated example of what is termed a “fire and forget it” projectile. Once launched, the missile “does its own stuff”. Whilst it has a fixed objective, which is reached using GPS and terrain following algorithms, the missile is not directly controlled by anyone after launch (see Hellström 2013 for a more general discussion of weapon autonomy). Indeed, it seems that Tomahawks and other such weapons can alter the way in which they reach their target by responding to local conditions in order to evade interception. It follows then that, within some simple parameters, a cruise missile is an agent, in that it can originate action in response to the conditions it encounters. Whilst, presumably, the potential responses are preprogrammed, the actual responses, actions, are chosen to suit a given situation.

Agents versus Media

I would want to argue, then, that the world can be divided into two categories (!); agents and media. It seems to me that the confusion within the animist position is to equate constraint with action. This is compounded by the fact that media do not simply constrain but are, as McLuhan (1964) said “the extensions of Man” [sic]. In fact we might want to say that constraint and extension are necessary counterparts; a bicycle facilitates rapid travel, but constrains my opportunity to read a book.

Material culture allows us to do things we could not otherwise do; planes let us fly where flapping our arms fails, but in a sense here constraint and facilitation converge to mediate action, but do not originate it. Moreover, of course, media exist in a continuum from the “natural” to the artificial; our primary medium is the earth’s atmosphere and, in particular, the 20.95% Oxygen that it contains. Again, here, there has been some sleight of hand accomplished by playing of the falsity of the distinction between the natural and the artificial. A boulder of the right size, and a wooden chair both afford (in Gibson’s 1979 sense) sitting. But to my mind this contributes nothing to our understanding of agency.

In his ecological approach to perception, Gibson (1979) stressed the importance of “invariance” in the environment as an antithesis to the Cartesian view that our surroundings are simply representations in the mind. Invariances specify information such that it need not be represented – it is just there. We do not need to have a mental representation of the 20.95% of Oxygen in the atmosphere in order to breathe. The problem, and one that Gibson failed to address, is that the environment is not static, that a great deal of it has been altered by other entities (people, animals, plants, etc.) and that we as social beings learn to use the invariances around us from other people (Costall 1995).

Canonical Affordances

The fact that not all physical things we encounter are “natural” is what tends, I think, to give credence to the idea of “material agency”. When I go to the supermarket and the door automatically opens, its easy to see the door as an agent in its own right. In a series of articles, Alan Costall (1995; 1997; Costall and Richards 2013) has argued that we need to distinguish the affordance that naturally arises in our environment, be it air, or gravity or boulders we can sit on, from those artefacts that afford in virtue of their design. And, moreover, that such affordances are canonical, in the sense that their affordances have been created with us in mind. Whilst a boulder is the product of geo-physical processes, a chair has been designed and made to be sat on and in general chairs conform to a set of expectations which I would term congruence (see Graves-Brown 1995a and b). Although some chairs do vary from the norm, most are made within a set of socially defined expectations as to what a chair is like. Moreover, from infancy, our interlocutors introduce us to canonical affordances – we learn appropriate and inappropriate ways to sit on a chair.

When it comes to agency, its attribution to chairs or keys is what Ryle (1949) called a category error; the canonical affordances of an object clearly imply intent, they have been originated by someone. But the mistake is to assume that the origination belongs to the chair, rather than the joiner that made it. Moreover, whilst the canonical character of designed affordance serves to define the social world, the absence of agency in inanimate objects means that they cannot, generally speaking, oblige us to act in certain ways – a shoe can be used as a hammer, a cigarette lighter to open beer bottles. In Preston’s (2000) terms, artefacts have “proper” functions; the canonical affordances that their maker intended. But they also have “system” functions, in that they can also afford actions not intended by their makers. Here, in a very clear sense, lies a clash between the power to originate action that resides in both maker and user. This brings me to my next point.

Transaction

Having abandoned the dualism of Descartes, where thought effectively precedes the world around us. And having rejected the extremes of free will, where we can fly by flapping our arms. We are left with an essentially phenomenological view of life in the world. But I want to suggest is that the relationship between agents and media goes beyond interaction to what Dewey and Bentley (1949; see also Dewey 1929) called transaction. This might seem a somewhat subtle distinction, but it seems to me that interaction does not offer a strong enough account of what life is like. In The Dead Poets Society, Robin Williams’ character remarks “we are no laughing at you, just near you” and to me interaction can seem like just acting near to others, who then act in return. Interaction can be seen as a purely mental, perceptual process. Conversely, in a transaction, something is always exchanged. Whatever we do in the world, both we ourselves and the world are changed. To act implies changing something in the world, but it also implies some alteration in the state of the actor.

To put it another way, artefacts and other objects mediate our ability to originate action, they both facilitate and frustrate our desires. But at the same time our transactions with the world also reflexively change our selves; objects mediate change in our own bodies and minds (Graves-Brown 2013b). Again it might appear that things are somehow “acting back” on us, but again I would argue that inanimate objects are only ever the channels through which genuine action is mediated.

Robot Morals

MQ-9_Reaper_CBPIn the 18th century, the construction of automata became a serious method in attempts to understand how people work – contrary to what you might think, they were more than just toys (Riskin 2004). In the 20th and 21st century, this programme has been revived in the fields of Artificial Inteligence and Cybernetics, with some interesting implications for our understanding of agency. In recent years this has crystalised around debates about the use of drones (UAVs), particularly the MQ-9 Reaper. In point of fact, most of this controversy has been misplaced, in that, although capable of some autonomy, attacks by drones have been directed by human “pilots” from remote locations. Nevertheless, these devices highlight the question of moral responsibility in relation to the agency of things.

It is generally argued that moral responsibility was first addressed in Aritstotle’s Nicomachean Ethics, and the basic issues remain the same. Here: “an agent is described as morally responsible for an action, if it is worthy of praise or blame for having performed the action.” (Hellström 2013: 102). Moreover:

Only a certain kind of agents can be ascribed of responsibility, namely those who possess a capacity for decision. Furthermore, the action has to be voluntary. According to Aristotle, a voluntary action has two distinctive features. First, the action must have its origin in the agent. Second, the agent must be aware of what it is doing.

It seems to me, therefore, that qualification as an agent implies moral responsibility. Along with Aristotle, I have argued that a true agent originates action. I would suggest that his other criteria, those of decision and awareness follow from this, although this leaves open questions as to the levels of awareness of, say, a Venus Flytrap or a shark. In other words, although we can debate the attribution of levels of moral responsibility (e.g. whether individuals or society are “to blame”) the fact remains that to be an agent implies responsibility for action, and that this cannot be the case for most “in-animate” objects. That, to paraphrase Harry Truman, the “buck stops” with an agent.

That these issues are taken seriously is indicated by a current campaign by a number of NGOs seeking a United Nations ban on so-called “killer robots” and that this was the topic of Chatham House conference in February 2014.

Conclusion

Cervantes’ Don Quixote famously mistook windmills for giants. I would suggest that those who draw an animistic conclusion from Actor Network theory are themselves tilting at windmills. At the beginning of the 17th century, wind was one of the few sources of power that did not originate with living things; by the end of the century, a new power, steam, was already on the point of entering the world. Quixote’s deranged mistake was to assume that motion implied agency, much as we feel that the automatic door of a supermarket, or the computer is an agent because it appears to act without human intervention. For Cervantes, the proposition that a windmill was an agent was patently absurd, and only a mad person could believe so. In our own era, it is easy to draw the same conclusion about a brick, but not so when we see robots assembling cars; the incorporation of power into things, the very fact that we talk of inanimate objects, creates ambiguity. Yet technological development has demonstrated, at best, that motion is a necessary condition for agency; and agent is an entity that acts, but to act is more than just to move.

Indeed, we might want to insist that sessile plants are also agents, even though they do not, generally, move in the way that animals or machines do. They grow, set leaves, flowers, seeds and propagate. Climbing plants spread, others spread through runners under the soil. What unites all living things is that they originate; they are a locus of change, rather than simply responding to an external force. In Foucauldian terms we might say that they posses power, rather than simply being powered. To posses power is to be responsible for action, and whilst it might seem absurd to claim that a tree is responsible for its roots undermining my house, this is only because we confuse responsibility with level of awareness thereof, which adds a moral dimension to action. And this is a whole other question, as to whether other animals are aware of “right” and “wrong” and the emerging, quite serious, discussion as to whether machines may one day be held morally responsible (Hellström 2013)

In this discussion, I have sought to emphasise that I am not trying, in Joerges words to posit a “world of actors devoid of things”. Rather I suggest that there is a qualitative difference between entities that act, and those that mediate action. I’m not saying that this is simple. For clearly all actors are themselves media – there are no causeless causes. The endless debates around DNA, nature:nurture, etc. make this abundantly clear. In human society there are endless instances; “we were only obeying orders” or the Milgram 37, where we ask why individuals do not “take responsibility” for their actions. Moreover, I have no desire to simply discard Actor Network Theory. Rather I would propose a Power Network Theory or perhaps a Responsibility Network Theory. When person A shoots person B there is an endless network involved, including the life histories of shooter and victim, those who designed, made and sold the gun, and a whole lot more. But within such a network, only some of the entities will have, and could have, originated actions that led to the shooting, and the gun itself is not one of them.

References

Costall, Alan 1995 Socialising affordances. Theory & Psychology 5(4) 467-481,

Costall, Alan 1997 The meaning of things. Social Analysis 41: 76-85

Costall, Alan and Richards, Ann. 2013. The things that things are for. In Paul Graves-Brown, Rodney Harrison and Angela Piccini (eds) The Oxford Handbook of the Archaeology of the Contemporary World.

De Landa, Manuel 1991 War in the Age of Inelligent Machines. Cambridge Mass.: MIT Press.

Dennett, Daniel 1984 Elbow Room: The Varieties of Free Will Worth Wanting. Cambridge Mass.: MIT Press.

Dewey, John. 1929 Experience and Nature. London: George, Allen and Unwin.

Dewey John and Bentley Arthur 1949 Knowing and the Known. Boston: The Beacon Press.

Gibson, J.J. 1979 The ecological approach to visual perception, Boston: Houghton Mifflin.

Graves-Brown, P. 1995a Fearful Symmetry. World Archaeology

Graves-Brown, P. 1995b Stone tools, dead sheep, saws and urinals. In John Schofield (ed) Lithics in Context HMSO.

Graves-Brown, P. (2013a), Review: Bjørnar Olsen. In defense of things: archaeology and the ontology of objects. Journal of the Royal Anthropological Institute, 19: 183–184

Graves-Brown, P 2013b Inside is out. In Alfredo González Ruibal (ed.), Reclaiming Archaeology: Beyond the Tropes of Modernity. Routledge

Graves-Brown, P. forthcoming a. The Blind Man’s Stick. Review of Malafouris. Journal of Contemporary Archaeology.

Graves-Brown, P. forthcoming b. The Fiends of Commerce smile. Discussion of Lemonnier’s Mundane Objects. HAU: Journal of Ethnographic Theory

Hellström, Thomas. 2013 On the moral responsibility of military robots. Ethics of Information Technology. 15:99–107

McLuhan, Marshall 1964 Understanding Media. McGraw Hill.

Malafouris, Lambros. 2013 How Things Shape the Mind., MIT Press.

Olsen, Bjornar 2010 In Defense of Things: Archaeology and the Ontology of Objects. Walnut Creek: Altamira Press.

Preston, Beth 2000 The functions of things. A philosophical perspective on material culture. In Paul Graves-Brown (ed) Matter, Materiality and Modern Culture. London: Routledge.

Riskin, J. 2004. The defecating duck, or, the ambiguous origins of artificial life. In Brown, B. (ed) Things, 99-133. Chicago, Chicago University Press.

Ryle, Gilbert 1949 The concept of mind. London: Hutchinson.

Sigaut, François 1991 Un couteau ne sert pas à couper, mais en coupant. Structure, fonctionnement et fonction dans l’analyse des objet. In Catherine Perlès (ed) 25 Ans D’Études technologiques en préhistoire. Juan-les-Pins: Éditions APDCA. Pp21-34.

Ucko, P. I. 1969. Penis sheaths: a comparative study. Proceedings of the Royal Anthropological Institute of Great Britain and Ireland 2: 27—67

Leave a Reply