More selected projects

The Octopus and the Machine
Sam Ludford

A newspaper article published in the Daily Telegraph in October 2008 describes the exploits of an octopus named Otto, who, apparently acquiring a taste for chaos after learning how to short circuit the light above his tank by squirting jets of water at it, embarked on a rampage of escalating mischief [1]. In addition to interfering with the electrics, by the end of the spree Otto’s misdemeanours included damaging his tank by throwing stones against the glass, juggling hermit crabs, rearranging his tank “much to the distress of his fellow occupants,” and squandering the resources of the aquarium staff, who took some time to get to the bottom of the matter.

The aquarium’s director was quick to discredit Otto’s character. “We’ve put the light a bit higher now so he shouldn’t be able to reach it. But Otto is constantly craving for attention and always comes up with new stunts.” Another spokesperson stressed the moral gravity of the issue: “It was a serious matter because it shorted the electricity supply to the whole aquarium that threatened the lives of the other animals when the water pumps ceased to work.” Despite admitting that he was likely bored due to the winter closure of the aquarium, the responses make clear that responsibility for the danger to fellow sea critters, not to mention the trauma endured by those poor hermit crabs, lay squarely with Otto. 

The inhabitants of the Sea Star aquarium (of Coburg, Germany) depend for their continued existence on a technological bureaucracy comprised of both human and nonhuman components. These include cooling and heating systems, automatic spotlights, tanks with precise structural properties, animal specialists and security guards, timetables, routines, and so on. Unsatisfied with being merely maintained, Otto had other plans. His crime was to co-opt the apparatus of the bureaucracy for ends other than those intended by its human architects. The authoritarian clampdown that followed takes the smooth functioning of this regime as axiomatic, its moral probing either off-limits or genuinely unthinkable by those enforcing it. Nowhere is it considered that responsibility for the risk to animals existing in an artificial environment highly sensitive to technological malfunction might lie with whoever put them there in the first place. 

As a reader of the article it is difficult not to root for Otto. The journalist, well aware of this, delivers the story with a tongue-in-cheek deadpan highlighting the absurdity of the earnest displayed by the aquarium’s representatives. The comic tension lies in the fact that in order to lay the blame at Otto’s door the institution must characterise him in a manner that only makes him more relatable. Describing him as bored, they convey a being possessing its own interests and agendas. Described as attention seeking, Otto is imagined enjoying interaction for its own sake. Portrayed as a moral agent Otto also becomes a moral patient, a victim of the aquarium and its repressive technological bureaucracy. No wonder he starts trashing the place.

Otto’s pranks are not unique among octopi. In a recent book on the evolution of the mind in cephalopods, Peter Godfrey-Smith relays a series of anecdotes arising from animal intelligence researchers about octopi stalling attempts to assess their cognitive faculties by failing to conform to the assumptions built into their experiments. [2]

In some early scientific work on animal learning by Peter Dews [3], three octopi were rewarded with a piece of fish whenever they pulled a lever hanging down into their tank. Two of the octopi—Albert and Bertram—learned to do so with reasonable consistency, while the third—Charles—was not so compliant. The paper documenting the experiment describes Charles as squirting the experimenters when they approached the tank, absconding with the light, and applying so much force to the lever that it broke, “forcing the early termination of the experiment.”

Behaviourism, the prevailing paradigm guiding Dews’ research when it was conducted in 1959, is reflected both in design of the experiment and in his response to Charles’ behaviour. The introduction to the paper makes clear that the experiment is intended to extend the demonstration of ‘operant conditioning’ model of learning, already enjoying success in research on mammals, to other animal groups relatively unstudied at the time. The experiment attempts to modify the behaviour of the octopi by rewarding lever pulling actions, which would add evidential weight to the theory that operant conditioning is the mechanism behind learned behaviour. 

Discussing this from the heady vantage of 2016, in a climate distinctly more frosty to the behaviourist program, Godfrey-Smith points out that one of the assumptions guiding experiments of this kind is that animals of a given species will behave in more or less the same way in all cases except where particular behaviours have been learned. Hence Dews’ need to account for Charles’ behaviour with factors in his experience instead of factors in his temperament. But one of the (rare) clear lessons from octopus research is that, at least in their case, this assumption is false: temperament varies greatly from individual to individual [4]. Perhaps Charles was just more feisty, or more playful, or more prone to the grumps.

The degree of individuality displayed by octopi represents a major difficulty in subjecting them to behavioural analyses. It casts doubt on whether observations made on particular sets of octopi can be generalised to others. To use a phrase of Michel Callon’s [5], it undermines the ability of particular individuals or groups to act as ‘representatives’ for those not presently accessible for observation by scientists. In Callon’s own work on the attempts to introduce Japanese techniques of scallop cultivation to France, efforts to seed scallop larvae on the sea bed using a towline of ‘collectors’ made from netted bags were initially successful, generating funding for the project from interested parties—but then failed spectacularly in all subsequent years. It is unknown whether this ‘mutiny’ stems from the presence of some hidden variables or from a failure of representation by the scallops of the initial year, since funding was withdrawn and the research abandoned. But it serves as a cautionary tale: variability in the temperament of octopi ensures that similar mutinies are a very real possibility wherever conclusions are drawn based on inductive generalisations drawn from sets of particular octopi (as they always must be). The heterogeneity of the octopus population provides a resistance to its consolidation as a single object of scientific study [6].

There is another complication thrown up by the research, concerning the propensity of octopi to interfere with the experimental apparatus. In the case of Dews’ experiment Charles broke the lever, stole the light, and disrupted the experimenters by firing jets of water at them. Clearly it is a problem when the thing being studied affects the experimental setup. But with the octopus it runs deeper than this. It seems, in fact, that the attributes which make it so difficult to scientifically capture the nature of octopus intelligence are those most suggestive of a highly developed mind: a capacity for play, the ability to distinguish and respond variably to different humans, the use of tools, the ability to shift priorities, an often uncanny awareness of captivity, and so on. A striking example comes from octopus research Jean Boal, conveyed to Godfrey-Smith in a private correspondence, who relates the tale like this.

“Octopuses love to eat crab, but in the lab they are often fed on thawed-out frozen shrimp or squid. It takes octopuses a while to get used to these second-rate foods, but eventually they do. One day Boal was walking down a row of tanks, feeding each octopus a piece of thawed squid as she passed. On reaching the end of the row, she walked back the way she’d come. The octopus in the first tank, though, seemed to be waiting for her. It had not eaten its squid, but instead was holding it conspicuously. As Boal stood there, the octopus made its way slowly across the tank toward the outflow pipe, watching her all the way. When it reached the outflow pipe, still watching her, it dumped the scrap of squid down the drain.” [7]

What was this octopus doing? Making a point? Conducting an experiment of its own? Whatever the reasons, it is difficult to explain this story without attributing complex intentions of some sort or another to the octopus. These kinds of occurrences are likely impossible to recreate reliably in experimental conditions, and those experiments which aim to test some kind of cognitive capacity in octopi are often met with indifference or outright hostility by their subjects. The cunning of the octopus eludes quantification not because it is an immaterial cunning, but because in the situations that count it is used to disrupt the means of quantification.

I want to suggest in this essay that these two difficulties in the study of octopus intelligence—the failure of generalisation deriving from their heterogeneity as a group, and the difficulty of placing them in experimental conditions due to their propensity to act disruptively or differently—represent two distinct but deeply entwined conflicts between octopus and machine. Moreover this conflict, which the octopus seems to be winning, reveals through its structure and the way it manifests a conflict between human and machine—a conflict we may well be losing.

The behaviourist program was an early attempt to subject the mind to a thoroughgoing scientific analysis, aiming to explain the behaviour of an organism in reference to external causes [8]. This would reduce mentalistic explanations of behaviour (Albert pulled the lever because he desired some fish and believed that pulling the lever would get him some) to causal explanations phrased in terms of dispositions (Albert pulled the lever because he had been conditioned to have a disposition for lever pulling by being plied with fish as a reward for doing so in the past). As a fully fledged philosophical program behaviourism attempted to identify mental items such as beliefs, desires and intentions with dispositions to behave. Albert’s belief that the lever would get him some fish just is his disposition to pull the lever in the presence of a desire to eat fish; Albert’s desire to eat fish just is his disposition to yank nearby pieces of fish into his beak.  

‘Disposition’ is clearly a key concept here. Discussing the transition from Aristotelean to modern science, Wilfrid Sellars points out that ‘disposition’ is a development of the Aristotelean notion of a ‘capacity.’

“We must say not only that a thing could have behaved otherwise than it did, but also that it would have behaved differently if the circumstances had been different. Water is not only “capable” of freezing; it would freeze if the temperature were reduced to 32F. To clarify this we must distinguish between the concepts of capacity and dispositional property. Both involve reference to the laws of nature, but the former is a weaker notion and is essentially negative. To say that a thing is capable of being in a certain state is to say that its being in that state is contrary to no law of nature. On the other hand, to speak of a dispositional property of the thing with respect to that state involves a positive reference to a law associating that state with a certain kind of circumstance.” [9]
The stronger sense of the counterfactual embodied in the concept of a disposition turned scientific explanation from a business of relating things via their natures and capacities to a business of relating events via their causal relationships and the laws governing the transition between states. Taken in combination with reductionism, this picture forms the mechanistic image of the world. According to this image understanding a system is a question of understanding what it does—how it behaves, its dispositions—by decomposing it into its component parts and piecing together the relationships between its parts in terms of how they behave. 

Behaviourism was an attempt to apply this kind of mechanistic analysis to the mind. By resisting it, as well as subsequent theories which follow the same underlying rationale, the octopus resists being understood on the terms of the machine. 

One way the octopus achieves this is by being radically amorphous. The heterogeneity of temperament is one example, throwing up barriers to those looking for general principles of octopus psychology. But this amorphousness also occurs at the level of the individual. For example octopi elude theories of embodied cognition [10], which argue that certain kinds of knowledge are coded into the structure of the body, simply because their bodies have so little structure. This formlessness of octopus bodies also makes them difficult to photograph. In Godfrey-Smith’s book a description of a diving site where strange octopus behaviour was observed is accompanied by a photograph on the following page, “taken just off the edge of the site, […] to give you a sense of how these animals look.” [11] But looking at it does not give much sense of anything. The components are all there: the tentacles, the bulbous head, yet their configuration in the image seems to consist of nothing more than accidents of the moment. Absent is an essential form, some kind of deep structure. In each subsequent photo this form is totally altered, multiplying the absences. Equally unsatisfying, the article describing Otto’s antics is accompanied by a grainy image of a nondescript octopus clutching a jar. The caption leaves it ambiguous whether this is a stock photo or an authentic mugshot of the culprit. 

Video, too, is often unreliable. A CNN report from 2016 features phone footage of an octopus leaping onto a rock to wrestle a crab back into the water. [12] A crab on a rock, a rapid flurry of movement; then, eventually, no more crab. Despite the camera lingering on the seething mass of tentacles and claws for some time it all seems somehow unreal, as if it can never be fully clear what exactly is being witnessed. The reporters seem sufficiently doubtful that they feel the need to assure the audience that scientists have confirmed that what was just seen is in fact possible.

The second way the octopus defies the machine seems at first glance to be more straightforward: they do so by stealing, appropriating, tampering with, and sometimes destroying actual machines. This occurs outside the laboratory as well as inside. One of the most popular octopus videos on Youtube shows an octopus acrobatically breaking into a metal crab trap, gobbling all the crabs, and escaping at the last moment through a tube specifically designed to keep things in, thus thwarting Steve, the fisherman [13]. In effect it hacks the trap, repurposing it for its own needs. 

In the laboratory, however, this kind of behaviour takes on a special poignancy. When an octopus interferes with the mechanisms of an experimental setup, it is interfering with the very apparatus being used to subject it to mechanistic analysis. In these situations it can be seen that the amorphousness of octopi and their reputation for abusing machinery are two sides of the same coin. The formlessness of the octopus is also the limit of human technology to produce knowledge of the form of the octopus. This refers both to the mechanical technologies used in experiments and keeping octopi in captivity, and to the conceptual technologies aiming to reveal the form of their nature. 

These points in mind, it is time to return to Otto. What was really going on? Why was it so natural for the aquarium staff to treat Otto as a responsible agent? And why is the gut response of an onlooker to root for Otto? 

I’d like to suggest an answer to this question: by portraying Otto as an active agent, the aquarium staff are deflecting (consciously or otherwise) from a third actor: the technological bureaucracy itself. In particular, they are deflecting from the pull it exerts not over the animals it incarcerates and maintains, but the humans that tend to it. Imagine that no agency was attributed to Otto, that he was seen as a piece of the environment, like a special kind of weather. If this was the case then when the tanks started getting rearranged and the electrics started shorting the aquarium staff would have to conclude that they were faced with a malfunction. But malfunction is precisely the moment at which a technology ceases to be part of the environment and starts exerting its own agency, making demands on humans, forcing them to do things they had never intended to and typically don’t want to do. When electricians have to camp out on the aquarium floor to find the cause of the blackout, the blame can only lie with themselves, or at least with whoever designed or built the aquarium’s systems. This conclusion is resisted by shifting the agency of the machines onto Otto: the lights and the tank are no longer indicative of a malfunction, but the interference with perfectly well functioning machinery of a malicious actor. Otto is portrayed this way because it is in the interests of the aquarium staff to do so.

What, then, of the warmth felt for Otto by the wider audience? I’ve suggested above that the attribution of agency to the machine is averted by offloading it onto Otto. In neither of these cases are humans considered significant actors. As such, the situation represents a certain kind of powerlessness of the humans involved, a denial of their own agency. As representatives of the aquarium defence of its technological systems is axiomatic, so an actor that interferes with them is necessarily a bad actor. But for anyone who is not professionally invested, there is no reason Otto must be a bad actor. The aquarium is in many ways a microcosm for a much larger technological bureaucracy, one which encompasses and permeates the whole of human society. It includes computers, everyday gadgets, cars, industrial machinery, scientific apparatus, the mechanistic ideas guiding scientific discovery, machine-oriented theories of the mind, and so on. Irrespective of the truth, goodness or value of these things, they come with their own demands, their own agency manifesting as unintended and sometimes undesirable effects. A sense of powerlessness in the face of this is hardly surprising, and nor, for that matter, is the welcome of a creature that appears capable of defying the regime. 

The octopus is such a creature. The professional loyalties of the aquarium staff forced them to attribute a negative agency to Otto, but this need not be the case for everyone else. Anyone who is less than convinced that the global technological regime is a wholly good thing is more likely to see the octopus in a more positive light—perhaps, even, as an emancipatory actor. In a 1984 essay, Thomas Pynchon described how the figure of Ned Lud took on a mythical potency for the Luddite movement famous for destroying the machinery used in cotton mills which threatened their livelihoods.

“By the time his name was taken up by the frame-breakers of 1812, historical Ned Lud was well absorbed into the more or less sarcastic nickname ‘King (or Captain) Ludd,’ and was now all mystery, resonance and dark fun: a more-than-human presence, out in the night, roaming the hosiery districts of England, possessed by a single comic schtick—every time he spots a stocking-frame he goes crazy and proceeds to trash it.” [14]

At the end of the essay Pynchon wonders whether the computer age can expect any kind of rekindling of Luddite hostility to technology. He suggests not, reasoning that the technological bureaucracy is already so vast and all-pervasive that humans are so dependent on it that we are no longer able even to imagine ways of resisting it. This may be true, but need that be the end of the story? Perhaps the response to human powerlessness to resist lies in finding that power in nonhuman actors, such as the octopus, who continue to elude the machine.




  1. “Otto the octopus wreaks havoc,” The Telegraph, accessed March 28, 2018,
  2. Peter Godfrey-Smith, Other Minds: The Octopus and the Evolution of Intelligent Life on Earth (London: William Collins, 2016)
  3. Peter Dews, “Some observations on an Operant in the Octopus,” Journal of the Experimental Analysis of Behaviour 2, no. 1 (1959): 57-63
  4. Godfrey-Smith, Other Minds, 53-54.
  5. John Law, “Notes on Theory of the Actor-Network: Ordering, Strategy and Heterogeneity”, accessed March 28, 2018,
  6. Michel Callon, “Some elements of a sociology of translation: domestication of scallops and the fishermen of St Brieuc Bay,” published in John Law, Power action and belief: a new sociology of knowledge? (London: Routledge, 1986): 196-223.
  7. Godfrey-Smith, Other Minds, 57.
  8. Malcolm Ashmore, “Behaviour Modification of a Catflap”,, accessed March 28, 2018,
  9. Wilfrid Sellars, “Aristotelian Philosophies of Mind,” Philosophy for the Future, The Quest of Modern Materialism, edited by Roy Wood Sellars, V.J. McGill, and Marvin Farber (New York: The MacMillan Co., 1949): 544-70.
  10. Godfrey-Smith, Other Minds, 75.
  11. Godfrey-Smith, Other Minds, 61-62.
  12. “Octopus leaps out of water, grabs crab,” CNN, accessed March 28, 2018,
  13. “Octopus Steals Crab From Fisherman | Super Smart Animals | BBC Earth,” BBC Earth, accessed March 28, 2018,
  14. Thomas Pynchon, “Is It OK to be a Luddite?” New York Times, accessed March 28, 2018,