Algorithmic Fanboy

I’m​ a neurotic rule-follower. Whenever I fly I anguish about possible minor transgressions. Is my hand baggage less than the maximum permitted depth of 23 cm? Is my tube of toothpaste under the regulation 100 ml? Is the transparent bag in which I’ve put it transparent enough? Do I have to take my shoes off now? Then there’s the horror of the automated passport reader with its plastic sheep-dip gates. Is my passport the right way up? Are my feet exactly on the yellow sticky-backed plastic foot shapes on the floor, and does it matter? Is my nose really that big? Finally there’s the barely concealed relief when my features, contorted by anxiety, defeat the AI facial recognition system, and I have to go and smile at AN ACTUAL PERSON (am I allowed to smile here? Are THEY allowed to smile? Can’t we just be PEOPLE at this point?) to see if I am who my passport says I am and can be allowed to go home.

Rules are all around us, and they get inside us. Algorithmic rules determine what our phones show us, and what faces do and do not get recognised by the automated gatekeepers of the nation-state. When we look up from the screen, the rule of ‘keep out’ is everywhere, either explicitly in so many words stuck on a barbed-wire fence, or in the deeply internalised rule that you don’t walk into your neighbours’ flat to inspect the paint colour in their kitchen, even if the door is open.

Lorraine Daston’s book about the history of rules is as various as its subject is broad. It will tell you, with great wisdom and humour, about the regulation of traffic jams in 18th-century Paris, about early computers, about the historical emergence of the ‘laws of nature’, about arguments over a ruler’s power to dispense with the strict application of the letter of the law, and about the way in which tailors and consumers bypassed ever more intricate laws restricting who could wear what kinds of clothes in early modern Italy. It will also tell you quite a bit about what it means to be modern.

Daston divides ‘rules’ into three broad kinds: tools of measurement, models or paradigms, and laws. The first and third of these are more or less self-evident: one set of things we call ‘rulers’ provide consistent measurements of lines, while the other, lawmakers, impose consistency in the judgment of behaviour. Both display and encourage what we call ‘regularity’. But the second of these three types of rule – the rule which takes the form of a paradigmatic example – is less familiar. In the pre-modern world an individual or an instance of a building or a statue could serve as a guide to later examples. Such a paradigmatic instance of an art or an entity might be described as a ‘model’ or a ‘canon’ – and that word (canon) derives from the name of the giant cane plant, which grows dead straight and so could be used as a ruler or tool of measurement. Pliny the Elder described Polykleitos’s statue of a spear-bearer as a ‘canon’ or ‘model statue’ that could be used as a guide or rule for later artists in representing the human form. The actions of a model of conduct (Jesus, or St Benedict) could also provide a broad ‘rule’ for behaviour. Daston argues that we’ve almost lost the capacity to think of a ‘model’ as a rule in this broad sense, and one aim of her book is to explain and recuperate this conception of what a rule might be.

A rule in the form of a ‘model’ or a ‘canon’ doesn’t prescribe a particular form of behaviour, since a ‘canonical’ sculpture could act as a model for later statues in a different stone or with curlier hair. That makes it a kind of rule which Daston would describe as ‘thick’ or ‘supple’, since it embraces many possible particularities. ‘Thick’ rules of this kind ‘did not aim to anticipate all particulars. Rather, they drew attention to the range and kinds of deviations and exceptions to the rules,’ and they required discretion and experience of a number of analogous instances in order to apply them. A pre-modern ‘thick rule’ might be akin to the instructions you’d give someone who knows how to cook when telling them how to make your favourite pasta. They’d have the experience to be able to supply the unstated details (chop up the garlic before you put it in the pan, get a saucepan out, turn on the gas). ‘Thin’ rules, by contrast, typically address very specific circumstances, and would include the kinds of detailed regulation that stop you taking your hamster or your shaving foam on a flight to New York.

One of Daston’s wider historical claims is that ‘thick’ rules – rules in the form of broad paradigms to emulate or as general guides to practice – tend to give way to the ‘thin’ rules of modernity, designed to placate the officious traffic warden who lurks within all authorities, which ‘assume complete uniformity in execution and conditions of application’. She relates this gradual thinning down of rules to the development of the modern nation-state, which both required and enabled unity of practice across broad geographical regions. A single zone with one army could determine consistency of currency and of weights and measures, while the development of international and colonial trade further encouraged the development of supranational laws and protocols that were invariant within the boundaries of the trading zone.

Those changes in attitudes towards rules went along with a set of interconnected scientific, ethical and mathematical developments in Western Europe. Newton’s ‘laws’ of gravity and of motion and the arguments of political and ethical theorists who sought to describe human behaviour as intrinsically subject to a range of ‘natural laws’ – either of internal human tendencies or divinely ordained ethical principles – variously codified understandings of the human and natural world. These developments were roughly concurrent with a growing political scepticism about absolute rulers, whose gracious power to issue pardons, or to use the prerogative power of equity purely at their own discretion to modify or correct the letter of the law, came to be seen as inimical to ‘the rule of law’, under which lawmakers and subjects alike were bound by law’s dominion. This wide array of interconnected shifts in thinking about rules not only dethroned (or at least qualified the authority of) absolute monarchs, but also delivered a blow to the notion that a single person might serve as a ‘rule’ or ‘model’ of conduct, from instances of whose behaviour precepts for actions could be derived. Why bind yourself to a ‘canon’ or subject yourself to a monarch, when the democratic nation-state could establish a whole series of rules and regulations which tied up reality into a predictable and tidy shape? It also dethroned miracles, which came to be seen as so contrary to the hard necessity with which the ‘laws’ of nature operated that they could only be (for most thinkers) impossibilities or human inventions.

Thin rules now rule, and for Daston the thinnest of all rules is the algorithm – a specific formula directed at solving a single clearly defined problem. The word ‘algorithm’ derives from the name of the Persian polymath Muhammad ibn Musa al-Kharizmi, and originally meant no more than a mathematical operation carried out in what are still usually called Arabic (though they were originally Indian) numerals. Early modern students might learn such operations by practising on a series of examples rather than from learning a mathematical formula. Later versions of the algorithm were, Daston argues, ‘thinner’ and more specific, and she relates their rise to the development of capitalist economies. The first things called ‘computers’ were the (mostly male) assistants to astronomers who did the lower-level calculations for the books that predicted the paths of the stars. The development of automated calculating machines led to the replacement of these relatively well-paid and skilled men by larger numbers of less well-paid and less highly skilled women, whose job it was to punch the cards which provided the rules for the ancestors of what we now call ‘computers’. The development of mechanised algorithmic methods of problem-solving were, Daston argues, one aspect of the tendency of market capitalism to break down the complex processes of manufacture into constituent simple rules, which could be followed by a relatively unskilled and therefore cheap labour force. Looms and computing both went the same way: from large-scale acts of complex human craft towards thin rules for tiny and tidy operations dispersed among many workers who were increasingly replaceable by machines.

The book is an exemplary intellectual history: a rangy, quirky, lucid and profound discussion of the twisted pathways that make us as much of a mess as we are now. It displays a (cautiously voiced) nostalgia for the ‘pre-modern’ flexible or supple rule, which includes within itself a complex of examples and potential instances, and a corresponding suspicion of the rigidities of the algorithmic, ‘thin’ rule-bound nature of modernity. It makes me wonder why I don’t dream more often than I do of living in a medieval village in which hard rules were just soft and fuzzy guidelines – provided, of course, that I could also enjoy all those nice things rules can get us (like antibiotics and vaccines and accident and emergency departments and planning regulations to stop my neighbours taking all my light and a better than Dogberry police force) but minus the horrors of Twitter and Elon Musk. It also makes me wonder whether the world of rules that now surrounds us is quite as thin and cold as Daston sometimes implies. Is the concept of a rule in the form of a ‘model’ or paradigmatic instance quite so remote from our way of thinking about rules as she suggests? Or to put that another way, are we moderns perhaps a little more pre-modern than we might want to believe?

Probably. The earliest English use of the word ‘model’ occurred in a translation of Euclid by the mathematician John Dee in 1570, and through the 17th century the word could be used to mean, variously, a plan or design, an ‘ideal example’ and ‘a unit of scale’. So in this last sense an architectural feature which conformed to the golden ratio of 1.62:1 might be twenty ‘models’ wide and twelve ‘models’ high. Since a ‘model’ in the sense of a unit of scale might correspond to a yard or a cubit or a millimetre or a mile or a kilometre, such an architectural design could be scalable to any dimension – from what we now call a ‘model’, as in Airfix or Hornby, to a full-on Heathrow Airport nightmare of looming vastness. A ‘model’ wouldn’t determine the size of what you built, or what colour you painted it, but it would determine its mathematical proportions. The history of the word ‘model’, that is, might suggest that the idea of a ‘canon’ or exemplary guide to future practice always had the potential within it to become a mathematical rule which prescribed how widely spaced the Corinthian columns on your façade should be. But other usages of the word indicate that we can still think of rules as embedded in a paradigmatic instance. A ‘model’ in the Vogue sense can sometimes be regarded as a physical ideal. A ‘role model’ is someone who embodies a general goal for living, rather than prescribing a specific set of rules about how to live.

But it’s certainly true that the mathematical origins of the word ‘model’ have come to dominate the looser, exemplary sense. Biologists and economists and mathematicians routinely describe themselves as making ‘models’ of complex phenomena. Even that usage, though, retains something of the combination of mathematical rigidity and variability according to circumstance that’s implicit in the earlier history of the word. A mathematical ‘model’ of the proliferation of long-tailed tits in a particular habitat, or of the relationship between inflation and average income in a developed economy, wouldn’t claim to be rigidly predictive of all those phenomena down to the last farthing or the last feather in the nest of the last brood of chicks. Rather it would claim to be heuristic, or broadly predictive of what is probable. Only a bad or mad economist (or an astrologer) would claim to be able to predict by rule, model or algorithm exactly what your income will be by 2025 (though the smart money says it will be lower in real terms than it is now), while a good one would say that within a range of probability or certain parameters it might be x or y or more probably x minus quite a lot of y. The ‘model’ driven by complex algorithms, that is, might often be a kind of probable approximation aspiring to the condition of a rule, and it might be rather less ‘modern’ in the sense of ‘thin’, ‘detailed’ and ‘invariable’ than Daston allows, and a bit more akin to the classical idea of a ‘canon’, or broad guide to the shape of things to come, which might be modified in the light of newly discovered data.

Various forms of algorithm can indeed serve capitalist economies – by breaking down complex operations into simple ones, or by predicting with a near aim the movements of markets. They can make a lot of money. But maybe we are so keen to see algorithmic rules in general as dominant forces in our age because they serve a wider human fantasy: that we can model operations on the outer edge of, or even beyond, predictability (human desires, the weather, the daftness of markets, the way different people write) by creating an intricate set of thin rules that can replicate the complexity of the thing they seek to model. The 21st century is in love with that fantasy, which like many fantasies has inside it a nightmare struggling to get out.

The nightmare is that we ourselves might be no more than an elaborate set of rules, and that we might even be able to write ourselves out of reality by modelling reality so well that we’re no longer necessary to it. ChatGPT, the perfect dim cheat’s AI essay-writing tool, is a set of thin rules aspiring to be so thick (and boy does it often succeed in sounding thick …) that they can generate words which seem to have been written by a human agent. But like all models, ChatGPT doesn’t generate a replica of human behaviour. It generates a predictive approximation of what human-like behaviour might be in a given situation. The algorithmically driven ‘intelligence’ of ChatGPT is a pain because it might enable lazy students to be even lazier, but it’s strange that it’s given rise to such histrionic expressions of terror, since its chief genius is its ability to use words to say nothing at length. If I ask it to ‘write an essay in the style of Colin Burrow’ it delivers a piece of pap about Shakespeare’s sonnets and how lovely they are, which says ‘in conclusion’ that ‘Colin Burrow’s meticulous attention to detail and clear, concise prose would surely appreciate the depth and complexity of Shakespeare’s exploration of love in the Sonnets, and its enduring impact on literature and culture.’ At last I have an algorithmic fanboy – even if it clearly hasn’t grasped the concept of what it is to write ‘in the style of X’, and even if it hasn’t really grasped the implicit rule that an essay can be an exploration which goes off to odd places en route to somewhere strange, and so doesn’t have to end with ‘in conclusion’.

We might think there’s something distinctively modern about the way we are tied up in a web of fussy regulations, and we might get the apocalyptic wobblies about artificial intelligence destroying the concept of the human by breaking us down into a finite series of thin rules. But reality remains gorgeously difficult to regulate, just as complex human behaviours remain more or less impossible to model because we humans are a bunch of stubborn-minded bastards who just won’t do what we’re told. That’s probably a good thing, though future events may of course provide exceptions to that general rule.

Even if​ Daston can sometimes make us seem less pre-modern in our thinking than we are, this is nonetheless a book for our times. It reminds us that developed societies within wide and uniform territories tend to multiply regulations. As specific regulations increase, so the principle that all rules may imply exceptions and therefore invite judgment and discretion in their application comes into question. This means that the proliferation of thin rules can imperil the concept of the thick and flexible rule, and can both express and generate fear of ‘discretion’, broadly conceived, in the application of a rule. There is good reason for that fear: a broad rule which has to be applied with discretionary flexibility by a wise ruler could also be bent to suit the will of an absolute tyrant, or quietly twisted to fit the prejudices of those tasked with enforcing it. So a broad rule according to which a police officer can stop and search anyone at any time on ‘reasonable grounds’ both invites discretion in its application and enables prejudice to be enacted in the choice of people whom the officer believes it reasonable to stop, and in such cases distinguishing the ‘reasonable’ from the ‘prejudiced’ may become practically difficult as well as politically contentious. The general tendency in Western responses to the fear that tyrants can manipulate thick rules has been to produce increasingly specific regulations which restrict the elasticity of general principles, and therefore circumscribe the discretionary power of political authorities, and it’s not hard to see why that tendency should prevail within democratic states. But a general rule about rules is that one rule breeds another rule developed to catch an exception to the first rule, and so (potentially) ad infinitum, until there are so many darn rules that nobody can be bothered to grasp or obey most of them.

That recursive process can generate in human subjects an irritated revulsion from the multiplicity of micro-regulations – or at least from micro-regulations which are seen to be imposed by an external authority rather than those which have had time to be internalised in the conduct of subjects. Such irritation can be inflamed by populist autocrats who claim that they want to cut through the red tape which restricts ‘our’ freedoms in the name of ‘common sense’ or ‘liberty’, or by the Jacob Rees-Moggs of the world, who offer the seductive warmth of a bonfire of regulations, which will warm and revive ‘our’ freedoms. In this political dynamic (which is probably intrinsic to democratic states during an extended period of peace) one can be fairly sure that among the regulations to be chucked into the bonfire will be those which just happen not to be in the interests of those who ignited it – expensive workers’ rights, or tedious little bits of red tape like planning regulations on commercial developments, or the large and generous and therefore also expensive principle that refugees should be given shelter.

Thin rules tend to drive out thick, and not just the thick rules which they seek to refine, but the general principle that rules can or even should be as thick as is compatible with their underlying purpose. So Article 38 of the Charter of Fundamental Rights of the European Union states simply that ‘Union policies shall ensure a high level of consumer protection.’ The specific regulatory implementations of this broad and flexible principle include the EU Novel Foods Regulation, which sets down, inter alia, the specific regulation that Antarctic krill oil from Euphausia superba should be labelled as ‘lipid extract from the crustacean Antarctic krill (Euphausia superba)’. The proliferation of such regulations – or at least the fantastical multiplication of them by those hostile to the EU (the claim that bendy bananas are banned, for instance) – combined with a widespread fear at both ends of the political spectrum that broadly framed rules can be flexed according to the political motives of those empowered to interpret them has contributed to the gradual erosion of willingness to accept such thick, flexible and humane rules as Article 15.2 of the charter (‘Every citizen of the union has the freedom to seek employment, to work, to exercise the right of establishment and to provide services in any member state’), and Article 19.2: ‘No one may be removed, expelled or extradited to a state where there is a serious risk that he or she would be subjected to the death penalty, torture or other inhuman or degrading treatment or punishment.’ Thin rules tend not just to drive out thick ones, but (in a diseased political environment in which the instruments of government are not trusted to apply general rules for the common good) can be made to gnaw away at the broad principles beneath the thick rules. As Daston reminds us, modernity is fragile, and it is either reassuring or terrifying to be told that ‘when the background conditions for thin, rigid rules suddenly collapse, thick, flexible rules return, no matter what the epoch.’