What I’ve been reading:
Brian Christian and Tom Griffiths’ Algorithms to Live By
Laura Beatty’s Looking For Theophrastus
Binyamin Applebaum’s The Economists’ Hour
I've enjoyed reading Brian Christian and Tom Griffith's *Algorithms To Live By*. I've written previously about Christian's The Alignment Problem and its relationship to wider questions of philosophy and history, and of course that's broadly how I've read this book. This book is a tad different; less about AI and more about the most common algorithms used in programming / statistics in general, and though the title suggests it's a "use algorithms to improve your life" a la Newport or Atomic Habits, in reality the book's a high level introduction to statistical methods and computation. A good bridge, in other words, for humanists wondering how to think through programming. In this post, I want to use Christian and Griffith to think a bit about how reliance on data (analytics/science/forecasting/machine learning etc etc) accelerates the sclerosis of political institutions. I argue that a party in government accumulates information, slows in its attempts to parse that data, and mistakes the map for the territory. This links to Fukuyama's reading of Huntington, and helps explain why parties and states tend to become senescent. I then take a step back, and think more generally about how this relates to parliamentary timekeeping and democratic systems in general.
Statistics aims to find meaningful patterns in the world. In order to do this, you need to decide which factors to consider and how you might capture these factors into variables. In other words, to find patterns in the world you first need to model it, and so the patterns in the world can best be described as patterns in your model of the world. Statistics, then, is as much of an art as a science; figuring out what you want to track, how you're going to measure it, and where you're going to draw the line is profoundly difficult. Too few variables and you won't find any patterns. Too many variables however and you'll *always* find a pattern.
Griffith and Christian call this overfitting "a kind of idolatry of data":
Throughout history, religious texts have warned their followers against idolatry: the worshippers of statues, paintings, relics, and other tangible artefacts in lieu of the intangible deities those artefacts represent. The First Commandment, for instance, warns against bowing down to "any graven image, or any likeness of any thing that is in heaven." And in the book of kings, a bronze snake made at God's orders becomes an object of worship and incense burning, instead of god himself. (God is not amused.) Fundamentally, overfitting is a kind of idolatry of data, a consequence of focussing on what we've been able to measure rather than what matters. (156)
Idolatry / overfitting is when statisticians focus so much on their models at the expense of the world itself. In a way of course, it's part and parcel of trying to map out new territory; you make observations, record them, then notice what's changed the next time around. You enter a sort of cycle of iterative "improvement." In so doing you begin to build up a map of the world that appears to better describe the territory than the territory itself. The problem - idolatry - is when you mistake that map for the territory "offering prayers to the bronze snake of data rather than the larger force behind it". As the religious simile employed here implies, Christian and Griffith think this is an a intrinsic problem in statistical methods precisely because it's a fundamental human tendency.
Although Christian and Griffith don't explicitly link the two, the problem of idolatry is related to their later argument that age related cognitive "decline" is a problem of organisation rather than decay. As humans age, the quantity of information retained from experience necessarily balloons. "The larger a memory is, the more time it takes to search for and extract a piece of information." (102) "The fundamental challenge of memory really is one of organisation rather than storage." (103). Slowdown, then, is:
an unavoidable consequence of the amount of information we have to navigate getting bigger and bigger. [Michael Ramscar's ] group demonstrated the impact of extra information on human memory by focussing on the case of language. Through a series of simulations, the researchers showed that simply knowing more makes things harder when it comes to recognising words, names, and even letters. No matter how good your organisation scheme is, having to search through more things will inevitably take longer. It's not that we're forgetting, it's that we're remembering. We're becoming archives. (103)
The infrequency of memory lapses "is a testament to how much you know [and] how well you've arranged it." (104). The more experience over time, the more information retained, and the longer it takes to access infrequent associations. This is surely a type of idolatry: since the brain optimises for best models of the world and thus privileges information that's required more frequently, the more information you retain the more difficult it becomes to accommodate new or unusual information into those mental maps. This in turn suggests that as you acquire data/experience your mental model tends towards crystallisation. Anything with memory - whether biological, mechanical, statistical - therefore necessarily becomes less flexible over time.
This is painfully evident in political and social institutions. Francis Fukuyama argues that that the U.S. Forest Service's faith in scientific management exemplifies how institutions become liable to ossification and decay. Seeking to move away from patronage networks controlled by congress, Fernow and Pinchot staffed the U.S. Forest Service on the basis of merit and ran it according to the late 19th century belief in scientific management. "For many years afterward [it] remained the shining example of a high-quality American bureaucracy" (456). Until, of course, it didn't. The problem was that the service was meant to regulate the exploitation of timber. Forest fires, according to the scientific theory of the late 19th century, needed to be fought at all cost; after all, they destroyed valuable resources and threatened homes. Yet these fires only became more ferocious, and over time occupied a greater percentage of the service's resources. By the 1980s the agency was spending a billion a year on firefighting alone. Meanwhile, ecologists had gathered overwhelming evidence that fires formed part of natural forest lifecycles. Yet by the time they were able to convince the agency that fires were critical to timber production, the expansion of suburbia into the forests had made the "let it burn" policy politically untenable. The forestry service, founded to produce and protect timber, now spends billions protecting property by interrupting part of the natural forest cycle and drastically reducing its own timber yield. Yet another limiting factor comes from the 1964 Wilderness Act, which means the Forest Service is simultaneously tasked with protecting the forests it's meant to harvest, while concurrently interrupting forest growth to protect suburban property of the voters who can set its priorities. The "scientific" model of the forest therefore became more real than the actually existing ecosystem. By the time the model was overturned however, the damage of idolatry had been done: a service unwilling to update its model has now become unable to respond because of the secondary effects of its own inflexibility.
Political decay, idolatry of data, and human senescence therefore all describe the same problem. "Political decay," Fukuyama argues, "is therefore in many ways a condition of political development: the old has to breakdown in order to make way for the new." These institutions, "according to Huntington, are "stable, valued, recurring patterns of behaviour"" that "facilitate human collective action". In other words, institutions are social models of the world, schemas which are "genetically hardwired into the human brain." Decay, in turn, is an emergent property of any reasonably accurate model of the world:
Institutions fail to adapt to changing circumstances for a number of reasons. The first is cognitive. Human beings follow institutional rules for reasons that are not entirely rational. Sociologists and anthropologists have speculated, for example, that various religious rules have rational roots in different functional needs - for example, the need to regulate sexuality and reproduction, the requirements for conveying property, organisation for warfare, etc. But fervent religious believers will not abandon their beliefs simply in the face of evidence that they are wrong or lead to bad outcomes. This kind of cognitive rigidity extends well beyond religion, of course. Everyone creates and uses shared mental models of how the world works, and sticks to them in the face of contradictory evidence. This was just as true of Marxism - an avowedly secular and "scientific" doctrine - as of contemporary neoclassical economics. We saw a vivid case of this in the U.S. Forest Service's belief that it possessed "scientific" knowledge about forest management, which led it to persist in its fire suppression policy in the face of accumulating evidence that this was undermining the goal of forest sustainability. 463.
One reason why institutions become brittle, then, is that the longer they survive, the longer their experience crystallises their model of the world. Political decay happens when institutions become victims of their own success. Science becomes less of a verb and more of a noun, a great golden snake which demands tribute because it has statically endured through time. In turn, Fukuyama and Christian's approach would suggest that all maps inevitably become the territory. The more accurate the map, the fewer times you feel the need to check your surroundings, and when your surroundings *do* appear to contradict the map, well, the map has been right before - so surely your experience of the territory is wrong? As Fukuyama goes on to argue moreover, the longer the institution has been around, the longer it has proved to be a useful rule of thumb, the greater the chance that a priestly caste around the idol emerges. All maps, all models, all institutions are born to die.
The Conservative Party can therefore be described as in the process of becoming its own archive. The Conservative Party currently looks to be on course for an electoral defeat the likes of which - according to William Atkinson at ConHome - haven't been seen in the UK since Labour's defeat in 1931 (where they lost 4/5ths of their seats), and which will in all probability compare to the Canadian Conservative Party's electoral annihilation of 1993 (from 167 seats to... 2). These poll numbers are not new. The UK did not wake up yesterday and decide, en masse, that the Conservatives' time was up. The question, then, becomes why the conservatives haven't bothered to readjust. Jonn Elledge suggests the answer lay in a cluster of reasons; the tories need to pretend to their donors, their MPs, their party faithful, that electoral survival is at least plausible for example, the fact that the Conservatives appear to be the natural party of government, that shocks like Canada '93 simply don't happen here within living memory, and finally the sense that 'things like this just don't happen here.' There's something to be said for the argument from experience. To return to Amitav Ghosh on climate change for a moment, a few years of Conservative turmoil and sheer incompetence ending in a catastrophic wipe out would just seem too much like some sort of West Wing to really happen in real life. Another answer, of course, lies back in the idolatry of data. In other words, it's worse than you think: the conservatives actually believe this. Like the U.S. Forest Service, the Post-Coalition Conservatives were elected "to get Brexit done." Each subsequent election, whether for the leader or at the polls, has served to reinforce the brexit-ultra model of the world. In this way then the Conservatives are victims of their own ruthlessness; they got brexit done by any means possible and they were rewarded by electoral success. But brexit is done; the map is out of date, even if the cartographers don't want to hear it.
Parliamentary success then, can be seen as a kind of idolatry. A party gets a mandate, which then sets the direction of travel for their policy for the next five or so years. A particularly successful campaign brings in more MPs - and in the case of brexit, that and subsequent successes tells the party and their outriders that there's one direction, one particular policy goal that the voting public care about above all else. Of course, that positive reinforcement is out of date as soon as the new PM takes office. The economy, foreign affairs, domestic scandals, assorted historical difficulties etc all play second fiddle to the gravity well of the previous election. Each subsequent election of the conservative government, and again we'll have to go back to Brexit and stretch the metaphor to breaking point, has been the equivalent of printing the map on nicer paper. The governments' memory has been Brexit-focussed, so its produced more knowledge about brexit and brexit alone, with a side jaunt into Covid. But the better it got at brexit and brexit-related activities the *worse* it became at actually governing, at being able to respond to normal historical events or even to keep the institutional lights on. The British Parliamentary system is essentially a duopoly; you either get the Conservatives or the Labour Party. The more elections in a row one wins, the greater the chance the next election will be fatal. This creates a sort of fatalism at the heart of political life; a Huntingtonian-Fukuyama-esque historical cycle where it's somehow *natural* that political parties get sclerotic and reach the end of their natural life. But this too is a kind of idolatry of the type that's eaten away at the Conservative Party and, as we've seen, since there's no real impetus for the Conservatives to adjust course there's no real understanding that the political order within the party itself is fragile. After all, defeat is always followed by rebirth.
The problem, then, is how to ensure political organisations are responsive to social and economic changes; how can we ensure the "natural" cycles of decay and rebirth in the parliamentary system are prevented from being reified at the expense of state provision. To be a bit more explicit, I'm worried that the belief in "natural" parliamentary cycles i've explored here is no longer suited to maintaining state capacity. To be more explicit, how do we ensure the government is able to rebuild infrastructure, respond to military threats, and keep Britain functioning as history has begun again? The answer is not in dreams of smashing the blob. You might thing you're hastening the cycle and starting again, but in reality you want to preserve some institutional knowledge in the largest sense - how contracts work, where the keys are, who runs what, that sort of thing. The other problem is that the call to smash the deep state, or whatever other america-brained phrase Right-Jacobins are running with, is what got us in to this mess in the first place. Don't worry, we know the map isn't the territory, that's why we have this *new* map we got a racist with an anime profile pic to draw. A broader approach, an answer that isn't "simply do y instead of x" is to massively expand the Houses of Parliament. Coupled with staggering elections - with MPs serving 4 years, elections running for half the house every 2 - there'd be a better feedback system in place. Since we're going fully into the land of fantasy, I'd quite like to see the honours system revamped; a barony or a lordship for the top taxpayer, for public service, for long service in certain careers, even. Anything to break the introspective tendency; the problem remains that if we're talking about the preservation of parliament, about the maintenance of a tradition, of a state, a nation, or even a civilisation, it's a sure sign that the rot is endemic. I'll finish with MacIntyre's After Virtue:
So when an institution–a university, say, or a farm, or a hospital–is the bearer of a tradition of practice or practices, its common life will be partly, but in a centrally important way, constituted by a continuous argument as to what a university is and ought to be or what good farming is or what good medicine is. Traditions, when vital, embody continuities of conflict. Indeed when a tradition becomes Burkean, it is always dying or dead.