All skin and bone: fragility in a changing world

Of late we’ve all read about flattening the curve of the pandemic; spreading out the burden of patients on the NHS by delaying infections, so that its limited capacity is not overwhelmed by all the severe illness happening at once. The idea is to go from this:

To this:

              This is a salient example of something that matters a lot, which is the fragility of systems. Typically when we create something, whether a company, a machine, a committee etc., we think of it as having a steady capacity, like the flat dotted line in the first picture for NHS capacity. The NHS had a certain number of intensive care beds and ventilators, and that’s what it can manage. The same applies to a company: it has staff employed for 40 hours a week, and we assume that those 40 hours are steady and identical, so the company has some multiple of 40 person-hours a week to use.

              As has been explained in hundreds of articles over the last few months, disease does not follow this pattern. Some problems have a reasonably steady rate of incidence, but infectious disease ebbs and flows. We are accustomed to the flu season every winter, which increases demand for ventilators, as vulnerable people need help to overcome the virus before summer rolls round and people stop crowding together indoors.

              So what capacity should the NHS have: level 1, level 5 or somewhere in between? All capacity costs money, so we can’t just pick the most capacity without suffering elsewhere. After all, that money could be spent on something else: tax cuts for the lobbyists, maybe, or other NHS treatments. If we have enough capacity to cope with the highest peak seen, or to cope with a potential higher peak, then over half of our capacity will, on average, be unused.

              Empty operating theatres, empty hospital beds… these are prime fodder for angry tabloid stories bemoaning government waste and bad health management. Why have we wasted money on stuff we’re not using when we’re not funding expensive cancer treatments for this one person we dug up?

              That, like most tabloid arguments, is utterly stupid. It could easily be that those beds, despite being empty some of the time, still save more lives than the equivalent price of an overpriced and oversold cancer drug. Tabloids never even attempt to consider that; they are oblivious to the variability in demand. While I’m on a little aside, that’s one reason why government programmes are often so expensive: they are, by their nature, universal. They need to reach everyone, and serve everyone. If some people are hard to access, difficult to deal with, or all need help all at once, then the whole point of government intervention is to be there in those difficult times. The private sector can serve the easiest, most profitable people, and actually does best by excluding difficult cases, as they reduce the average profit. Homogeneity is profitable… a point we might come back to.

              Anyway, I think we can all guess that we don’t have capacity at levels 1 or 2. We’ve gone through so many cost-cutting drives to find efficiencies, and 10 years of underfunding, that I’d not be surprised if we were lower than level 4. We don’t even have enough capacity for lowest levels of peak demand. This is the world we live in: we are making hard decisions about allowing some people to die in order to save money. Covid-19 has not suddenly created a situation in which need outstrips capacity; it has merely created a worse one than normal. If I haven’t been clear enough, we have already been letting people die in the name of efficiency. This is normal and necessary.

              Another aside: yes, I do mean necessary. The nature of healthcare is that there is always more that we can do. I do think that we should be doing an awful lot more, but even I recognise that some things that could conceivably be justified as healthcare are too expensive or disruptive to be worth it. So there will always be a balance between saving some health and allowing people to truly live.

              The immediate point is that the NHS had had capacity cut until it was already overburdened, because that ensures the most efficient use of assets by one measure of efficiency. If operating theatres are always in use; beds never unoccupied; doctors’ time always filled (to the extent that many are working double hours), then it necessarily follows that sometimes there will be unmet need and people will die. They won’t always die from famous, newsworthy diseases like a new global pandemic. They will be elderly or infirm people who just got unlucky. They got the flu or had a fall at the wrong time. Or maybe, like my mother, had a stroke and couldn’t keep a bed on the stroke ward with proper rehabilitation and physiotherapy because there were too many strokes, and so after recovering then got dumped on sedatives instead and died without ever really waking up again. At 63 years old.

              God forbid that we have enough stroke beds, and sometimes have them empty, with nurses and physiotherapists having an extra half hour a day to rest or devote to fewer patients.

              One more general point we can see from this example: fragility. I think that there’s more to fragility than simply gradually going over capacity. Fragility is about a system breaking; there are set points beyond which things don’t just get that little bit worse, but a whole lot worse. If you bend a twig, it gradually flexes out of shape, but at some point it’s under so much strain that when you bend it more it snaps. It’s important to be aware of fragility, identify it, expect it and prevent it.

              All systems need spare capacity. In my kitchen, for example, I need work surface to do my chopping, mixing etc., and to put newly dirtied pots and bowls. If there isn’t enough work surface, I have to spend a fair amount of time shifting things around just to get, say, the chopping board next to the bowl I want to put the chopped stuff in. And if, as in one shared house I lived in, all the surfaces are used as storage, it suddenly becomes impossible to cook at all. The free space that my housemates regarded as fair game for dumping cereal packets on was actually better empty.

              In the NHS, empty beds or operating theatre slots are important. Emergencies can crop up, and a spare slot allows managers to rearrange everything. That’s why in those little sliding piece puzzles one slot is empty: you need to be able to move a piece somewhere empty in order to free up its space. If, in the name of efficiency, you fill that slot, but the pieces need to be rearranged, you’re stuck. You’ve broken the system. And if you need to get things in place fast, you might be best off with two free spaces; that makes the puzzles much easier, whether it’s a child’s game or organisational management.

              Systems aren’t often like the picture below:

              Fragility doesn’t just come from inflexibility. There are all sorts of negative feedbacks that might emerge. If, for example, patients start to clog hospital corridors, staff might be less able to get to where they need to be, losing time on the patients they actually had time to book in. Or those extra patients will transmit disease faster to everyone than they would if in a ward space. Without flexibility, a delay somewhere, which becomes very likely, will lead to complete cancellations elsewhere, because staff or equipment suddenly aren’t available…

              That means that systems behave more like this:

              So, having explained how we’ve been killing people for years, what more important things can we apply the same concept to? Fragility is a broad concept that doesn’t just apply to life-saving work in healthcare, but to everything. Before we get to the economy, let’s start with games

              When you play a game with your family, you’re typically faced with risky or less risky choices. Monopoly, for all its flaws as a game, is widely known, so let’s think about that. Should you buy another house on your set or keep a bigger cushion of money in case you land on someone else’s? If you keep a massive financial cushion you’ll never get enough income from your under-improved properties and you’re guaranteed to lose eventually, but if you aim for a high income as soon as possible, and you spend all your cash on houses, then you’ll need to be lucky not to lose quickly. Ideally, with just your family playing, you judge your risks based on everyone else’s success: if someone else gets lucky, you need to spend your cushion to try to keep up, and take your chances now when they're better. With, say, 4 players, that can still lead to a game that lasts a while, although it’s very likely that someone gets unlucky early on and has to find something else to do.

              The same applies in any game. If you play pontoon (also called blackjack, or 21, or vingt-et-un…) you want to beat just one person, the dealer. That’s why the dealer should play quite safe; if he goes bust then everyone else knows they can just stick with what they’ve got and still win. But imagine you’re playing against everyone else. It’s really likely that someone out there will have a 20, even if there’s no obvious 5-card trick or 21. Playing it safe will just lose you your stake. You have to take a big risk to win.

              This is our capitalist market. You are in competition with everyone. In theory, the number of competitors can be modelled as infinite. That’s ‘ideal’. If you play it safe and hold something back for a rainy day, someone else will start up, cut that cost and sell things more cheaply at the same quality (not that quality is much of a signal in modern markets). That person will get all the customers; or else that person will sell at the same price but make more profit, and therefore get all the investors. Either way, your investment in resilience is wasted.

              The key to our current economy is competition (call it capitalism, neoliberalism, just plain badly-run or badly-regulated, or whatever you please; let’s save the argument about terminology for someone else). It’s about competition right now, for profits right now. That means cutting as many corners as possible to grab whatever share of the market you can right now: the future is too expensive to worry about.

              There are all sorts of theoretical answers to this counterproductive set of incentives. You might hope that investors have perfect information and therefore know which companies are more resilient, and therefore also know that those companies offer better long-term prospects. But perfect information is one of the theoretical assumptions that we’re furthest from. Even in the information age, people are more snowed under a weight of misinformation (also called PR or marketing) than they are well-informed.

              You might hope that the barriers to entry and obvious market volatility lead all the cut-throat market entrants to be less risk-taking with their businesses. But there will always be risk-takers hoping to survive through luck rather than good planning, and therefore make a big profit.We have no shortage of desperate people trying to come good with yet another throw of the dice.

              And so we are likely to have repeated new entrants starting up, making profits in the good times, and going broke in the bad times. And if new entrants are providing that competition, big, established companies must follow the same strategy or lose out. And that’s how we get big banks, ‘too big to fail’ taking massive risks and in the financial crisis easily becoming insolvent. The market system we have set up pushes all companies to as close to breaking as they can go in a desperate attempt to keep up.

              In a racing analogy, with only one race, that makes sense. If you’re thinking of a predator and prey, a nice macho analogy, the prey must run faster, no matter what the potential cost. But as soon as you think longer than one event, it doesn’t make sense. If you push your car, or horse, or yourself, far too hard in the first of many races, you’ll likely have an injury or breakdown and do far worse. That’s fragility. Horses are actually a great example. They have evolved to be running machines; creatures that big shouldn't be as fast. But that has come at a cost. They easily break their legs, which have no excess weight, and therefore strength. In the wild, that's just collateral damage of their evolved survival strategy. But the same fragility in society equates to human suffering.

              We choose to set up a system which forces this desperation and makes it difficult to invest for the future. That’s how our markets work, with instant stock trading, and a focus on share price by senior executives, whose reputations are determined by stock performance now, even if their stock options mature in the future.

              Yet, just as with infectious disease, markets are volatile. Demand varies, by season and economic cycle as well as consumer whim. We must therefore expect, as part of our economic system, an endless cycle of bankruptcies, failures and job losses, with only the lucky surviving for any length of time until they become famous enough and respected, or have enough reach and digital information, that they have a massive advantage over new entrants. They can then afford a little safety; in fact, once established, their advantage can be so great that we get a whole load of other market failures, from disinformation (‘marketing’) through to lobbying and regulatory capture.   

              The essential point is that everyone sensible must take those chances, and therefore be lucky. And those few who are successful owe it more to luck than taking the best chances. If I give myself a 2% chance of winning and 196 others are incompetent enough to share the remaining 98% at 0.5% each, I still owe a lot more to luck than to skill if I win. That's true even though I'm four times better than any other rival. In the same example, whichever 196 of us lose, all will still owe their situation more to the system that forced us into taking those chances than their own incompetence.

              In short, capitalism builds fragile systems: it is intrinsic to the way we run it. And when big companies fail, we have seen what happens: public bail-outs. So in the end, we all pay for this disaster of a system, while those who have been lucky enough to have enough to invest, and crucially, enough to pay for lobbyists, end up winning in the good times and not losing in the bad. There are even ways to make money from the bad times, if you have enough to invest in the first place, and enough to spread your risk widely.

              That’s what has happened to privatised services. I don’t know much about our railways, but I have seen enough of the arguments train drivers have had with management to cast that argument as one about fragility: guardless trains, staffed only by the driver, are a safety risk. It’s not that a train can’t be operated only by the driver, if everything works well. But people do get stuck in doors, disabled people need help with wheelchairs, and so on. Much of the time, perhaps, a conductor is not required, but without one the train can be delayed, and then it might miss its slot on a packed trainline which has no spare capacity itself… The problem is obvious.

              There are other criticisms of privatisation, but this is not the place.

We do have an industry which has dealt with fragility, and that is electricity production. A functional National Grid is really a wonder of the modern world. The electricity is on all the time, even though daily demand varies enormously (although reasonably predictably, which does make things a bit easier). How do generators balance supply with demand? Through a whole company (National Grid has been privatised) and over 16,000 pages of dense legal obligations on top of normal laws and regulations.

              If you’re tempted to take it for granted, look at some other countries to see the results. I have a friend in Zambia who is plagued by what they know as ‘load-shedding’; rolling, unpredictable blackouts that can happen as much as every day for hours, because electricity supply and demand are not matched (and oversupply is dangerous too, as it can blow fuses or equipment).

              Just a little change can short-circuit the whole system. We ensure that suppliers always supply through definitive legal obligations and considerable amounts of government subsidy, planning and intervention. One could argue that some of this is unnecessary, but without some form of payment for availability as well as for electricity, we would not be able to match demand to supply. Availability is a commodity that has value.

              Except, of course, when it doesn’t. And that’s when it’s individuals’ availability. Low wage workers are not typically paid to enter zero-hours contracts.

It’s a default for workers to work between 9 and 5 (or in many jobs nowadays, 8.30 and 6 or longer….). It doesn’t matter if they’re a bit ill and would benefit from a lie-in or a shorter day; the world of work, including office work that isn’t shift-based (when the timing is important, for overseeing machinery or serving customers), demands that people work set hours. But individuals do not have a steady capacity for work. Illness, life events, home trouble, a neighbour’s party: a vast range of things can make the cost of 9-5 on one day feel like more than on another. Presenteeism encourages workers to turn up even if they feel less than 100%, instead of doing work when best able to. And if it’s a demanding job, then it requires that they do the additional work of managing the timing of their capacity. No late dinners with friends: work must start at 9, not 10 tomorrow. No social life. Got ill? Well, tough luck. Take some drugs and hope the stimulants mask the fact that you should be resting more than usual.

We treat workers as substitutable components; Lego bricks all the same size and purpose in any role. Workers aren’t automata who can perform at (the same) 100% for precise, unbending periods of time. They have more or less capacity at different times and ideally each individual life would have flexibility to do more at some times at less at others. If we require people to operate at peak capacity all the time, at the same times, we can expect the system (their life) to break down.

And by require, I include demands of career advancement and promotion. If we reward dedication then we will end up rewarding those with the free time to dedicate to work. Those who can spare most capacity for work will do best, and those burdened with care, housework, poverty and budgeting constraints, illness or just hobbies and creative interests will all lose out, and everyone will end up, as happens with companies, putting in as much time as they can in order to avoid falling behind and trusting to luck that life doesn’t unravel. There will be no free time for creativity; no spare capacity for events. Spare capacity is a source of social shame: what are you doing with your time? Why are you wasting it? And at the same time as our culture demands you give all your time, researchers are finding that having too much to deal with reduces your performance at everything. This is most obvious with the poor and unfortunate: even when they have time for something, the stress of the rest of their lives, and the intrusive worries about all their other problems, reduce their performance. They have gone past their capacity and negative feedback has begun to demonstrate fragility in their enforced lifestyle.

And then we discuss burnout and mid-life crises. By the time burnt-out people have hustled their way to depression they’ve already trodden others down. The people who knew that they wanted a good work-life balance all along have been elbowed out of the way by those who used their spare capacity and were fortunate enough not to need it for anything else… but now find it might have been nice to have it for something else. Everyone has lost, even those who had it right all along.

Spare capacity should be a point of pride and thankfulness.

The flip side of a set 40-hour week is that employers frequently expect office staff to work longer hours at peak times. As we can see in the electricity market, that sort of extra capacity is hugely valuable when we (the people) buy it, but it has become a hidden expectation that we (the people) don’t get paid for when giving it away.

There’s a reason that contractors and consultants charge a lot, and it’s not always just their over-inflated brand value. They allow flexibility: they are a whole segment of the economy that exists to provide flexibility in capacity for organisations. The downside is the overheads of contract management, performance monitoring and lack of trust; or lack contractual flexibility in outsourcing.

The drive for leanness in systems comes from a lack of trust. Managers don’t know when they’re being taken advantage of… and are driven to take as much as they can from employees. Cut things away and see if anything breaks would be a good idea if only effects were immediate and stresses constant. There is wisdom in the old design maxim that you achieve perfection not when there is nothing left to add, but nothing left to take away (yes, if I spent more time on writing, I could apply that here). But wisdom only. You need to know what actually could be taken away.

My father always used to like the solid, real-world example of a bridge. 97% on a test isn’t good enough: a 97% survival rate when crossing your bridge isn’t good enough. The same applies here. You can save some materials and still have a bridge stand up when you first build it, but if it collapses when the first storm wind blows, you haven’t been clever. And maybe it’s more expensive to put emergency supports up every time the wind blows…

What is leanness? It used to mean just being thin: the opposite of fat. And we get fat for a reason: to survive hard times. Weightlifters go through bulk and cut phases because they can’t grow quickly without also getting a bit fat: fat is an intrinsic part of growing strong, and is not just waste. Obesity is unhealthy, but a human with no fat is seriously ill. The same applies to organisations and systems. The clue is right there in the words that modern management jargon uses. We should learn from Mother Nature: in volatile times, spare capacity matters.

Choice trees have more than a trunk

              I hate modern algorithms in my life. The only time I notice them (and I’m sure there are a lot of good ones I never notice) are when they’re limiting my choices in the name of personalisation. When I visit YouTube, for example, I’m flooded with junk I don’t want to watch. How do they regard this ability to hide interesting videos as an asset? I see the same on Netflix or Amazon. Amazon shows me the exact same thing I’ve just bought, or others like it. How many boxes of marzipan should I give my sister as a Christmas present? 1 or 20? Netflix tries to offer similar things in categories, but the categories are clogged up with videos I have already seen: it thinks that these are more relevant to my watching habits that something entirely novel. I’m trapped in a bubble: my last video looked like the one before it, so I’m offered the chance to watch the one I watched before, and then re-suggested the second one, ad infinitum.

              So I do. I re-watch good things and bad. I watch one marble run video on YouTube, because I really wanted one when I was a child, and then my homepage is flooded with dozens of marble races (strangely entrancing that they are). The default setting is too tiresome for me to change. I wouldn’t even know how to search for something interesting.

              Defaults affect us the world over. Someone offers you a free trial, but then wants your bank details. I remember that the first time this happened to me I was shocked. “Why do you need bank details if it’s free?”, I asked. So that they can set up a subscription and it’ll run unless I cancel. I ran away from that; I didn’t want a subscription. I wanted a free trial. My youthful naivety is long gone now: we’re all accustomed to such sales tactics. It’s just the way world works. But is it the best way?

              Imagine all the people having money and time leeched from them because they’ve lost track of all these defaults that they need to change… changes that are made deliberately difficult. I’ve complained to numerous companies over the years: van rentals, cancelled flights etc. No complaints process was easy. I can send the complaint, but then I get a robotic reply that doesn’t address the problem… we have to correspond, time goes by, nothing happens, I have to pursue it and forget… and then my complaint has gone away. No refund, bad service justified.

              This is just another way that people try to control our decisions. It’s not advice or nudging: it’s outright manipulation.

              So what would be better? Well, if we return to Amazon or YouTube, it’s easy to imagine what better would look like. Amazon should know when I’ve ordered something and remove it and similar items from display. Maybe it works this way to make money: lots of gullible idiots buy another copy, just in case, or because they forgot they’d already bought it. Or in some rare cases, because they genuinely could use a similar item at the same time. And then, just when Amazon forgets I ordered something, it should be remembering: when I’ve had a chance to use up, break or lose the last order, or decide it wasn’t good enough. That would be helpful. And if I don’t want something, it’d be lovely to tell Amazon to forget that I ever even looked at something like that: to remove the high ranking it has for me.

              I need something similar on YouTube. The flood of videos about X-box games because I once made a mistake and watched a video I thought was a general comment about computer games was a waste of space. I don’t own an X-Box. But I watched another one from that channel, so I must be interested! Well, of course I watched another one. I was looking for some mild diversion while chopping vegetables, and that was all YouTube was showing me. I need a way of deactivating the ranking for those videos: of down-rating them, to say ‘that was just noise, don’t treat it as hugely meaningful signal’. But without saying ‘this is something I would never watch: never show me this again’.

              What I need, in fact, is a tree structure. Broad categories, all available, which open into subcategories, like a folder system on an old computer. Only because we’re modern, we can use metadata and have some things show up in multiple places, but maybe with different rankings. An item might be 100% comedy, 50% news. So when I go to news, I might find pure news sources first, but see the option of a satirical report.

              I don’t want a helpful machine to learn that when I look for news, I really mean comedy. Maybe I really mean news but nothing was interesting enough, or maybe I’m trying to be more informed but the comedy seemed diverting enough this time. In essence, what I say is what my thinking, conscious brain has decided, and what I do is this decision, corrupted by temptation, emotions and the limited options available. If algorithms learn to bypass my thinking free will and feed the beast of my emotional, whimsical self, then they are not helping me: they are destroying what makes me ‘me’.

              Netflix does the same. I can’t simply say ‘reduce the ranking of things I’ve watched by 90%’, or ‘nothing I’ve watched before should be visible before scrolling’. Instead, 90% of my screen is full of things I’ve seen, helpfully displayed as most similar to the things I’ve watched. I assume that this is to hide the lack of content, and because they have found that when faced with the choice of re-watching something or spending 30 minutes searching for something new, people who have decided to watch Netflix rewatch something. So Netflix gets viewers, and can tell itself that it is adding value to their lives.

              But there is no option to customize for myself: to write my own algorithm, to tailor it to reduce temptation, to help me live the way I want. I must submit to someone else’s choices; to the tyranny of how they choose to interpret my history and desires, all based on the assumption that their calculations can understand me better than I understand myself, and that I will never change nor want to change. The only path is deeper into the rabbit hole.

              That’s how defaults are set up at the moment: traps for the unwary. We must devote cognitive attention to escaping them, and if we do escape, it’s usually extreme: complete removal or deletion of the option.

              Humans don’t work like that. There is noise in our decisions, and gradual shifts in our choices. I might watch one thing because it seems like an appropriate diversion right now; or because a friend is visiting and chose it; or because it was talked about at work, or any of a range of reasons. I might want to watch one or two on a subject, but not hundreds; and I might also want to watch things on other subjects. Some things are only just acceptable; others are brilliant. But the binary information of whether I do it or not doesn't reflect that. I'd do brilliant things more, but sadly they are rare. So I drop down to mediocre things. If that's a sale for whichever salesman or company I get it from, that's fine for them. More of that, please: get me on something that's good enough, and save time and effort that might have been wasted on getting me something great. Soon I'll come to accept that good enough is all I can get.

              I can’t let this discussion go without Facebook, which thinks its algorithm is so hugely valuable. Sure, it might help serve advertisements, but it turned Facebook from a brilliant way to see what every friend I had had done recently into a mediocre way to see what the friends I spend most time talking to had done recently… as if I needed that help.

              It still passes my threshold for use: nowhere else has any of my friends’ activities and thoughts displayed for me to engage with and maintain some social contact. But even when I turn this much-vaunted algorithm off, it looks very much like 480 of my 500 or so friends never do anything, until I look directly at their page. That’s not how I want to arrange my life; limited to as few as possible. I want to pick my interests myself.

              Maybe if I repeatedly click through a sequence of folders, I might want a shortcut to develop, as happens at work: Windows gives me five recently used folders, but only uses ones that I use at least a few times. They’ve realized that there’s noise in my activity. My neurons do the same thing: connections are strengthened or weakened over time, not created or severed.

              But even if algorithms learn to adapt properly: to deal with my interests as a network, with hub subjects and spoke subjects, all of which I want to be available, rather than the hub drowning out the spokes I already had, let alone potential new ones; even then, they can’t replace free choice. Humans thrive on a wonderful combination of self-directed spontaneity. Shops are wonderful because we see things we never thought about. That’s why people love Tiger or Pod, or department stores: for the novel nicknacks that ruin the environment but stimulate the mind. Algorithms, as they deal with me, would take my passing glance at a novelty pencil and teleport me to a pencil shop, smugly proclaiming that they were being helpful.

              The same is true in creativity. We need those infamous water-cooler moments where two different ideas collide that do not usually meet. How can an algorithm, based on history, ever hope to create such truly stimulating and productive moments? The actions of chance created evolution and the remarkable panoply of species that we’re killing off today. Chance moments spark people’s minds, bringing meaning and progress. Our choice systems are instead designed to create ruts and dig us deeper into them. It would be better to help us never dig a rut. Maybe when I’ve watched things, or bought things, it would support me to deliberately show me something new!

              The people who help us most are civil servants. Because they are rightly careful not to be illiberal, every nudge comes with requirements to give information and the right to opt out. We can opt out of pension schemes, but we don’t. Evaluating them and then filling in the paperwork is too much of a burden. Most of us can’t spare the time (and we all know that the administration of life has been gradually outsourced to the individual, while working hours and the demands on our time have increased).

              In summary, algorithms are designed to dig us deeper into whatever hole we start digging for ourselves, because it looks to be more profitable to keep us hooked on whatever just about passed our threshold for interest in the first place than to help us out of holes to search for new places to explore by ourselves. They contract our worldview to whatever is easiest to give us that we still accept using. They do not expand our horizons, helping us see further and better and to be more discerning. They make us small-minded, like the idiot programmers (or PR teams) who call them progress.We should be creating tools to make us better, as Steve Jobs said he did.