Ageing: what is dignity?

 

The phrase ‘growing old gracefully’ is well-established. It means not desperately clinging on to something that, to everyone else, is obviously gone. I find it strange. Since when has grace had anything to do with going with the flow; with meek acceptance?

              We mostly mean pleasing or attractive movement when say something is graceful nowadays, but this is a metaphorical meaning: with charm, perhaps? Or willingness? In a pleasing way?

              The dictionary isn’t much help, but we all intuitively know what is meant. On the surface it makes sense: dwelling on the past and not facing the unpleasant present is a sort of mental illness or cowardice. That’s why British patriots are scornful of Brexiters who hark back to supposed glory days of Empire: we do not have an empire now. We cannot behave as if we do without it looking like the senile ramblings of a demented grandpa.

              I dispute that approach to growing old. Ageing is a curse: life’s greatest horror, perhaps. The slow, inexorable doom that tortures you with gradual deterioration, always reminding you that life can only get worse; that your time is limited and your value depleted.

              One old Christian tradition says that we should accept our fate with meekness. God (who doesn’t exist, in case anyone was unsure) has ordained our fate, and His will is for the best, by definition, so we must accept everything and trust that we will be rewarded for our faith in the future. It’s the line of con-men everywhere: trust me, blindly follow, don’t worry because everything will work out, you’ll get all the benefits in the future. Again, rather like Brexit.

              Because of the way religions work, Christianity also has a more stern, heroic character to it. Something that can inspire people when that’s required, saving the meekness for when the masses must acquiesce to their rulers rather than when they must be inspired against their rulers’ enemies.

              The human mind can see both points of view. Hamlet, when discussing suicide, pondered either side of the problem: ‘whether ‘tis nobler…’ to endure troubles or take up arms against them. A very different context for a very different time, but a similar discussion.

              So, is it attractive or pleasing (i.e. graceful) to give in to ageing? Is dying one’s hair or still going out and having a good time graceless? It depends on whom one wants to please. Some people are wedded to tradition or never really thought that these things were worthwhile anyway. The phrase ‘grow old gracefully’ hides their bias about what is right and proper in life. What they really mean is ‘I always found being young distasteful’.

              There is an undercurrent of religious bias here. Whether consciously or subconsciously, people value settling down, not rocking the boat and doing the same as everyone else. Some people always looked forward to settling down and thought youth was just a phase before real life began. Good for them: what they like is perfectly fine. Should they impose it on others?

             

              So is it obviously better to acquiesce to the fading of one’s vigour, health and lifestyle? Ageing is a horrific curse: to see your own deterioration and live on, knowing that you will never be as able as you once were. It is fading away of what is precious: if not precious, then why do young people  not deliberately hobble themselves or buy wrinkling creams to make their skin look older?

 

Is life a boon?

If so, it must befall

That Death, whene’er he call,

Must call too soon.

- William Schwenck Gilbert

Ageing is the ultimate indignity, robbing noble and good people of their abilities, hopes and dreams. It is not a friend to be welcomed. Is it then an unbeatable enemy with whom it is better to make peace?

Of course not. There is nothing nice or good about being overwhelmed by pressure. Wrong is wrong, however uncontrollable. Milton’s devil makes clear that it is better to struggle against overwhelming force if it is wrong, even in the face of certain defeat. To stand up for one’s principles is heroic, not graceless. A good person shouldn’t go gently, but should ‘rage, rage against the dying of the light’.

              You might disagree. But let’s be clear about it: this isn’t about unseemly attempts to escape a just punishment. There is nothing just or right about ageing. This ‘graceless’ behaviour is about a refusal to let a curse take away your life; a refusal to do ageing’s work for it by fading away early.

              Every advance humanity has ever made, every triumph over adversity or injustice, has been because of people fighting against it. We do not regard Nazi collaborators as graceful or worthy. At best, we sympathise with the scary and precarious situation they found themselves in and understand their choices without condoning them.

              The idea of ageing gracefully is ludicrous. Meek acceptance is only a virtue for the oppressors, and only in the oppressed. What you gain in dignity through never showing physical defeat you lose from the moral defeat you have inflicted on yourself.

Malgorithms

 

Imagine an algorithm trying to be a restaurant waiter:

“Good evening, would you like to start with a starter?”

…

“Now that you’ve ordered a starter, would you like to re-order your starter, or try another one of our starter range?”

…

“Welcome back! Since you tried that starter last time, how about trying it again this time?”

              Has no-one in the technology sector ever considered the value of variety?

A.I. has been the hot new thing for a few decades now, with commentators often praising the geniuses who write such profitable algorithms, and occasionally criticising them.

Let me be clear for those new to the subject: the algorithms that are the basis of the world’s biggest technology companies are not A.I. They are computational statistics; just a lot of different datasets and computers powerful enough to test many possibilities and find correlations between results. Intelligence requires insight; the ability to understand mechanisms, form expectations about correlations and dismiss coincidences, interpret results and predict outcomes before testing begins is intelligence. Machines that do these things are still complete fiction.

Algorithms are not artificial intelligence: they are a tool to amplify human stupidity (or intelligence, if any can be found). It should be simple to do it better. Yet no-one does. Why, I can’t say. I can guess.

 

Examples of problems

Amazon

Now one of the biggest companies in the world and famed for mistreating its low-paid employees, everyone seems to be familiar with some of the Amazon website’s failings. Why would anyone want to buy exactly the same product they just bought… especially when they have a choice to buy multiple copies the first time round?

I have bought a mouse and been offered another one; a computer, and been offered another one; toothpaste, and been offered more… until a few months later when I’ve used up that batch, when Amazon has forgotten that I ‘like’ toothpaste products. I’ve looked at products, decided that I don’t want them, and had Amazon suggest them to me again.

Amazon recommends junk, is always trying to take extra money from me by tricking me into a PRIME subscription with a tiny little ‘skip’ button hidden somewhere to get normal delivery, and yet has such limited search that I had to scroll through endless bins noting the dimensions myself to find ones I liked that would also fit in the space in my kitchen.

 

Facebook

My world has shrunk over the last few years. My life is smaller and poorer because I relied on Facebook as a tool of choice.

 I originally found Facebook a fun tool for staying in touch with acquaintances: I could see what they posted and share my own life without the awkwardness of imposing on them, or the forwardness of e-mailing directly. It was a fantastic innovation, and one for which we should thank the inventors Zuckerberg stole it from.

However, Facebook has gradually decided that it isn’t a networking tool, but an anti-network tool. Acquaintances I liked reading about have disappeared; some of them still post on Facebook, but I never see that. I have no social network any more. The hundreds of friends I’ve accumulated from different lives and interests have shrunk to maybe half a dozen. Everything else, if I scroll for five minutes or fifty, is junk. Much of it is repeated, as if I’d rather read the thing I deliberately avoided 5 seconds ago than see something interesting.

I didn’t join a friendship network to follow celebrity or corporate content, but even Facebook’s ‘chronological’ option in the ‘news’ feed is not a chronological list of friends’ activity. It is the usual curated feed with the junk I’m not interested in re-ordered. Facebook refuses to show me the activity of the friends who are the reason I’m on Facebook.

I am not surprised that it is haemorrhaging users. It no longer serves any purpose even in my life, one of its more enthusiastic initial converts due to my social awkwardness at using other means of connection.

 

Netflix

This has rankled for ages. Netflix is a subscription service. It shouldn’t suffer from any competing incentives to providing me with the best value. The Netflix algorithm uses the bland categories of Blockbuster, a company that it bankrupted, and refuses to allow me to block things I’ve already watched. My whole opening screen is now things I have seen; it is the work of some minutes to search through the junk I’ll never watch for anything new. Their filing system is a disgrace. Just think of the advanced computational statistics you could run with modern computing: what part of a programme does this individual stop at? Is that gory, horrific, excruciatingly embarrassing, dull… could you write a programme to judge that? Did the person pick the show up again, or give up? How can we help this person get more of what he wants?

Instead, Netflix has gone with the easy approach of how to keep someone addicted with what their customers will be satisfied with – judging by the similarity of their approach to Amazon or YouTube. Maybe that’s just what the experts know how to do, having worked at those other companies.

If it stops me hiding shows then I won’t see how thin its offering is: I still don’t see things I want to watch, but I’m left with the possibility that there’s something new hidden amongst all the junk, making Netflix seem like a well-stocked provider.

 

YouTube

I grew up with dogs and recently watched a YouTube video of some working dogs ratting at a farm. I’d never seen dogs hunting before (except African wild dogs on a BBC nature documentary) and I was curious. YouTube decreed that I had an interest in rats! I’ve watched a few puzzle presentations, so now I have calculus as one of my suggested subjects. I watched a clip of a film scene I liked, so another suggestion is ‘fandango movieclips’.

There might be skilled programmers writing code in distant servers that allows Google’s almighty machines to crunch correlations, but if there’s any insight or intelligence in what I’ve seen, it’s hard to find. How could anyone be interested in ‘Fandango movieclips’? Why would anyone assume that I care about film clips in general? And if I watch the final scene of ‘The Good, the Bad and the Ugly’, why would I want to watch it again immediately afterwards on a different channel?

Sure, it might be a channel that posts clips of well-liked films, but it’s hardly an interest is it? Not like, say, stamp-collecting or painting.

YouTube is the biggest in my personal litany of computational statistics failures. I started by listening to music on YouTube, just to try to work out if I liked classical music enough to buy any. From there I found some occasional videos on other subjects. The videos are often diverting, but rarely truly valuable viewing – the same problem as everywhere else. Diverting is good enough for online engagement, but doesn’t actually improve lives. It steals them away. But viewing is easy to measure, so diverting enough to watch is what we get; genuine value is hard for a computer to measure and so no-one worries about it.

I have recently watched quite a few videos about computer games, and now I see endless videos repeating the same content at me. If I watch one video about the (genuinely) exciting release of Total War: Warhammer III, then I must want to watch every single commentator’s take on it, because they are all similar content.

So far, so familiar. But alongside these wastes of screen space is an insidious new one. I watched a few film reviews, and noticed that the reviewer has a thing against preachy, poor writing. Another online reviewer also has a thing against preachy, poor writing, but YouTube has combined my viewings of an anti-woke film reviewer with computer gaming and now wants me to watch Jordan Peterson, Laurence Fox and ‘Moron DESTROYS feminist!!11!’ videos. Facebook thinks I might like nutjob ideas from Nigel Farage or Turning Point.

 

The overall effect

The path is predetermined; I cannot choose my own way. If I watch one gaming video, I am compelled to watch them all. And once I’ve watched them all, I must be radicalised, shoehorned into the neat category that YouTube has chosen for me. My individuality, if I watch much more of this, will be erased. Strange that the lunatic Right, which cares so much about individuality and not being a sheeple, only exists because it can herd its recruits like livestock, using advertising and malgorithms.

Anything that might have been good in these online treasure-troves is buried beneath the malgorithm that serves as hellish gatekeeper, confounding and redirecting me from what I really want. I am in a constant state of warfare with an invisible enemy, yet all over the news I can read celebrations of ‘artificial intelligence’.

However clever the programming (and statistics, if any), this fundamentally misunderstands what benefits humanity. Maybe it’s good for distraction: the ‘content that’s just good enough to distract someone for a few more moments’ approach, or maybe ‘just enough to fool someone into a purchase at this moment’. But it’s not adding value to people’s lives. At best, the economists’ ‘consumer surplus’, which is the extra value the customers get above their purchase price, is minimised to 0. At worst, it’s taking advantage of people’s weaknesses to give less than no benefit.

 

As far as I can tell, the algorithms I encounter have two mechanisms: what connections have already been made (i.e. what content is consumed by other people who consumed this content) and offering more that is exactly like what has just been shown. There is no clever AI, no insight or genius: just data processing. I can get better advice from a 6-yr old.

 

The answers

The cause

I can see only two plausible options for why AI is so stupid: either 

i) the programmers are too caught up in the complicated mathematics and computing to think more broadly about what they’re actually doing (i.e. they’re not the geniuses some commentators like to say); or

ii)                           ii) there is more profit in having bad algorithms.

There’ll be some of each, varying from problem to problem, but in my experience it’s very easy to run even big and respected organisations without real insight, innovation or thoughtfulness. From the risk management failures that caused the financial crisis through to things I have directly observed, there is more than enough evidence to justify assuming institutional stupidity as a default explanation.

Of course, where there’s an obvious financial incentive, or documented evidence, we shouldn’t dismiss greed: the other thing that large organisations achieve is to separate bad consequences from the decision-makers, freeing them responsibility and guilt.  

              These two things are not entirely separate; trusting, obeying or creating a system frees people from the need to consider each decision, whether to work to invent a new answer or to choose a moral one. And organisations necessarily involve systems of control, communication and interaction, or they are not organisations but disorganisations. Anyway, that’s a bit off-topic… and yet system creation is rather relevant to a discussion of algorithms.

              Personally, I think it’s obvious that programmers were not geniuses who knew all the consequences and outcomes of their work beforehand. They merely had to be clever enough to do the work, not to understand all its implications too. But repeated testing and observation allows them to wilfully or thoughtlessly recreate what others chanced upon.

The idea of the wisdom of crowds, thoughts of market forces and simple unthinking hopefulness all combine to make people stick with assumptions and methods that merit more consideration. If something achieves results, it must be right. That’s why Newtonian physics has never been superseded by quantum mechanics. There is no improvement needed on the success we already have.

 

1.      Variety and radicalisation

Innovation, learning and wisdom come from exposure to new things. Humans need to expand their minds; those who find their mental worlds confined often turn to drugs for such expansion. And too much stagnation leads not only to depression but to senility. The malgorithms do the exact opposite of what humans need: they give us more of what we’ve already had.

If I have bought an item, I will be funnelled into more such items. What was a passing interest will become a life-consuming myopic tunnel – if I was tempted into giving you profits once by such content, I can be tempted again, if you bombard me enough until you catch me at a weak moment.

The only way I can escape these myopic tunnels is by being so temptable and free with my money/eyeballs that I’ll respond to whatever content I am shown, creating such a wide array of ‘interests’ that tunnelling is temporarily stymied (although it will always be the most stable outcome of the system). Even in this case my interests are still not my own: they become random, controlled by the great malgorithm in the sky once again, and this time hugely profitable, because the malgorithm can offer guaranteed profits from me to whoever pays for its promotion.

Innovation comes from linking disparate ideas and concepts together. The tech sector itself will tell you that fulfilment comes from solving problems; from creating your own answers or artworks. You will be pushed towards creativity by seeing myriad different things, not by being given exactly the same thing, either as a direct repeat (Netflix and Amazon), something that does the same thing (YouTube and Amazon) or something that’s next on a path that others have already beaten (YouTube and Amazon again).

Novelty is key to human flourishing. We know that the best way to learn is to be challenged just within your ability. Things need to be recognisable, but new. Algorithms leap on any expressed preference (of which more later…) and convert it into your only preference. Imagine a food algorithm: ‘you bought chocolate just now, so here’s a meal of chocolate. Ooh, that tempted you yesterday, so a meal of chocolate will tempt you today. Others who buy chocolate also buy cakes, so why not have a cake for pudding?’ And then a rash of news articles praising the geniuses who have written code to give people what they want, plus a few opinion pieces controversially suggesting that the population might be getting unhealthier…

When most of us eat food, we do not want that type of food for a while later: we have a refractory period for consumption of a particular product. The same for sex: once someone has had sex a couple of times, he needs a few minutes to recover. The same applies to viewing habits, or shopping habits: it’s not beyond the wit of man to note a preference, but only act on it later, when it might be helpful. Could a clever computer work out how long a tube of toothpaste might last - or even how long it lasted for that individual – and only suggest a repeat purchase that much time later? Could one note a gaming news/opinion video and only suggest another one when there’s a chance there’d be something new to say? There’s a product refractory period and the individual’s refractory period for that product… a whole new world of variables to calculate and feed back into an algorithm.

In the meantime, show us something new. Things distantly linked to our viewing: leap ten links down a chain if it’s too hard for a computer to grasp the concept of a video or product and suggest something intelligently. Instead of ‘people who bought this looked at this’, work through a chain: ‘people who looked at your thing also looked at content B, and people who looked at that also looked at C…. so here is item J’. If 10 steps doesn’t work, experiment, but with a weighting against lower links. Think of creating whole new markets for products, rather than simply strip-mining anything that looks like one.

I can search for very niche content. I don’t need AI crunching my data for that. AI should improve on search results rather than simply repeat them. AI often seems to me to be worse than the search box. I get to type things into the search box; AI has chosen what to type for me. The joy of going into a shop is seeing things I wouldn’t think to look for. I’m already capable of searching for what I already like. AI often removes that ‘shop joy’ – Facebook and YouTube being the biggest offenders. Amazon, amazingly enough, looks to have done some work to keep a little of it.

In sum, exaggerating expressed preference (I really want to emphasise that’s just a name for it: we could have a long argument about how much preference is expressed…) makes the system unstable with equilibrium at complete radicalisation (i.e. just one interest consumed all the time). Most people do not get that far because their usage of the algorithm is limited; because there are occasional, random additional inputs; and because even relaxed users eventually rebel against the mental confinement of the algorithm.

Imagine a programmer who put limits on what his algorithm would radicalise: time delays on repeat content, limits on the amount of display space dedicated to any one category of content, random content great leaps of connections away from the user’s previous usage… someone who would create a programme to help every individual reach their most preferred content and yet also help keep them healthy and alert by never getting stuck in a rut.

2.2.      Customisation

On the subject of finding ways to show us new things that aren’t entirely random, an algorithm for each individual makes more sense than one for the whole population, trying to treat us as all the same. How did supposedly intellectual people forget to include individual (meta-) preferences as a variable in their computational statistics? Is it beyond the wit of man to allow individuals to customise their preferences? Not in the Facebook manner of ‘news feed’ or ‘news feed [with a different name]’, or by selecting individual items and asking it to show more or less of them, without knowing how they’ve been categorised, but by choosing more complex sets of criteria?

Just imagine what you might learn about humanity if you gave people conscious control over what they were shown, rather than relying on revealed preferences! What if the programmers’ interpretation of what is revealed is itself wrong? If I watch a video someone placed in front of me, did I like it? Do I like it as much as one I sought out for myself? If I don’t engage with someone, is it because I dislike them? Or is it possible I’m not just a simpering moron thoughtlessly reacting to my passing whims: that I am giving that person space, not crowding them, worried about what message I will send them, not sure if it’d feel weird to invade something that might be meant for closer friends… talk about American bias! I know that Americans are taught to be outgoing at all costs, that being shy is practically evil and being pushy is being a balanced adult, but even many Americans manage to reject this weird propaganda, and the rest of us prefer to laugh at it than mimic it.

Imagine how much I would rely on an algorithm that supported my attempts to better myself rather than undermined them: one that allowed me to ask to stay in touch with friends and showed me posts from people I had NOT seen or interacted with for a while! How much joy would that bring to isolated people, trapped in lockdowns?

The difference between the customisation I envisage and search engines is that most complex systems have sensitivities or assumptions fed in. A population average fed into an algorithm might work reasonably well, but it might be even better if users could change those inputs to suit themselves. Show me more variety; place little weight on low numbers of interactions with content as I’m just testing it out; limit the amount of this type of content because I want help with self-control rather than attacks on it… I might genuinely only be interested in so much of a certain type of content: if a company knows when I reach my limit and offers new things at that point, it’s much more likely to keep my interest.

Trust the users. Rather than only ask them for data points and trust you to join the dots however you choose, allow them to suggest shapes. Give them a tool to tailor to enhance their lives rather than one that tries to force them to be like everyone else and never change. Using averages makes it easier to be average: deviations from what the algorithm expects are not catered for. Perhaps there are large tails of the distribution who could be served so much better?

People were probably initially thrilled to find any sort of meaning from population data; that itself seemed brilliant. And the algorithm’s self-referential ‘vicious cycle’ nature (see later) ensures that small signals become large ones, so no-one ever needed to think about improving the initial signal.

3.      Noise

Anyone who has ever done any analysis of anything, let alone real sound engineering, knows about noise. Is something a signal, or is it noise? Has the extra-sensitive algorithm detected a revealed preference from my behaviour, or was it mere chance that I did those things, and next time I’ll want to do something else? If next time the algorithm shows me the same things, that’s a response to my signal and we’ll never know whether the signal was true or not. The lack of any consideration of noise in all these algorithms is detestable, and yet understandable. They want to find chinks in our armour of self-control. They can leap on a potential weakness, and if they fail, leap on the next one without worry: it costs them nothing to try.

That’s the only way to understand the way an algorithm can decide I like rats after watching one video of working dogs; or decide that I like jigsaw puzzles after looking at just one puzzle (people only have one birthday a year, and I’m hardly going to buy puzzles for years in a row). But it would be more accurate to wait. Intelligent programming, rather than brute force attacks on users, would calculate the criteria for judging whether something was a thought for a birthday present or a burgeoning hobby.

Just think how much revenue you might get from me if you could show me things a friend might like in the few weeks before his birthday, rather than showing me things that might be good for the friend who just had his birthday!

What if a ‘like’ on Facebook didn’t mean ‘give me more interactions with the person who posted this content’, but had intelligent analysis behind it, trying to place it into one or more of, say, the following categories: ‘this is just a random interaction’; ‘I enjoy things on this subject no matter who posts it’; ‘I am interested in many things this person has to say’; and ‘I enjoyed this particular instance of this sort of content because it was better than others’.

If Facebook ever had anyone trying such a thing, he must not have got very far. GIGO is the greatest rule of statistics, or any analysis: ‘garbage in, garbage out’. I have always used my ‘likes’ sparingly, being far too laconic and shy for the modern bare-all world. For me they were a way to signal quality, the last option on my list: ‘I enjoyed this particular instance of this sort of content because it was better than others’.

Programmers need to find a way to weight every single interaction that is currently used merely to contribute to ‘show more of our categorisation’ as contributing varyingly to my categories above. And leave dummy variables open in case their categorisations are not the ones I am using. For example, when I pick my viewing, I can very simply define my television choices as a) a preference for action and adventure of any sort; b) a firm veto of anything that uses cringeworthiness (for drama or comedy) and c) decent scriptwriting without massive plotholes or idiots as protagonists. I am surprisingly accurate at judging (b) and (c) from a mere written description, let alone a short clip. Netflix cannot comprehend (b) or (c) and so its attempts to find content for me are doomed.

 

4.      Quality

If technology companies have a notion of quality, it seems to be only to get people hooked: show the most popular videos, and then show more of exactly the same thing once someone has watched them. I doubt that they have a separate rating from ‘views’ or sales.

What if I only want to watch quality videos, and only one or two on any one subject?

We could try to work out if quality could be teased out from popularity, especially given the runaway nature of popularity, which often feeds more on itself than intrinsic quality. Are there some people, like me, who give a better signal of quality by rarely interacting with things? How do you judge quality before you have many likes, and once you have some data, how do you taper your analysis into ever greater certainty as more and more data accumulate?

I would start with measuring quality by the level of interaction by people not usually interested in that content. Or perhaps by finished viewings. Amazon can use reviews, perhaps noting the proportion of total purchases who review items, but YouTube, Facebook and the like are stuck with ‘likes’, a very poor measure of quality.

Dating apps have ‘superlikes’. These have limited supply and are intended to make users spend on buying more, as they also circumvent some barriers to matching. But the limited supply should also make them a better signal of attractiveness.

Clever data analysts would weight ‘likes’ more from those who use less and are less interested in the content. A like from someone who gives out two a day should be 500 times as valuable as from someone who gives out 1,000. Although I can imagine that perhaps first exposure to some content might generate a like no matter how good it is. For example, someone might like an anti-vaccination video because it seems to share important new information no matter whether it was well-produced or a basement rant. And, in fact, more persuasive anti-vaccination content is actually worse content, as it better promotes harmful lies.

Can you measure truth? Probably not without any human assessment. But you could rate truth values of some content, and then check who likes truthful content but not false content, and give their ‘likes’ or interactions high value in automated judgements. You could multiply your inputs.

There are plenty of places to start. I’ve seen no evidence anyone has even tried!

If people are still not convinced that popularity is a poor measure of quality, may I invoke the Nazis, as every online debate must always do? They were popular, but not good.

What should algorithms do with a measure of quality? That’s another interesting question. Established quality might drown out new, unrated content. But I doubt tech companies really care about such high-level, long-term side effects. Quality might be essential for a healthy society. Quality might help drive user engagement. After all, if you really can measure it well, then you don’t have to avoid any addictive or otherwise dodgy practices; you can simply do all the usual malpractice but with higher-quality content!

Good content isn’t actually needed. Popularity is actually what is required to keep users engaged; good content was simply how people used to try to be popular. If popularity can be achieved some other way (c.f. the Conservative party and Donald Trump) then quality is entirely irrelevant to the business in hand, despite it being far more beneficial for people’s lives. But imagine if you could have both!

              The incentive not to care about quality comes from problems with acting on that judgement. If even a derived variable in an algorithm looks like a judgement of quality, it becomes harder, mentally, to claim not to be a publisher. This fear might be one of PR rather than reality, as using popularity is just as much a judgement of quality as any more complex calculation: it’s merely a very poor one.

 

5.      Feedback

The feedback loop of algorithmic control ruins the data, rather than reinforcing it. If I exert control over something, I can no longer treat its new state as the natural state; as raw data. If I chop a tree down, I can’t conclude that it grows horizontally. These feedback loops exist elsewhere too: in the stock markets, where the ‘wisdom’ of crowds is itself a signal, and all that matters is outperforming the crowd, the feedback loop drowns out any price discovery. In any popularity contest, across friendship, politics, media influencers, writing or music, success begets more success: people are drawn to, and more likely to hear from, those who are already popular.

Yet algorithms are built on the strange idea that if people respond to the algorithm then they must have been interested in that to begin with. At the same time as trumpeting the power of algorithms to persuade people (to advertisers), the data analysts are assuming that no power was exerted and the results are all innate preference.

Imagine you’ve just bought a new computer (and maybe you’ve moved to a new house and new ISP and whatever else companies measure). The algorithm doesn’t know you yet. You’re at the top of a hill with a whole world to explore. What direction do you choose? Some people know where they want to go: they have one hobby and it’s all-consuming. But others might prefer to explore. Tough luck: once your ball starts rolling downhill, that’s it. The algorithm will whizz you into the deepest rut in that direction and keep you there. Did you like computer gaming? Well, now you hate women and want to massacre school students in America. If you want to see what’s in the other direction, or even in the next valley over; if perhaps there’s a mine of information elsewhere that you might prefer, tough luck. You have to work your way out of the hole yourself. The malgorithm will try to feed you misogyny, over-eager to guess where you’re going, desperate to push you into the rut, even though the nature of ruts in the road is that you don’t want to get stuck in the mud everyone else has created. Your first tiny push and willingness to explore has been turned into a ‘signal’ that you want to wallow in filth.

It's an untenable position. Knowing humanity well, and these companies a little, I feel confident asserting that very few people within them have even thought about this. It’s a fundamental logical inconsistency that has got lost in the rush to complete discrete subroutines or ‘improvements’.

But one could, if feeling more charitable, assume that the controlling ‘intelligence’ in the companies is a little intelligent. Perhaps they know that exerting power changes the environment and that therefore their conclusions about people’s preferences are unfounded. If that’s the case, there is only one possibility remaining: that they are evil, exactly as the backlash against the tech sector claims. They know that they are distorting, rather than discovering, people’s preferences, and continue down that path because it is profitable.

              Maybe they believe that distorting people’s preferences: bombarding them in the hope of a moment of weakness, or showing them content that is just distracting enough, rather than as worthwhile as possible, is more profitable than adding value to people’s lives by discovering and servicing their real preferences. I think that’s an unproven assumption.

              This is what most advertising is: it is about creating new preferences. But it’s a strangely self-consuming world in which these content-providers steal more and more of your life through distorting your preferences in order to show you more adverts for other things which want to fit into your dwindling life! The advertising platforms are crowding out the products that want to use them.

              But if it really is true that people’s real, consistent preferences, in the most virtuous versions of themselves that they aspire to be, are for less consumption, it says a lot about our world that the companies promoting global destruction through massive consumption – through encouraging people to be the impulsive, lesser versions of themselves - are valued so highly, and praised as the height of innovation. If, on the other hand, people do want to waste their lives and the planet, unguided by malign influences, perhaps nudging us to be better, rather than indulging our weaknesses, would be the right thing to do.