THEY’RE NOT THAT INNOCENT

Oops! They Did It Again. And Again. And Again.

Unintended and perverse consequences in recent Medium policy decisions — a peer-reviewed academic study*

*Disclaimer: not actually peer-reviewed. Or academic. Or a study. Other than that, every word of the subtitle is true.

Odd way to start an argument, perhaps. But in this case it makes sense. Because it highlights the fundamental issue that my other points rely on: people lie. They bullshit. They say and do things not because they are right or true, but to gain an advantage of some kind.

They are genetically programmed to do it, and they are constantly incentivised to do it by systems with rules that are poorly conceived, weakly implemented, or usually both.

There is a rock-solid solution that would eliminate such unpleasantness at a stroke. Millions of them, in fact. But none has yet followed a trajectory that would result in a direct impact on the planet we temporarily call ‘home’. So in the meantime, we’re left constantly scrubbing the shit stains from the rich and varied tapestry of human existence in this fast-paced world of technology and content generation.

And it seems to me that certain decisions taken by Medium since I’ve been delving into its fascinating world, have resulted in precisely the kind of unintended consequences that undermine their own stated or implicit aims.

I say ‘unintended’ rather than ‘unforeseen’, because I think that in many cases they could absolutely have been foreseen with a little more psychological insight. My consultancy services are available for a very reasonable monthly retainer, but for cheapskates, here’s the free TL;DR executive summary: people only care about themselves.

So how are such human failings reflected in the real-world problems of Medium’s perverse pivots?

1. The Boost — prioritising quality content via publications

That’s the idea, right? We raise the standard above the level of a free-for-all what-I-had-for-breakfast blogging platform by establishing certain premier publications which only contain high-quality content.

Kind of like a return to the professional flagship titles the site had before it pivoted away from that model last time.

I know that’s not the stated policy, but it’s a logical inference. If you handpick certain publications, and tell writers ‘these are the places to go for the best rewards’, while funnelling their output to readers via the algo feed, that is exactly what you are doing.

But this aim of selection and consolidation, of feeding the big sharks, undermines itself.

If the only feasible way to generate earnings from the site is to be a Boost nominator, and the only feasible way to be chosen is by being a pub owner, you’ve created an incentive for complete fragmentation of the pub scene. Everyone wants to set up a new pub — despite the already crowded and overlapping market — because they may as well get themselves a ticket in the lottery.

Which undermines the funnel system that the Boost needed in order to function, with its ordered hierarchy. Good writers, frustrated with the system, reach the conclusion ‘if you can’t beat ’em, join ‘em’, and start publishing under their own titles.

By definition, they are now very unlikely to be boosted — at least unless they get the nod several months hence — , so the system has in fact reduced the volume of writing available during the phase when the Boost was meant to be enticing new users in.

2. The Boost — encouraging quality through incentives

Incentives are everything in economics. Get them right, and Homo economicus will of their own volition do exactly what you want them to do. Misjudge the numbers, and you’re herding cats instead.

So the idea here is that if we massively increase the rewards for Boosted articles, people will stop writing all that off-the-top-of-their-head crap, and instead spend time researching, crafting and honing fewer but higher-quality pieces.

Fine in theory.

So long as the effort/reward ratio works out. Let’s say a regular piece takes me one hour to write and nets me $3 on average. A Boosted piece brings in $50, and takes four hours. Bingo: I’m getting over four times as much per unit of effort!

But let’s say that only one in ten of those lovingly crafted 4-hour pieces actually gets Boosted. Suddenly my returns have dropped to less than half my bog-standard word salad rate. So I may as well give up on crafting, and just stick to grafting.

3. The Boost — outsourcing quality assurance

Here’s another given: outsource anything, and you lose control over your product and your reputation. You’d better be damn sure you can trust your subcontractors, and have a watertight SLA in place that you can reliably monitor and enforce.

Because if they screw up, it’s your good name flushed down the pan.

This must, again, have seemed like a good idea. The blah-blah corporate line is ‘because our publication owners are the true subject specialists in their chosen fields, and have the direct editorial relationship with their writers’.

But the bottom line for the bean counters is: it’s cheaper. And if their picks have to be greenlighted inhouse then the ‘subject specialist’ argument is undermined anyway.

Now, that greenlighting might be seen as a way of bypassing the outsourcing jeopardy. But again, does it work in practice? We know that stuff gets through the net. Some of which seriously cannot have been looked at for more than 5 seconds, if that.

The danger here is that you end up with the worst of both worlds. You can’t guarantee the quality of the external picks, but your attempts to do so rob the inhouse curators of the time to make their own.

And not just time — it’s also about mental freshness and enthusiasm. The more you turn a process into a production line, the more you deaden the soul of those involved.

4. The Boost — casting the net wider

‘Uncovering hidden gems’ — that is another of the programme’s supposed virtues. But again, the incentives work against it. We’ve all seen nominators talk of and show off their spreadsheets. Their hit rates for different topics and writers.

It’s worth remembering that a perfect 100% every month for a year would bring you in $10,800. Now that’s certainly an incentive. But is it the right incentive for the professed goal of ‘detecting unseen talent and voices’?

Because if the bulk of my earnings from Medium come from my accepted nominations, my priority becomes maximising that strike rate. It’s Moneyball, baby! And my spreadsheet tells me that Writer A is a good player for my team, but Writer B’s average sucks of late.

The tendency will be to narrow the selection pool to a few writers who produce the kind of stuff that is an easy win. Consequences: the boosted work — which is all that the algorithm promotes — becomes stale and repetitive. But that might be hard to see because each individual piece is decent enough in itself.

More importantly, perhaps, those writers with more eclectic, distinctive and challenging voices are overlooked and become disillusioned. The site loses variety, but also risks losing content.

You could counter this by setting arbitrary limits, but that again would be counterproductive — you would be rejecting work simply because that writer or topic has already filled their quota for the month.

5. AI filler — the elephant that shat on the tapestry

OK, filling the site with AI crap isn’t technically Medium policy. But you could certainly argue that the lukewarm, tardy and apparently ineffective response from management indicates that the recent proliferation is at least partly the direct consequence of policy decisions.

And unfortunately it operates in toxic synergy with the Boost. You essentially have a situation where above street level, all we can see at the top of our feeds is Boosted stuff. ‘Selected for you’ is these days synonymous with ‘Boosted in a topic area of assumed relevance to you’.

Meanwhile, in the murky depths beneath the streets of Gotham, the city sewers are clogged with an enormous fatberg of derivative, vacuous, stinky sludge. My guess is that the combination of the two phenomena has a lot to do with the widespread complaint that views and reads have gone round the U-bend of late.

6. Bots — a legal menace

Again, this comes down to ‘inaction’ or ‘inability to take action’, rather than active policy. It’s a tough one — no one has got a proper handle on this, and Medium’s shoestring team can’t be expected to work miracles.

But unfortunately they will have to live with the consequences of the bots, fakes, spiders and scrapers swarming the site. Because one of the consequences of all the above points is that writers are saying to themselves ‘If I can’t find an audience/monetise my work here, I need to look into placing it on other platforms’.

As we all know, the time lag between thinking that and acting on it may well be a matter of months, maybe more.

But in the meantime, some will — and I know I count myself among their number — be wondering whether this is a safe place to post work that they might want to reuse elsewhere in the near future.

Because if anything published here is being placed in the shop window to be plagiarised and reposted by third parties, that potentially jeopardises its value. The question for writers is: ‘Do I want to put my best work here, and risk losing control over it for the sake of 17 reads and $1.28?’.

And our personal answers to that question could again serve to further reduce the overall relative quality of the content on the site.

I don’t claim that Medium brought about all these consequences deliberately, nor that they could have entirely prevented certain such aspects from arising. Much of the above has always simply been a fact of life, and has been heightened by ‘the fast-moving technological world of technology in which we live in’.

But I do think they should all be priorities for the Medium team to address. And if they are already doing so, they need to let us, as users, know about it, with more convincing and reassuring words than ‘we are aware of the issue — we hope to have a solution soon’. Or simply ‘that’s a feature, not a bug’.

[Rant ends]