Episode Seventy Two: Symptom Masquerading As Disruption (2); The Model Is The Modeled; Labour Not Employment; Superstar Ratings, Here We Go; Not Swarm

by danhon

0.0 Station Ident

Today’s episode is mainly a response to what I wrote in yesterday’s episode. It feels like I unearthed a bit of a thread, and I got some great notes from readers. So let’s go pull on it.

1.0 Symptom Masquerading As Disruption (2)

Hat tips to Rowan Beentje, Kim Pallister, John V Willshire, Tim Maly and Matt Haughey for responding to the zero-hours, externalisation of employees point of yesterday’s Symptom Masquerading As Disruption. It’s clear there’s a range of opinion on this, from the belief that capitalism will deliver us *from* evil, to capitalism delivering us *to* evil: that it will either induce an unemployment rate of up to 50% in the near future, or that things will shake out and we’ll go back to a period of growth and what’s happening now is a mere blip.

1.1 The Model Is The Modeled

But first, John V Willshire’s observation, that I mentioned on Twitter kind of blew my mind. Now, John *has* studied economics, and the point he made was this: this “stack” view of people – that there are those who now think of people as virtualised substitutable AWS EC2 instances that can be activated, spun up, assigned a parcel of work, and then demobilised, “is the way that economists have always liked to think of people anyway – little atoms of meat who must behave in predictable ways.”

Yes, OK, so what we have is our humans as rational actors and, in a sense, what Uber and Airbnb have done is not necessarily produced an API that controls the world, but an API that instead controls other humans. We reach out and use these services, and our requests get translated, mediated, into instructions for other humans to perform for us. You can see a sort of spectrum-disorder response to this in Hacker News comments where occasionally someone will call for an even better version of Uber where there is literally no need to interact or converse with your driver at all, and essentially the human is totally abstracted away behind a piece of glass-fronted interface.

But John’s *best* point for me, was when he said:

“What if rather than being a way to describe the world, economics has unwittingly become a way to proscribe the world. Then we’re fucked.”

Abstract it away and it’s kind of saying this: a model of a subject that is so successful at describing the subject that the subject takes on the attributes of the model. The model becomes the thing being modeled.

This is a thing, now. Seeing the world as addressable stacks. A kind of mankind’s dominion over a computer-addressable, insructable directable world. There was someone at work who got super excited about “an API for the world!” and I think that’s kind of the problem for me: an API for the world abstracts the world so that you can deal with it and manipulate it, which is great, but the thing is we have a super high bandwidth low-latency interface for the world that’s super multi-modal. And I think it’s fair to say that our APIs for the world right now are really coarse and in that way, treat the objects (note! objects! Not people!) that they interact with in a necessarily coarse way. And humans aren’t coarse. Humans are many splendored things.

And maybe this is part of the whole “design with empathy” mini-crusade that I’m on. Sure, APIs that allow you to instruct humans to do things like Uber and Airbnb are successful right now, but I’m questioning whether they’re successful good, or successful because of a symptom of changes in the labour market, or, honestly, a combination of the two. And, you know, first attempt at providing an API layer for humans that’s more nuanced, I think, than Mechanical Turk, which I should’ve referenced earlier. But I like to think that an empathic API that’s more considerate of humans will do better than one that is less considerate. Remember this, hackers of the Bay Area: you do not like being thought of as replaceable resource units, and there aren’t many people who think “yeah, Human Resources is totally the best name for that department”.

Christ. Am I starting to sound like Jaron Lanier?

1.2 Labour But Not Employment

Most of the comments were about the fallacy that you can’t compare Uber or Airbnb with Hilton or a taxicab/car service because you need to count the hosts (and all the work they do, or that they outsource) with hotel staff and drivers with platform/despatch staff.

The point here in both cases is of a disruptive service (sure, creating or finding capacity and repurposing it and serving a need) that is *also* provided via a low-cost, arguably sub-cost-of living wages farmed out to *people who are willing to perform that labour at that price*. And yes, you can see both sides of the coin: comfy overpaid cabbies protected by unions who aren’t progressive, and simultaneously fair wages that will now drop to sub-minimum wage.

Kim’s not worried, because he has faith in capitalism working the way it should: either Uber has to compete and provide higher “wages” and reduce its take, or the overall service that Uber provides plummets as the quality of the labour workforce Uber’s able to attract decreases.

The correct point that Kim makes is that a job is a job is a job, whether you’re doing it for zero-hours or you’re doing it under contract with benefits and protection, that “whether you hammer license plates in a prison yard or a factory or in the comfort of your own living room between commercial breaks of Survivor, it’s still swinging a hammer.” Yes: that’s the *labour* side of it, but I think the point is that we’re trying to have some sort of civil society here, and it’s the fringe benefits that come with the hammer swinging that are at stake here.

Tim Maly’s contribution was to examine the new labour market: that there’s obviously a brand/communications story and agenda around “the kind of people who Uber and Airbnb benefit” and the reality and it would do us good to pay attention to the reality more closely. America, I feel, has always been an entrepreneurial, go-get-them state. I can’t remember which book it was I was reading a long time ago, but the thought that Russia couldn’t win because they didn’t inculcate kids into capitalism from an early age with rituals like the lemonade stand. So ideologically, I think Americans are predisposed to the “extra ways to achieve the American Dream” concept.

The point here is that there’s a romanticised “side job”, and there’s a reality “side job”. Again, it’s not necessarily Uber or Airbnb’s fault that there’s a side job labour market in the first place. In fact, well done for spotting it! But, how does this play out? Do “main” employers realise that they can be much more dynamic in terms of employment and scale back hours because now not only will the state pick up the slack with welfare, but so will other dynamic humans-as-a-service, er, services? Is the endpoint, as Maly suggested to me, a world where the majority of people have 2-3 unpredictable jobs that they must do well in order to maintain high star ratings? Because, and again, I’m *pretty* sure I’ll have agreement from the software engineers who read this: you kind of perform better when you’re able to concentrate on a singular task. Atomised, dispersed and on-demand tasks performed by humans may not be the best way to have those tasks performed well. And the threat of a bad star rating and less employment is not necessarily the best way to incentivise “good” performance. In any case, I think we can agree that star ratings, if considered in isolation as a feedback mechanism, are a route to local optima and not necessarily the best case.

1.3 Superstar Ratings, Here We Go

Rowan Beentje responded on Twitter with the thought that star ratings can be thought of as “crowdsourcing discrimination”, and in response to that, in a coarse way, I can see them as performing some sort of denormalising/anonymisation/dereferencing operation on *actual* feedback. I mean, we can agree that the regular star feedback mechanic is pretty sucky and not that useful[1]. On the other hand, yeah, good enough for Netflix. And sure, maybe better than not having any at all. Or, I don’t know, the happy face or sad face feedback machines that you see at airport immigration and customs these days.

On this, we can ask the question: are we using these algorithms to achieve the objectives we want in society, as well as in business? You can think of this as the Strong AI problem, the one that says that algorithms absolutely have needs and wants because they’re designed by humans, but also need to be understood in a special way. Let me paraphrase Elizer Yudokowsky: “The [algorithm] does not hate you, nor does it love you, but you are made out of atoms which it can use for something else.”

The defence here is that the algorithm is merely reflecting reality and hey, it’s just following orders. It wouldn’t be the algorithm’s fault that non-white drivers happen to receive worse ratings than white drivers from Uber’s audience. But algorithms, because they have the capacity to be blind *if they’re designed that way* can help, and not just reflect existing biases. The promise of big data, of course, is that the algorithms in a way could help spot such biases and correct for them. If, of course, we decide to design them that way.

One of the points I made in an earlier newsletter was that not doing something (“Not Trying Is A Signal”[2]) could and should be construed as a weak signal. When Dropbox takes the time to build a low barrier to entry web flow for opting out of dispute resolution compared to other companies requiring you to print out a PDF and mail it to a certain address, you can see that Dropbox “cares” more about its customers and the way it acts in the world.

And how you act in the world is, after all, something you – whether you’re a person or a corporate person – have control over.

[1] http://xkcd.com/1098/
[2] http://newsletter.danhon.com/episode-nineteen-not-trying-is-a-signal-peak-game-easyhard-snapchat/

2.0 Not Swarm

Meg Pickard[1], whom I’ve been friends with ever since blogging was “web logging”, sent me a note in response to the point on Foursquare’s naming of their new app, Swarm, and its insectile connotations.

What if Swarm were called Murmuration?[2]. As Meg says, the concept of murmuration brings forth “lots of individual units coming together to make amazing, unreal and stunning pattens. Unique ones. Natural ones. Organic ones.” A very different feeling, coming from a very different place.

Of course, the counter-argument is that “Tweet” is a verb now and so is “Google” and we don’t seem to have much of a problem with that, or even “Photoshopping” things.

But, I contend that Swarm is a name of a thing, and names come with power.

[1] http://www.megpickard.com
[2] https://www.flickr.com/search/?q=murmuration

Okay! Notes! Apologies to those of whom who sent notes and that I’m replying to through the medium of this newsletter. And there are *still* new people. This time from Sweden! Hello! Say hi!

Have a good weekend, everyone.

Best,

Dan