There’s a part of working in a fully remote environment that I am still struggling to understand, as a reformed philosopher and general skeptic.
Automattic, where I work, works. We get an astonishing amount of shit done, and we do so across a number of product lines and continents. We all work from home, or wherever we can find a flat surface and a clear wifi signal.
One of the things that surprises people the most about Automattic is the immense level of trust that is placed in every employee; it’s baked into the culture. We are committed to transparency and communication. We’re trusted to do our very best work and to find motivation intrinsically, rather than having it nudged upon us by bureaucratic processes or micromanagement.
As a student of philosophy, especially social choices, something about this model has always seemed rough around the edges to me. It sounded good, and it seemed like people believed it, but in theory it seemed like it should not work.
We can parse this disbelief of mine in the terms of the classic prisoner’s dilemma, what’s called a collective action problem.
Shipping, supporting, and improving something like WordPress.com is a huge task. It takes lots of people working together, collectively, to get it done. There are problems associated with this kind of collective action – one of them, the basic building block, is the prisoner’s dilemma.
The prisoner’s dilemma goes like this:
- You have two prisoners, X and Y. They’re kept separate.
- They’re offered plea deals: if X rats on Y, he can get off with a reduced sentence. If Y rats on X, he gets the same deal.
- Both X and Y know that the cops don’t have enough evidence without someone ratting.
- If neither X nor Y betrays the other, they both serve 1 year for a lesser sentence.
- If X and Y both betray each other, they each serve 2 years.
- If X betrays Y but Y stays silent, X goes free and Y gets 3 years.
Short story long, in a prisoner’s dilemma, ratting leads to suboptimal outcomes in terms of overall jail time. The best outcome across both actors (2 total years in jail) require both to remain silent – but the only way to pursue your own maximal outcome (no years in jail) is, by definition, removing the best outcome for both prisoners.
Individuals acting rationally to maximize their outcome ensures that the optimal collective outcome is not reached.
If your aim is to guarantee the lowest amount of jail time you get, the lowest possible jail time comes from the combination of your betrayal and your comrade’s remaining silent. The worst possible outcome results from your keeping silent while your erstwhile comrade betrays you. In the prisoner’s dilemma, betrayal always has the better calculus.
I’m cutting this short because this is a blog post, but collective action problems are super duper interesting.
If you want to dive down this rabbit hole (and if you do then we have the same kind of brain!), here’s a great place to start.
We can see the prisoner’s dilemma crop up all over the place, but especially when it comes to things like a public good. Let’s say your neighborhood wants to build a playground. You’d like a playground in your neighborhood and folks are taking up a collection (or possibly a Kickstarter, it is 2016 after all). The very best outcome for you would be if the playground gets built with no donation from you, if others contribute but you do not – because the good that is created is enjoyable for all and cost you nothing.
You’d be a free rider. But no one would know. Then you could spend your would-be donation money on champagne and Oreos.
It’s the prisoner’s dilemma that gives me pause about working fully remotely, because within the company, we have a bunch of really outstanding public goods. There are the obvious ones: unlimited vacation, being able to set your own schedule, outstanding health benefits.
When we dig a bit deeper, even more public goods come into focus; access to top talent who are generous with their time, the ability to work where you see the most impact – or to not work when you so choose.
When we think about these public goods, especially in the context of a results-only work environment with personally set schedules, we get to a place where it looks like the rational behavior is to not work that hard.
Stay with me, now.
Look, if we imagine a perfectly rational actor, who seeks to maximize their own utility at all times, and they see some benefit in the public goods on offer within Automattic, what is their course of action?
In this case, the public good that they’re seeking isn’t really relevant – if they want to move from design to development and want to maximize the time they can spend with a development mentor, or if they want to become the next World Champion of Hearthstone and as such simply want to minimize time spent in non-Hearthstone activities, whatever it is, the pursuit of it stays the same, right?
For our rational actor, to maximize their utility, it is in their best interest to do the amount of work that would get them fired, plus one unit of work. If they skate that edge, they will not get fired but will still enjoy the maximum utility available from the (awesome) public goods available within the company.
We can imagine on another hand, a different actor, motivated in some other way, who spends all of their time heads-down in their work, whether it’s live chat support or RSS feed debugging or expense reports or whatever – in virtue of the facts that a.) work takes time and b.) time is scarce, that means that our second actor here, a hard worker who is dedicated to the company, enjoys fewer of the public goods.
Unless the work itself represents the highest utility for an actor, according to the prisoner’s dilemma the rational actor at Automattic (and similarly organized companies) will always choose to skate one unit above being fired, gobbling up the public goods which are significantly less available to their more dedicated and harder working colleague.
And yet, it works. We work together. We ship software and support hundreds of thousands of customers. I can’t think of even one person I’ve worked with at Automattic who fits the description of the rational actor, above.
I don’t know if this is a problem with the prisoner’s dilemma, or if it should instead represent to me a sturdy underlining of a conclusion that I’ve always suspected:
Automatticians, and other folks finding success in the remote employment world, are by nature something other than purely rational.
I couldn’t be happier with this conclusion. Part of the secret sauce of Automattic is accepting this irrationality and letting each individual sit with it. This is what makes the trust so important. This is part of what makes remote work so special – it helps us move outside a purely selfish position of utility maximizing, and into something else.
Something better.