Hack your incentives for fun and profit

There is a video that periodically makes the rounds on business social media sites like Linkedin. It shows a Southwest pilot hanging out of the cockpit, calmly cleaning the outside of his window. I find it funny because this clip works as an excellent Rorschach test of how people see the workplace. The pilot draws praise or condemnation depending on what people choose to see:

• A noble pilot with a can-do attitude literally rolling up his sleeves to solve a problem.
• A failure of management that forces a pilot to do work that either (a) he’s not trained for, or (b) is beneath his pay grade.
• A pilot hanging 20ft above the ground with no harness is surely too cavalier with risk to be trusted with a passenger plane.
• A voyeuristic culture that insists on recording and posting every mundane occurrence captured on smartphones.

Lost in almost all of these strongly-felt opinions about the moral character of the pilot, airline management, and the videographer is a much more interesting question:

Why was the pilot cleaning the window?

The obvious answer is “it was dirty,” but the question goes much deeper. Adam Smith wrote The Wealth of Nations almost 250 years ago, and in so doing he founded the field of economics. One of the central principles in his book, and of economics in general is:

People respond to their incentives.

There is plenty of evidence that the vast majority of people, in the vast majority of situations, act according to their incentives. Nonetheless, we’re so attuned to making moral judgements that we ignore the importance of incentives when evaluating situations. This is such an important mistake we make that in psychology it’s known as the Fundamental Attribution Error.

If you want to become a better manager and leader, you can take a big step in that direction by thinking more about incentives and less about moral judgements.

Incentives at work

Have any of these situations happened to you?

• You ask the office admin for help filling out an expense form, and he tells you there are documents on the intranet explaining how to do it.
• You’ve emailed a co-worker asking him to check in a source file so that you can proceed with some tests. He agrees but doesn’t do anything. Two days go by without response, and then you find out he’s on vacation for two weeks.
• Your boss asks you to finish work on two urgent projects. When you ask which one to prioritize, she tells you they’re both important and need to get done.
• You discover a major bug in a piece of code you wrote. You think about how to solve it without telling anyone, and in the meantime you get a glowing end-of-year review.

Is the office admin lazy? Is your co-worker selfish, your boss indecisive, and are you deceitful? You certainly can go down that road, assigning moral worth and moral failings to yourself and those around you. Or you can look to people’s incentives:

• The office admin received a directive from management to guide people to the intranet for everyday questions.
• Your co-worker knows that if he checks in that code, he’ll trigger a bunch of merge conflicts and review that he doesn’t have time for before his vacation.
• Your boss has three bosses breathing down her neck. Prioritizing one project over the other will piss off someone, and she doesn’t want that headache.
• Your end-of-year review defines your annual bonus, and delaying reporting the bug will ensure a good bonus for this year. Besides, the bug will be largely forgotten by the end of the next year.

Every person in these situations is responding to the incentives under which they’re operating. And if you want to improve the behavior of people in your organization, you’re probably better off examining their incentives than giving them Sunday-school homilies about good behavior. In each of these cases, changes to the incentive structure will probably bring about positive changes in behavior.

• The office admin gets a new directive, which is to help people with their queries and also show them where to find that information themselves.
• Required review for merge-conflicting code can be resolved in more ways than having the person who caused the conflict resolve it directly.
• Re-organizing the reporting structure of the group, so that your boss has only one person to report to, will give her more clarity on priorities and thereby give you that same clarity.
• Removing the disincentive to find bugs in one’s code seems like an important first step. Finding bugs should be encouraged, though of course we have to be mindful of the opposite incentive. We don’t want to incentivize the creation of bugs in order to find them, after all.

The point of all of this is to examine the way the company operates, and to iteratively improve it over time by recognizing the incentives that lead people to behave as they do.

AIs love hacking incentives

It’s not just people that are good at hacking their incentive structure. Anyone that has tried to create an autonomous computer or robot agent to do anything interesting has learned that automatic optimization algorithms are really good at finding loopholes in their reward functions (i.e. incentives).

A recent example is a robot that’s trained to move a block on a table deciding instead to move the table itself. The problem is that the robot was rewarded by the block’s absolute position. Easier to shove a big table than do the fine manipulation required to move the block. And this reward-hacking isn’t new. There’s a famous 1994 paper where a system that evolves simulated walkers and gets rewarded for speed decides to win the contest by creating really tall walkers and then letting them fall over.

The point of these humorous incidents is that (a) it’s really hard to create good incentive structures even for simple situations, so (b) you’re probably going to need to take a few cracks at it.

The value of principles

As you might imagine, it’s phenomenally difficult to create good incentive structures in complex environments like real companies. So much so that, while it’s important to create good incentives and to improve them over time, even that’s not enough. We need something more: something that gives useful guidance in the absence of explicit instructions. What we need are principles.

The principles we have in mind aren’t some deep moral code, nor are they airy (and worthless) generalities like “don’t be evil.” In order to be useful, principles can’t be so general as to be vacuous. They have to be able to specifically guide behavior in desired directions. “Great customer service at any cost” is a difficult principle to actually enact: should a customer service agent charter a helicopter to pick up and replace a defective pair of shoes? Better something like “When in doubt, err on the side of giving customers what they want.” The nuance is that most of the time, you do what you are trained to do, but in marginal or odd situations, the principle guides the person in a useful direction.

As a leader, it’s important for you to actually enact the principles in your daily actions. It’s particularly important to refer to principles when explaining and defending decisions. This creates good alignment with the people affected by the decisions: employees, partners, customers. By continually reinforcing the use of organizational principles, leaders embed culture in the minds of everyone, and principles become more than empty words intoned at a once-yearly meeting.

Incentives for everyone

Successfully organizing a bunch of people with different backgrounds, opinions, ideas, and motivations is no easy task. It sometimes seems like a miracle that it happens at all, and we’ve all experienced dysfunctional workplaces where peoples’ incentives actually work against the success of the organization.

The key to all great, long-lasting organizations is that they’ve managed to solve this incentive problem. Your organization can become one of them, but the first step is recognizing the path. And as we’ve seen, that path is paved with good incentives.