April 12, 2019

Anti-Patterns from Systems Thinking - Perverse Incentives and The Cobra Effect

Anti-Patterns from Systems Thinking - Perverse Incentives and The Cobra Effect

One anti-pattern from systems thinking we have probably all seen in action is perverse incentives. But yet, we seem to fall for this one over and over again.

Perverse incentives is what happens when you try to stimulate a system to achieve a certain goal by giving incentives towards that goal, but that incentive ends up having unintended, negative side effects. This is also known as the Cobra Effect, due to a famous anecdote which displays this particular quirk of systems.

The story goes that during British rule in India, the was a problem with a large number of venomous cobra snakes. Therefore the rules issued a bounty for every dead cobra handed in. While this worked initially, they soon noticed an increase in the amount of cobras being handed it in exchange for bounties. Investigation into the increase revealed that local entrepreneurs had started breeding cobras for income. Upon discovering this the government shut the reward program down, which made the cobra farms worthless. This led to the cobras being set free, and the result of the whole ordeal was that there were now more cobras than before.

While I can't guarantee for the authenticity of this anecdote, it does serve as a great example for how wrong it can go when you offer the wrong incentives - no matter how logical they might appear at first.

Another typical example is the way we reward people for working a lot, but not necessarily for working efficiently. We've all seen someone clocking in extra pay after working overtime to fix a screw-up, while avoiding the screw-up in the first place is seldom noticed.

There are also many good examples from larger systems, like our society. Measuring and rewarding students by grading them will provide students that get good grades. This is not necessarily the same as students learning in the best possible way. If welfare is measured by the country's GNP (Gross National Product), the country will focus on maximizing GNP. A high GNP is not necessarily the same as a high welfare, in fact, sometimes it might produce the inverse result.

When it comes to motivating people, there are two more topics I'd like to mention. First, it's hard to talk about incentives and measurements without mentioning Goodhart's Law. There's not much more to say about it, but everyone should be aware of it. Keep in mind, rewarding measurements does not necessarily mean giving monetary rewards. If teams producing a high velocity are rewarded with social status or higher autonomy and trust, velocity is now a measurement and people will often game the system to maximize their velocity (by padding estimate, cutting agility like I mentioned in my post on sub-optimization, etc).
Second, if you're interested in how to motivate people, I'd highly recommend the book Drive by Daniel Pink. I won't use this post to summarize an entire book, but two important things to learn from the books are included in point 2 and 3 below.

So, how do we create incentives, motivation and measurements in our organization? It's hard to provide hard rules to follow, but here are a few heuristics you can use when trying to find which direction to go:

  1. Make sure you are measuring is what you really want to achieve and the thing that will improve the welfare of the system.
  2. To motivate people, focus on intrinsic by allowing people to work autonomously, work towards achieving mastery and focusing on the purpose of their work.
  3. Be extremely careful with external awards. Although they do have their place, applying them in the wrong way can have catastrophic, non-reversible consequences. Don't confuse people putting in effort with people producing results
  4. Interestingly, one way to enable this is to focus on actual business outcomes, just like I mentioned in ways to avoid sub-optimization. This allows people to see the purpose of their work, and the autonomy to solve it in their preferred way.
  5. Continually look for the incentives in your system. There are very likely to be incentives there that you are not aware of.
  6. Continually re-evaluate the incentives and measurements you want to create. Keep in mind organizations are complex systems where you should make small changes and evaluate the effects of your actions.