A city pays $1 for every dead rat brought in. This will reduce the rat population—right?

💭 How to Think About This

A PERVERSE INCENTIVE is a reward that accidentally encourages the wrong behavior. Pay people to catch rats? They might breed rats. Reward doctors per procedure? They might do unnecessary ones. The cobra effect shows how incentives can backfire when we don't think about how people will actually respond.

What do you think will happen?

The logic seems sound: More dead rats = more money for rat-catchers = more rats caught = fewer rats. Simple supply and demand for rat-catching services. But this assumes people will only CATCH existing rats. What if there's another way to get dead rats...?

This actually happened in British India with cobras. People BRED cobras for the bounty—it was easier than hunting wild ones! When the program ended, they released the now-worthless snakes. Result: MORE cobras than before. The incentive rewarded dead cobras, not fewer cobras.

The city WANTED fewer rats. But it PAID for dead rats. These aren't the same thing! People respond to what you actually pay for, not what you hope for. You can get a dead rat by (1) catching a wild rat, OR (2) breeding rats and killing them. Option 2 might be easier...

"When a measure becomes a target, it ceases to be a good measure." Dead rats measured rat-catching... until they became the target. Then people found ways to produce dead rats that had nothing to do with reducing rats. This happens EVERYWHERE incentives exist.

The optimistic view misses how people actually respond to incentives. The bounty doesn't reward "fewer rats"—it rewards "dead rats." Those aren't the same thing when breeding rats is possible.

• Obvious logic: more bounty = more catching = fewer rats

• Hidden response: breeding rats for bounty is easier

• The cobra effect: this exactly happened with snakes

• Goodhart's Law: measure → target → gamed measure

• People optimize for what you pay, not what you want

You spotted the trap! If catching wild rats pays $1, and breeding rats is cheaper than $1 per rat... people will breed rats. They're not evil—they're rational. The incentive creates a rat-breeding industry when it meant to create a rat-catching industry.

This is called the COBRA EFFECT. In British India, bounties for dead cobras led to cobra breeding. When the program ended, people released worthless snakes—MORE cobras than before! The policy achieved the exact OPPOSITE of its intent. Good intentions, terrible outcome.

The GOAL was fewer rats. The METRIC was dead rats. These seem related... but they're not the same. People respond to the metric, not the goal. "Dead rats" can be achieved by catching OR breeding. Whenever metric ≠ goal, perverse incentives appear.

To avoid perverse incentives: (1) Ask "How would a devious person game this?" (2) Align metric with actual goal; (3) Pay for OUTCOMES not activities; (4) Monitor for gaming. Maybe: pay for verified rat population REDUCTION in an area, not individual dead rats.

You recognized the perverse incentive! When you pay for dead rats instead of fewer rats, you might create a rat-breeding industry. The cobra effect shows this exact backfire.

• Bounties get gamed: breeding is cheaper than catching

• The cobra effect: this happened historically with snakes

• Metric ≠ goal: dead rats ≠ fewer rats

• Better design: pay for outcomes, not activities

• Always ask: "How would someone game this?"

Good instinct to look for details! Key questions: Can rats be bred easily? Is the bounty high enough to make breeding profitable? Can the city verify where rats came from? But here's the thing—if rats CAN be bred and the bounty exceeds breeding costs, people WILL breed them.

Historical example: British India paid for dead cobras. The details: cobras bred easily, bounty was worthwhile, verification was impossible. Result: cobra breeding industry. When program ended: released snakes. Outcome: MORE cobras. The details you'd want to check? They were bad.

The deeper principle: when you reward a METRIC, people optimize for the metric, not your underlying goal. "Dead rats" is a metric. "Fewer rats" is the goal. Any gap between metric and goal creates gaming opportunities. This is Goodhart's Law: measure → target → corrupted measure.

The "details" that make bounties work: when gaming is impossible or unprofitable. Better designs: pay for verified population reduction, use multiple metrics, monitor for gaming, think like a devious optimizer. Assume people will respond to incentives literally, not hopefully.

Looking for details is smart—but the key detail is whether the metric (dead rats) can be gamed to defeat the goal (fewer rats). If breeding is possible and profitable, it will happen.

• Key details: breeding feasibility, verification ability

• Cobra effect: the bad details were present, backfired badly

• Goodhart's Law: any gap between metric and goal invites gaming

• Good design: align metric with outcome, monitor for gaming

• Assume people respond literally to incentives

🤔 Which thinking lens(es) did you use?

Select all the lenses you used:

Learn more about the 7 Thinking Lenses →

👨‍👩‍👧 For Parents & Teachers

🌱 A Small Everyday Story

"We'll pay you $1 per book you read this summer!"
"Cool!"
[Later: Reads 20 very short, easy books]
"That's not what we meant..."
"You said per BOOK. Not per page. Not per difficulty."
"We wanted you to read more challenging..."
"Then the incentive should have said that.
I optimized for what you actually rewarded."

See more guidance →

🧠 Thinking habits this builds:

Anticipating behavioral responses to incentives
Distinguishing intended from actual outcomes
Thinking about how systems can be gamed
Designing more robust incentive structures

🌿 Behaviors you may notice (and reinforce):

"How might people game this?"
Spotting misaligned incentives in policies
Understanding unintended consequences
Thinking ahead to behavioral responses

How to reinforce: When discussing rules or policies, ask: "If people respond exactly to the incentive, what happens?" Practice spotting perverse incentives in news stories, school policies, or family rules.

🔄 When ideas are still forming:

Some learners may become cynical ("All incentives are manipulative!") or use this to justify gaming systems. Help them see that understanding perverse incentives is for designing BETTER systems, not exploiting bad ones.

Helpful response: "Yes, you could game that incentive. But understanding this helps us design better systems—and recognize when WE might accidentally reward the wrong thing. It's a tool for improvement, not exploitation."

🔬 If you want to go deeper:

Study Goodhart's Law and Campbell's Law
Explore historical policy failures
Research incentive design in organizations

Key concepts (for adults): Perverse incentives, cobra effect, Goodhart's Law, Campbell's Law, unintended consequences, gaming metrics, mechanism design.

A city pays $1 for every dead rat brought in. This will reduce the rat population—right?

Compare All Perspectives

🎯 "It will work"

💥 "It might backfire"

⚖️ "It depends"

🤔 Which thinking lens(es) did you use?