Incentive Hacking

mechanism

When a subsidy rewards an easily fabricated proxy, participants can manufacture qualifying activity and collect the reward without creating the value the system intended to buy.

A restaurant did not need cooks, tables, or even customers to profit from Food Panda’s discounts. Someone could create a fake restaurant, place subsidized orders, mark them delivered, and pocket the platform’s venture-funded top-up. The transaction looked successful to the system precisely because the system was paying for the appearance of a transaction.

E1

The proxy becomes the product

The platform wanted genuine orders, but its simple rule rewarded a measurable substitute: an order recorded as delivered. Once the subsidy exceeded the cost of fabricating that record, fake demand became profitable. The participant did not have to defeat the rule; complying with its visible conditions was enough.

That distinguishes incentive hacking from ordinary rule-breaking. Read the payoff actually offered, and you can see the moving parts: a valuable reward, a cheap-to-produce signal, and weak verification of the underlying value. Repeated feedback then teaches participants which signals reliably release the reward. The result is a sharp form of perverse incentive: measured activity rises while the intended outcome does not.

E1

Simple rewards are not automatically broken

A subsidy is not hackable merely because it uses a simple rule. The exploit depends on artificial activity being cheaper and easier than the reward is worth. Strong verification can close that gap, but it adds friction and expense for legitimate users. That is the central design trade-off: make participation easy enough to encourage real behavior without making its proxy cheap to counterfeit.

E1

Price the cheapest fake

Before subsidizing an action, write down the lowest-cost way someone could produce the recorded signal without producing the intended value. For an order subsidy, trace what proves that a real restaurant, customer, payment, and delivery existed. Then redesign the rule or verification until manufacturing the proxy costs more than the available reward.

E1

Episodes that teach this