If-Then Rewards
Pink distinguished sharply between two reward structures. If-then rewards are prospective and contingent: the reward is promised before the work begins, specifying in advance what behavior will earn it. Now-that rewards are retrospective and noncontingent: the reward arrives after the fact, as acknowledgment of quality that has already been demonstrated. The distinction is not semantic. If-then rewards systematically trigger the overjustification effect for intrinsically motivated work because they redefine the purpose of the work in external terms before the work begins. Now-that rewards can enhance intrinsic motivation because they confirm that internally generated quality is valued, without making external acknowledgment the goal. The prescription for AI-age organizational design flows directly from this distinction — metrics systems must shift from prospective incentivization to retrospective recognition.
In the AI Story
The if-then structure is ubiquitous in corporate management: commission plans, performance bonuses contingent on specified metrics, grades for specified behaviors, the entire apparatus of measured-and-rewarded output that constitutes modern HR practice.
For algorithmic work — tasks with clear instructions and single correct outcomes — if-then rewards produce reliable performance improvements. The mechanism is robust and uncontroversial. Pay someone more to stuff envelopes faster and the envelopes get stuffed faster.
For heuristic work — creative, judgment-intensive, requiring flexibility — if-then rewards produce the overjustification effect. The Type I motivation that produces the best creative work is converted into Type X compliance that produces mediocre output at reliable rates.
The AI economy intensifies this dynamic because AI work is almost entirely heuristic. The algorithmic components have been absorbed by the tool. What remains is the judgment work that if-then rewards systematically corrupt.
Origin
Pink formalized the distinction in Drive (2009), drawing on Deci and Ryan's decades of self-determination research showing that the contingent nature of rewards — not merely their existence — triggered the motivational corruption.
The framework extended earlier distinctions in the motivation literature between controlling and informational rewards, specifying the temporal and contingency structure that distinguishes the two.
Key Ideas
Contingency is the poison. The reward's prospective specification, not its existence, triggers the overjustification effect.
Now-that rewards are safe. Retrospective acknowledgment of quality can enhance rather than undermine intrinsic motivation.
Heuristic-algorithmic asymmetry. If-then rewards work for algorithmic tasks and fail for heuristic ones.
Metrics are if-then rewards. Even without explicit bonuses, dashboards that measure specified behaviors function as contingent reward systems.
AI intensifies the stakes. The work remaining for humans is exactly the work if-then rewards most systematically corrupt.
Appears in the Orange Pill Cycle
Further reading
- Daniel H. Pink, Drive (2009), Chapter 2
- Edward L. Deci, Richard Koestner, and Richard M. Ryan, meta-analysis on extrinsic rewards (1999)
- Alfie Kohn, Punished by Rewards (Houghton Mifflin, 1993)