Heuristics and Satisficing¶

Why It Matters¶

Real decisions face computational and information limits. The right question isn't "what's optimal?" but "what's a good enough answer I can actually find?"

The Key Move¶

Ask: what's the cost of finding the optimum vs the cost of a good-enough solution? If search costs exceed the improvement from optimizing, satisfice. Use heuristics that perform well on average, accepting occasional suboptimality.

Classic Application¶

Chess. The game tree is too large to fully compute. Players use heuristics (control the center, develop pieces, king safety) that work well without being provably optimal.

Surprising Application¶

Hiring. Evaluating all candidates fully is infeasible. Satisficing: interview until you find someone good enough, then stop. Heuristics: proxies that correlate with success without being perfect (signals, credentials, referrals).

Failure Modes¶

• Searching for optimum when satisficing would serve better
• Using heuristics without understanding when they fail
• Confusing "simple" with "bad" (simple heuristics often outperform complex models)

Go Deeper¶

Gigerenzer's work on fast and frugal heuristics. Simon on bounded rationality.