Cheap Talk Discounting¶

Why It Matters¶

Most communication is cheap talk, and most people treat it as informative. "I promise," "I guarantee," "This is the best," "You can trust me" - all costless, all mimicable, all nearly worthless in zero-sum settings. Failing to discount cheap talk leads to systematic exploitation. Cheap talk discounting is defensive reasoning for strategic environments.

The Key Move¶

When someone makes a claim, ask: could someone with opposite characteristics make the same claim at the same cost? If yes, discount the claim heavily. Look instead for actions that bind, costs that were paid, or mechanisms that would punish lies. Treat words as weak evidence at best unless backed by credible commitment.

Classic Application¶

Used car sales. The seller says "This car is in great condition." This is cheap talk - any seller of any quality car would say the same thing. Rational buyers ignore it and rely on inspections, warranties, and reputation. The words carry zero information; only the seller's willingness to offer a warranty (costly if the car is bad) or the existence of third-party certification (costly to fake) provides signal.

Surprising Application¶

Academic letters of recommendation. In theory, these should be informative signals. In practice, they've become cheap talk - nearly everyone gets glowing letters, negative letters are legally risky, and the cost of writing "excellent" versus "adequate" is zero. Hiring committees increasingly discount them unless they contain specific costly details ("I hired this person for my own lab," "I'm betting my reputation") or come from referees with known standards (reputation as a costly signal).

Failure Modes¶

Overcorrection to cynicism: cheap talk can be informative when interests align or when repeated interaction creates reputation costs. Don't discount all communication, just adversarial cheap talk.
Missing mixed strategies: sometimes cheap talk works because receivers can't perfectly distinguish aligned from misaligned speakers. Bayesian updating, not binary ignore-or-believe.
Ignoring coordination: in pure coordination games (no conflict of interest), cheap talk can solve common knowledge problems even though it's costless. Context matters.

Go Deeper¶

Crawford & Sobel, "Strategic Information Transmission" (Econometrica, 1982) - formal theory of cheap talk
Farrell & Rabin, "Cheap Talk" (Journal of Economic Perspectives, 1996) - accessible overview

Bayesian Updating - Information Economics: Reasoning Under Asymmetry and Uncertainty