"Words are, of course, the most powerful drug used by mankind." - Rudyard Kipling. Powerful enough to build up or break down empires, yet game theorists talk about 'cheap talk'. How can the most powerful drug known to mankind be 'cheap'?
In game theory, whether words matter depends less on what is said and more on whether it changes what others believe.
Cheap talk is a message sent by a player in a game to other players at no cost to themselves, and nothing forces it to be truthful, and nothing directly happens because of it (no direct impact on payoffs). Would such communication make any real difference in a game? For example, suppose there are two software engineers James and John that must independently deploy either version A or version B of a software. If they deploy different versions, the system crashes, so they need to pick the same one. They're indifferent between the two, so both engineers struggle to decide which version to deploy. Upon deeper inspection of the codebase, James discovers version A fixes a bug that B does not and communicates it to John. This is cheap talk: James is not forced to tell the truth, and his words don't force John into making the same choice. However, John would rationally listen as it is always best for them to pick the same version as he knows James will pick A.
A simple, non-binding, non-verifiable message doesn't seem like it could do much, but it ensured that James and John picked the same software version. It did so by reducing the incomplete information problem the engineers previously had (they didn't know which one the other would pick). Cheap talk works because it gives players more information to work with. However, not all information is equal. Some are credible sources of information, others, not so much. The classic Prisoners Dilemma game is a good example of varying informational quality.
It is a very well known result that each prisoner, acting in their own self-interest, would defect and betray their partner in crime, so we end up with (Defect, Defect) as the final outcome, even though (Cooperate, Cooperate) is clearly a better choice for both players. The standard PD assumes the prisoners can't talk to each other and hence make choices independently. Now suppose they're in the same room, strategizing together on what to do. The cops won't beat them up if they lie, and their communication itself doesn't change payoffs (only the final decision does). This gives us a case of cheap talk, but does it change the outcome?
Suppose prisoner 1 tells prisoner 2 that they're surely going to cooperate. After all, it's far better than the outcome where both defect, right? Prisoner 2 finds Prisoner 1's rationale quite convincing and also decides to go ahead and cooperate, looking forward to the better payoff. Now that prisoner 1 believes prisoner 2 will cooperate, the set of possible outcomes is restricted to the first column. But notice that the payoff from defect (5) is greater than the payoff from cooperate (3) in the first column. Prisoner 1 has an incentive to mislead. Prisoner 1 picks defect. In reality, prisoner 2 would rationally also pick defect as that gives them the greatest payoff if player 1 cooperates. Once again, we end up at (Defect, Defect), nothing ever happens. Despite the addition of cheap talk, the coordination failure still persists because both players have an incentive to mislead. This is called a babbling equilibrium: messages are sent, but the informational value of the message falls to zero.
Nowhere is this more visible than online dating. We all know that writing a bio for your profile takes minimal effort, and there is no penalty for presenting an idealised version of yourself. Since there is no cost for overstating your strengths or understating your weaknesses, profiles start to sound similar. You will see the same phrases "travel, good vibes only, chill" across profiles. Individually these descriptions come across as sincere, but when these are repeated on so many profiles, they sound indistinguishable. Dating app users have adapted their behavior accordingly: instead of relying on written descriptions, they place more weight on pictures, mutuals, texting style, and behavioural compatibility. The silence of what is not said sometimes becomes more informative than what is.
What is interesting here is that communication weakens not because of speakers, but because listeners are perceptive; they understand the incentives behind the message and respond cautiously. This is why this same pattern can be seen in political speeches, corporate mission statements, public apologies and advertisements. Why do we find these to be pointless? Because the words simply aren't convincing enough. Their attempt at communication reduces to a mere performance, one that audiences have learnt to see through.
Consider corporate sustainability claims. If a company publicly announces that it is "deeply committed to environmental responsibility" while continuing harmful practices, consumers disregard the message. Over time, such repeated exaggerations reduce the informational value of similar statements across firms. As a result, people rely more on observational actions, such as actual carbon disclosures, third party certifications or measurable commitments, rather than verbal promises. Behaviour replaces speech as the more credible signal. In some cases, even restraint becomes strategic as saying less can further prevent loss of credibility. Silence, paradoxically, regains meaning.
At the heart of cheap talk lies a simple insight: words carry weight only when the speaker has something real to lose by lying. When the cost of deception is zero, promises degrade into noise. In a world saturated with communication, what we choose not to say - and what our actions reveal without words - often speaks the loudest.