Information Design & Bayesian Persuasion

You can persuade a rational person without changing the underlying facts by strategically packaging truthful information. This concept, called information design, involves simplifying or framing data to guide someone toward a particular conclusion, without lying. The simplified signal becomes a new, persuasive piece of information.

For example, a movie review site using a simple "Fresh" or "Rotten" score instead of a detailed percentage is designing information. This binary choice persuades more people to watch "Fresh" movies than a nuanced score of, say, 76% might.

Try this: When you see a simple pass/fail grade, a green/red health indicator, or a thumbs-up/down rating, ask what complex information is being simplified and why.

Information design is the process of structuring information to influence the decisions of rational actors. At its core is Bayesian persuasion, a framework where a "sender" (e.g., a company) commits to an information system (a "signal") to persuade a "receiver" (e.g., a consumer) to act in a way that benefits the sender.

The key is that the sender cannot lie, and the receiver knows the rules of the signal. The sender designs the test or report, and commits to revealing its outcome truthfully, whatever it may be. The receiver then rationally updates their beliefs based on the outcome, which constitutes new information for them.

Consider a medical diagnostics company developing a new test. Instead of just giving a complex probability score for a disease, they can design the test to return a simple "Action Recommended" or "Monitoring Recommended" result. A doctor, understanding the test's criteria, is persuaded to act on the first result, fulfilling the company's goal of facilitating clear clinical decisions.

This applies in policy as well. A government agency might grade restaurants "A," "B," or "C" for hygiene. This is more persuasive for the public than publishing the full inspector's checklist. The simplified grade designs the information to drive a clear choice: eat at 'A' restaurants, avoid 'C's.

The entire system relies on the sender's credible commitment. If the restaurant could bribe the inspector to change the grade after the inspection, the public would lose trust and the system would fail.

Bayesian persuasion is a formal economic model where a sender designs an information system (a signal) to influence a receiver's actions. The sender commits to the signal's rules, and the receiver rationally updates their beliefs based on the outcome, which persuades them to act in a way that, on average, benefits the sender.

• Kamenica & Gentzkow (2011) established that persuasion is possible if the sender can design a signal that splits the receiver's initial beliefs into new beliefs (posteriors) that induce the desired actions. [1]
• Bergemann & Morris (2019) situate this concept within the broader field of Information Design, which includes more complex scenarios like multiple senders or receivers. [2]
• Kolotilin et al. (2017) demonstrate that simple signals, like pass/fail grades or direct recommendations, are often the optimal tools for persuasion, particularly when the receiver has private information. [3]

The model has key limitations. It assumes receivers are perfect Bayesian updaters, ignoring cognitive biases. It also relies on the sender's ability to credibly commit to an information structure and can be computationally complex in practice.