Baudoin's Razor

Don’t flip a coin, choose information.

• You are commuting somewhere, you always take road A, but you know road B is likely just as fast, but you never drove there before. Choose road B.
• You often order meal A. You’re thinking meal B would be nice too, in fact, likely just as nice. Choose meal B.

Coin flip

• These are situations where both outcomes are just as preferable (or the lack thereof) that you are indifferent to the choice of the outcomes. Not the outcomes themselves, but the choice or consequences of the outcomes.
• Assuming a 50% preference for either outcome (why you are reaching towards the coin), why choose the (relatively more) known outcome? The (more) unknown outcome has more value, namely in information.

A razor

• Because it cuts the potential actions and presents a clean choice.

The bandit problem

• It is similar to the bandit problem in machine learning but the bandit problem has learning as its primary objective. Baudoin’s razor has learning as a secondary objective, a simple means to improve upon the coin flip.