Labs

Short exercises against the computer, each built around one idea from game theory or decision-making: when to trust, when to be unpredictable, how not to overpay, how to work back from a goal, how to test an idea, where a deal lies, how to weigh what you see, how wording sways a choice, what a rule rewards, and how a group should estimate. Each takes about five minutes and ends with a debrief of what happened and why. Points only, never money.

• Prisoner's Dilemma

  5 minutes · trains Cooperate or defend

  Cooperation pays when you will meet again, nobody knows which round is the last, and the other side answers in kind.

  About ten rounds against each of five bots: cooperate or defect, and watch the points.

• Rock-paper-scissors

  5 minutes · trains Be unpredictable

  If you are predictable, someone can exploit you. A random mix cannot be beaten.

  30 throws against a bot that learns your habits as you play.

• Sealed-bid auction

  5 minutes · trains Value and bid

  The winner of an auction for something of uncertain value is usually the one who overestimated it.

  Five jars of coins, one sealed bid each, against four bots.

• Nim

  5 minutes · trains Calculate

  Work backwards from the goal: find the positions you want to leave, and the first move follows.

  Take one to three counters from a pile; whoever takes the last one wins. Five games against a bot that plays perfectly when it can.

• Rule finder

  5 minutes · trains Read hidden information

  To find out whether you are right, include tests that would prove you wrong, not only ones that would confirm you.

  Guess the hidden rule behind a few numbers by trying your own; announce it when you are sure.

• Split the pie

  5 minutes · trains Value and bid

  Know your walk-away point and theirs before you bargain: the deal lies between them.

  Five negotiations over 100 points against bots, each side with its own walk-away point.

• Which bag?

  5 minutes · trains Weigh odds

  Start from how common something is, then adjust for what you see. And a streak of luck usually ends.

  Guess which bag a ball came from, then pick players for the next round after a lucky one.

• Gain or loss?

  5 minutes · trains Keep a cool head

  The same choice, put as a gain or as a loss, gets different answers. Restate it the other way before you decide.

  Eight choices between a sure thing and a gamble, then see where your answers changed with the wording alone.

• The bonus rule

  5 minutes · trains Think ahead

  People do what a rule rewards, not what you hoped for. Before you set one, ask what is the cheapest way to score on it.

  Four weeks setting the bonus for a support team of bots, and seeing what they do to earn it.

• Ask the panel

  5 minutes · trains Read hidden information

  Independent guesses, averaged, cancel each other’s errors. Guesses that follow the first voice do not.

  Six boxes of dots to estimate with a panel of bots: some rounds alone, some after hearing a first guess.

• Pick the same

  5 minutes · trains Cooperate or defend

  When everyone gains by choosing alike, pick what stands out to everyone, not your favourite. To get a group to aim high, go first where all can see.

  Four meet-ups with a crowd of bots and no way to talk, then eight hunts where a stag is caught only if all five hunters go for it.