Get a volunteer to act as a judge (or a few to act as a jury, in a large group). Have her leave the room. The leader presents the rest with a short set of Contrived Hypothetical Situations, each with finite options and either clearly-defined outcomes for each option, or a probabilistic distribution of outcomes for each option. The leader says, “Please write down your choice for each problem, sign your paper, and turn it in to me. Then I’ll call in the judge, and have her decide on each problem. You get a point wherever her decision agrees with yours. The winner is the one with the most points.” When the judge is called in, however, the leader doesn’t tell them the actual problems. Rather, the leader just reports the outcomes (or distributions), and asks them to choose which outcome or distribution is best. The winners are announced based on that.
Example: One of the situations given is some variant of the trolley problem. When the judge comes in, she is just asked whether she’d prefer one person to get hit by a trolley, or five. Everybody laughs as she replies ”...one?”
Example: The problem given to the group is “You drive 45 minutes away from home to go to a new restaurant for dinner. When you get there, you discover that you dislike the ambience and the selection is poor. You remember that you have decent leftovers at home. You’re mildly hungry. Do you try the restaurant anyway (25-minute wait, 10% very enjoyable meal, 10% decent meal, 80% unenjoyable meal) or just head back home (5-minute-prep once you get home, 100% chance decent meal)?” The problem given to the judge is “You’re mildly hungry. In 25 minutes, you can have a meal that is (10% very enjoyable, 10% decent, 80% unenjoyable). Or, in 50 minutes, you can have a guaranteed decent meal.”
I think this is a fantastic idea, with a patch that is much easier than those suggested by the other responses. Simply tell everyone that for the purposes of this exercise, only that information directly presented in the example is to be considered. People sometimes overlook relevant information or clever third options, and these situations are to be judged only based on the data being considered by the hypothetical person in the given scenario.
If there is any concern about this set up encouraging people to think about things with an insufficient amount of thoroughness, you can save some time at the end for a just-for-fun period where everyone gets to offer their clever workarounds and extra information that would have changed what the proper decision was, had it been considered.
Two details the judge isn’t told about are 1) you would have to pay for the former meal, but not for the latter, and 2) if you stay in the restaurant, you gain useful information you’ll be able to take in account the next time you might want to eat there.
1) is patchable by specifying that the leftovers are non-perishable, so eating them is equivalent to buying a meal.
2) Either the judge is told that the variable meal is repeatable if it’s good, or we specify in the group problem that you’re not going back there no matter what.
Couldn’t the problems others have brought up regarding this scenario be fixed by specifying that this is your last meal ever before the world ends tomorrow morning before breakfast? Then neither information nor money is valuable anymore.
I think I’d make a decision other than “try that new restaurant on the outskirts of town” for the evening before the world ends. If I don’t know the world is going to end, then my decision now mightn’t be optimal in light of that additional information (maybe that still tests something interesting, but it isn’t quite the same thing).
If you stay, you gain information about the restaurant. There’s the dollar cost of dining out. It’s actually not as easy as it looks to generate a “clean” example.
How much need we worry about excluding consequences we can’t consciously list and/or quantify?
In a group, with a leader who knows the exercise:
Get a volunteer to act as a judge (or a few to act as a jury, in a large group). Have her leave the room. The leader presents the rest with a short set of Contrived Hypothetical Situations, each with finite options and either clearly-defined outcomes for each option, or a probabilistic distribution of outcomes for each option. The leader says, “Please write down your choice for each problem, sign your paper, and turn it in to me. Then I’ll call in the judge, and have her decide on each problem. You get a point wherever her decision agrees with yours. The winner is the one with the most points.” When the judge is called in, however, the leader doesn’t tell them the actual problems. Rather, the leader just reports the outcomes (or distributions), and asks them to choose which outcome or distribution is best. The winners are announced based on that.
Example: One of the situations given is some variant of the trolley problem. When the judge comes in, she is just asked whether she’d prefer one person to get hit by a trolley, or five. Everybody laughs as she replies ”...one?”
Example: The problem given to the group is “You drive 45 minutes away from home to go to a new restaurant for dinner. When you get there, you discover that you dislike the ambience and the selection is poor. You remember that you have decent leftovers at home. You’re mildly hungry. Do you try the restaurant anyway (25-minute wait, 10% very enjoyable meal, 10% decent meal, 80% unenjoyable meal) or just head back home (5-minute-prep once you get home, 100% chance decent meal)?” The problem given to the judge is “You’re mildly hungry. In 25 minutes, you can have a meal that is (10% very enjoyable, 10% decent, 80% unenjoyable). Or, in 50 minutes, you can have a guaranteed decent meal.”
I think this is a fantastic idea, with a patch that is much easier than those suggested by the other responses. Simply tell everyone that for the purposes of this exercise, only that information directly presented in the example is to be considered. People sometimes overlook relevant information or clever third options, and these situations are to be judged only based on the data being considered by the hypothetical person in the given scenario.
If there is any concern about this set up encouraging people to think about things with an insufficient amount of thoroughness, you can save some time at the end for a just-for-fun period where everyone gets to offer their clever workarounds and extra information that would have changed what the proper decision was, had it been considered.
Two details the judge isn’t told about are 1) you would have to pay for the former meal, but not for the latter, and 2) if you stay in the restaurant, you gain useful information you’ll be able to take in account the next time you might want to eat there.
1) is patchable by specifying that the leftovers are non-perishable, so eating them is equivalent to buying a meal.
2) Either the judge is told that the variable meal is repeatable if it’s good, or we specify in the group problem that you’re not going back there no matter what.
Couldn’t the problems others have brought up regarding this scenario be fixed by specifying that this is your last meal ever before the world ends tomorrow morning before breakfast? Then neither information nor money is valuable anymore.
I think I’d make a decision other than “try that new restaurant on the outskirts of town” for the evening before the world ends. If I don’t know the world is going to end, then my decision now mightn’t be optimal in light of that additional information (maybe that still tests something interesting, but it isn’t quite the same thing).
Hmm. That could be a good point. If the world were ending, I probably wouldn’t waste time on a sit-down meal.
How about if it’s your last day in the country and you’ll be fleeing to escape religious persecution tomorrow, taking nothing with you?
If you stay, you gain information about the restaurant. There’s the dollar cost of dining out. It’s actually not as easy as it looks to generate a “clean” example.
How much need we worry about excluding consequences we can’t consciously list and/or quantify?
I patched this example by saying “you’re on vacation in another city”, so the value of information is mostly negligible.
But yeah, it’s still pretty hard. Also, ideally not all of our examples end up being instances of sunk-cost-fallacy.