Concreteness Game
The object of this game is to train players to generate examples for explaining concepts quickly. The game requires at least two people, but may work better with groups of three or four.
To play, one of the players (Asker) names a concept to be explained, such as “How do you traverse a linked nodal network”, “Explain the law of Conservation of Energy”, or “What constitutes good financial advice?”
The other player (Generator) then tries to explain the concept/skill/other by using a nearby object to assist. The Generator should relate the example back to the original query and explain how the example demonstrates the experimental predictions that the concept makes (see Extended Example below). The Asker listens to the explanation, and once the Asker feels as though the generator has been able to explain the concept fully, they indicate to the Generator that the Generator should pick another object and start again. Another option the Asker should exercise is to ask follow-up questions about the Generator’s example, asking for clarification or elaboration by interacting with their object in some way (examples in the Extended Example section below)
Extended Example:
Asker: “Okay, let’s try this. Explain Newton’s Third Law.”
Generator: “All right, hmm… Okay, take that big old oak tree as an example! Now, imagine that I want to push this tree over. If I push on it, it doesn’t really move anywhere, and neither do I. That’s because the tree has a really strong root system to prevent it from being tipped over, while I’m braced against the ground. That means the force I put on the tree doesn’t go anywhere. Now, what happens if I don’t brace against the tree? I’m going to try and push with the same strength that I did just now. If the tree didn’t exert any force on me, then why did it force my arms away strongly enough to tip me over?”
Asker: “What happens if you were standing on ice instead of dirt?”
Generator: “Hmm… If I wanted to avoid spinning away from the tree, then I couldn’t push against it too hard. If I did want to, though, then I could turn this whole area into a pinball machine by pushing off of all the trees!”
Asker: “Okay, another example this time!”
Generator: “All right, look at that guy in his kayak over there. Every time he uses his paddle to push himself forward, you see a whirlpool where his paddle was. That’s because he’s pushing against a lot of water with the flat of the paddle. The paddle pushes on the water, and the water pushes back on the paddle. You can see this because he moves forward while the water is disturbed. If the law didn’t apply then the water shouldn’t-”
Asker: “Okay, you’ve got that one, give me another one!”
Generator: “Hmm...”
Through playtesting, we’ve noticed that there are two broad categories of useful questions in this game. The first type are “You have a rule, now apply it with the objects in front of us”. For example, one might describe good financial advice by pointing at a nearly-broken-down refrigerator: “Some advice you have to use right away like take-out food before it spoils. Other advice will keep basically forever, like ketchup maybe, while other advice will let you repair the refigerator to let it keep everything fresh for longer. What you want is to avoid poisoning yourself and keep healthy, so good advice is anything that keeps from poisoning you or your friends. Stock tips are like take-out food, while better mental models are like fixing the fridge.” The other type of useful question seems to be descriptive in nature. “Explain the life-cycle of a caterpillar”: “Well, imagine riding around on a cheap kid’s bicycle and picking up tubing, gears, and other supplies as you ride around. Then you take the bike and everything you collected into a room, work for a while, and you come out with an awesome racing bike that lets you do things you never could before”.
Example questions:
Explain Conservation of Energy.
What constitutes good scientific practice?
Explain the sunk cost fallacy.
How do you lift heavy things safely?
What constitutes an efficient algorithm?
Explain (Hansonian) signaling.
Explain priming and how expectations about the quality of a thing can affect your assessment of its quality.
Describe how Omega might optimize for happiness.
Explain what a set (in set theory) is and some of its basic properties.
Explain the fallacy of gray.
Theory (SPOILER ALERT! This section contains material likely to prime your reactions)
This game is designed to help people provide concrete examples on demand. The expectation is that: Forcing the players to compare their mental models against physical objects makes their explanations more concrete because physical objects can be interacted with. If a Generator relates a tree to AI theory, the question “What is a branch and what happens when I push on it?” seems to yield concrete answers more often in practice than “What are the features of AI theory and why do they matter?”.
The concept of follow-up questions seems to greatly increase the fun of the game. Many more grins were observed when the Asker occasionally quit saying “Okay, got it, another example now!” and instead interacted with the physical model in some way.
Playtesting seemed to also show that people really enjoyed coming up with examples up to their third, and the fourth became difficult to generate while the fifth was simply not all that fun to force ourselves to come up with. Priming may have been an issue, but initial results suggest asking for roughly three examples is the most fun before moving to a different question.
Concreteness Game The object of this game is to train players to generate examples for explaining concepts quickly. The game requires at least two people, but may work better with groups of three or four.
To play, one of the players (Asker) names a concept to be explained, such as “How do you traverse a linked nodal network”, “Explain the law of Conservation of Energy”, or “What constitutes good financial advice?”
The other player (Generator) then tries to explain the concept/skill/other by using a nearby object to assist. The Generator should relate the example back to the original query and explain how the example demonstrates the experimental predictions that the concept makes (see Extended Example below). The Asker listens to the explanation, and once the Asker feels as though the generator has been able to explain the concept fully, they indicate to the Generator that the Generator should pick another object and start again. Another option the Asker should exercise is to ask follow-up questions about the Generator’s example, asking for clarification or elaboration by interacting with their object in some way (examples in the Extended Example section below)
Extended Example:
Asker: “Okay, let’s try this. Explain Newton’s Third Law.” Generator: “All right, hmm… Okay, take that big old oak tree as an example! Now, imagine that I want to push this tree over. If I push on it, it doesn’t really move anywhere, and neither do I. That’s because the tree has a really strong root system to prevent it from being tipped over, while I’m braced against the ground. That means the force I put on the tree doesn’t go anywhere. Now, what happens if I don’t brace against the tree? I’m going to try and push with the same strength that I did just now. If the tree didn’t exert any force on me, then why did it force my arms away strongly enough to tip me over?” Asker: “What happens if you were standing on ice instead of dirt?” Generator: “Hmm… If I wanted to avoid spinning away from the tree, then I couldn’t push against it too hard. If I did want to, though, then I could turn this whole area into a pinball machine by pushing off of all the trees!” Asker: “Okay, another example this time!” Generator: “All right, look at that guy in his kayak over there. Every time he uses his paddle to push himself forward, you see a whirlpool where his paddle was. That’s because he’s pushing against a lot of water with the flat of the paddle. The paddle pushes on the water, and the water pushes back on the paddle. You can see this because he moves forward while the water is disturbed. If the law didn’t apply then the water shouldn’t-” Asker: “Okay, you’ve got that one, give me another one!” Generator: “Hmm...”
Through playtesting, we’ve noticed that there are two broad categories of useful questions in this game. The first type are “You have a rule, now apply it with the objects in front of us”. For example, one might describe good financial advice by pointing at a nearly-broken-down refrigerator: “Some advice you have to use right away like take-out food before it spoils. Other advice will keep basically forever, like ketchup maybe, while other advice will let you repair the refigerator to let it keep everything fresh for longer. What you want is to avoid poisoning yourself and keep healthy, so good advice is anything that keeps from poisoning you or your friends. Stock tips are like take-out food, while better mental models are like fixing the fridge.” The other type of useful question seems to be descriptive in nature. “Explain the life-cycle of a caterpillar”: “Well, imagine riding around on a cheap kid’s bicycle and picking up tubing, gears, and other supplies as you ride around. Then you take the bike and everything you collected into a room, work for a while, and you come out with an awesome racing bike that lets you do things you never could before”.
Example questions: Explain Conservation of Energy. What constitutes good scientific practice? Explain the sunk cost fallacy. How do you lift heavy things safely? What constitutes an efficient algorithm? Explain (Hansonian) signaling. Explain priming and how expectations about the quality of a thing can affect your assessment of its quality. Describe how Omega might optimize for happiness. Explain what a set (in set theory) is and some of its basic properties. Explain the fallacy of gray.
Theory (SPOILER ALERT! This section contains material likely to prime your reactions) This game is designed to help people provide concrete examples on demand. The expectation is that: Forcing the players to compare their mental models against physical objects makes their explanations more concrete because physical objects can be interacted with. If a Generator relates a tree to AI theory, the question “What is a branch and what happens when I push on it?” seems to yield concrete answers more often in practice than “What are the features of AI theory and why do they matter?”. The concept of follow-up questions seems to greatly increase the fun of the game. Many more grins were observed when the Asker occasionally quit saying “Okay, got it, another example now!” and instead interacted with the physical model in some way. Playtesting seemed to also show that people really enjoyed coming up with examples up to their third, and the fourth became difficult to generate while the fifth was simply not all that fun to force ourselves to come up with. Priming may have been an issue, but initial results suggest asking for roughly three examples is the most fun before moving to a different question.
EDIT: Formatting