I recently had an idea that seemed interesting enough to post here: “Shut Up and Multiply!”, the video game
The basic idea of this game is that before each level you are told some probabilities, and then when the level starts you need to use these probabilities in real time to achieve the best expected outcome in a given situation.
The first example I thought of is a level where people are drowning, and you need to choose who to save first, or possibly which method to use to try to save them, in order to maximize the total number of people saved.
Different levels could have different scenarios and objectives.
You are given time to examine the probabilities before the level starts, but once it starts you need to make your decisions in real-time.
Another twist: You see the actual outcome of your actions, randomly generated by the probability formulas. However, you aren’t scored based on the actual outcome, but instead you are scored on the expected outcome of your actions, using the expected utility formula. I originally intended this to prevent people from getting a high score just by luck, or to prevent low scores caused by bad luck. Though I later realized that this doesn’t actually fix the problem—you can still play repeatedly until, by luck, you happen to guess high-scoring actions. Still, I think it would be a good idea to show both scores.
Other random ideas:
The first few levels should be a tutorial. Showing how to do the calculations in order to maximize your expected score. Or there could be a separate turotial mode. Or maybe the game itself is a bad idea, but the tutorial might still be useful.
During each level you need to make your decisions as quickly as possible—the longer you wait the worse you score. Though maybe only some levels should be like this.
Later levels require more options to choose from, and more complex scenarios.
As much content as possible should be generated randomly, to prevent the game from being the same if you play it again.
Maybe the player could also be scored based on some calculations they do before the level starts? Or just integrate this with the tutorial?
And most importantly: Specifically design the game so that the player must learn to overcome some of the standard biases, in order to maximize their score. We should try to work in as many of these biases as possible. And also plenty of generally useful advice for working with probabilities.
So, now that I posted this idea, I’ll let you decide what, if anything, we should do with this idea.
First, is this a stupid idea, that couldn’t possibly work as described?
Or is it a good idea, but a low priority, compared to the other projects we’re working on?
Should this be a group project? Does anyone volunteer to lead the project? Does anyone want to take on the project entirely on their own? Or should I lead the project, or work on it on my own?
What language would be best to implement this in? Flash? Java? PHP? Python? Something else?
I still haven’t earned much karma on this site (only 1 point actually, when I originally posted this). Mainly because I don’t expect to have anything original to say. And so I’m posting this here as a comment in the Open Thread, rather than making an actual post of it. If this comment gets enough upvotes for me to be able to make this into its own post, I plan to do so, unless someone objects. Or if anyone would like to take over the idea, please feel free to do so. I don’t care about credit, and generally prefer to avoid it. Possibly by the flawed reasoning “credit = responsibility = blame”, which I suppose might deserve a post of its own.
The first example I thought of is a level where people are drowning, and you need to choose who to save first, or possibly which method to use to try to save them, in order to maximize the total number of people saved.
I do competitive lifeguarding (possibly the world’s weirdest recreational activity) and there is actually an event like this, called Priority Assessment or PA. Your team walks in and finds an area of the pool with a bunch of people drowning (for a team of 4 rescuers, usually it’s about 12 victims.) The scoresheets are set up so that you get more points for rescuing the victims who are more likely to survive–i.e. non-swimmers and injured swimmers have a much higher point multiplier than unconscious, submerged victims. PA involves a lot of strategy–it’s not always the teams of fast swimmers that win, although that helps. There is an optimal strategy, which has to be worked out in advance because it’s a two-minute event.
I’ve been thinking about educational games as well. The main problem, it seems to me, is that trying to make learning fun for someone who isn’t already interested and motivated is a waste of time because you’re just trying to hide the teaching under a sugarcoating of computer game, and that never works. On the other hand trying to make learning fun for someone who is already interested and motivated is pointless, because they already want to learn and the game just adds needless hassle like completing levels in order to reach the next piece of knowledge, or whatever game mechanic you’re using. It’s a pity, because I think of the way games like Portal build up complex puzzles from simpler ones and use the level itself to ask the player leading questions, like a kind of visual/spatial socratic method, and I think there must be a way to use that to teach, espiecally mathematics where visual/spatial metaphors could easily translate into mathematical metaphors… but I just can’t come up with a concrete version of the idea that wouldn’t be boring or stupid.
Lately I’ve been thinking that the fastest way to get to grips with a new subject is probably just to memorise big chunks of information without trying to understand it, using techniques like a memory palace and spaced learning programs like Mnemosyne and Anki, then think about what you’ve learned later, and insight might strike you. This would be espiecally effective if you combined it with a social precommitment to teach your knowledge to someone else, or to take part in a competitive quiz.
Lately I’ve been thinking that the fastest way to get to grips with a new subject is probably just to memorise big chunks of information without trying to understand it, using techniques like a memory palace and spaced learning programs like Mnemosyne and Anki, then think about what you’ve learned later, and insight might strike you.
If this is true and you aren’t too much of an outlier, it would go a decent way to explain the failure of a good chunk of educational theory in the past few centuries.
If this is true and you aren’t too much of an outlier, it would go a decent way to explain the failure of a good chunk of educational theory in the past few centuries.
The basic idea works for me, but I think Tom’s simplifying it. It’s not about “add to Anki with 0% understanding” vs “gain 100% understanding when first learning”; instead, it’s more like “add to Anki with 80% understanding” vs “gain 90% understanding, having had to spend several hours for the extra 10%.”
Far too many people tend to get hung up over that one thing in a chapter that they can’t understand. More often than not, it’s something they could understand perfectly if they just said “meh” and read a few pages ahead, but no; they just stay stuck on that one spot, thinking “wtf is this??!!”
Also, far too many people read books word-by-word when they could get essentially the same amount of information by skimming over the pages. Anki helps here, as it forces you to extract the relevant pieces of information from the text, (or at least stuff that looks important) instead of letting you comfortably wade through a wall of text and believe you’ve understood it.
(It seems, on first glance, that these two paragraphs contradict each other, but they actually don’t. The third one is talking about stuff that looks easy but actually isn’t, the second one’s talking about stuff that looks difficult, but wouldn’t if you’d just read ahead.)
Also, you generally don’t have to wait until later for insight into whatever you’ve failed to understand… More often than not, insight strikes even as you’re adding the cards.
I’d suggest making it so you get scored by the number of people you save, but the game is long enough that luck doesn’t make a difference.
I wonder if it would be a good idea to give it multiple scores. For example, lives saved, life-years saved, quality-adjusted life-years saved, etc.. This way, you won’t have as many problems with people disagreeing with the scoring system.
Alternately, you could just have it so you could change the scoring mode in the options part. It would also act somewhat as a difficulty setting. It would get harder when you have to weigh a destitute child, who will live longer, vs. a middle-class adult, who will be happier.
Reading that, it occurred to me that in all the computer games I’ve played, it is possible to totally succeed. Kill every monster, pick up every piece of treasure, solve every puzzle, gain the perfect score. (I have not actually played all that many, but I also read some of the games press and it seems to be the usual case.) Even the old arcade games fit the pattern: you can’t win them, only endure as long as possible, but until then your goal is to impeccably handle every challenge that comes at you.
Are there any games in which this is deliberately made impossible? For example, PeerInfinity’s suggestion of trying to save drowning people, in a scenario that makes it impossible to save all of them (and which is varied every time you play the game, so you can’t simply search out the optimal walk-through). Military operations in a city, fighting door-to-door, with innocent civilians everywhere whose lives matter in game terms.
I’m not sure the suggestion of a game in which one cannot ‘get the top score’ makes sense. It seems contradictory - ‘is there an optimal path through the game which is not the optimal path through the game?’
Can you have games where the ‘path’ to a top score, the optimal play, varies from game to game? Sure. Not every game carries it to quite the extent of Nethack, but most do it to some extent. Non-random games like go or chess are generally the exception, and they can be trivially randomized. But each specific game can be seen as ultimately deterministic: given the output of the random number generator this time, the ideal path is such-and-such. You, the player, may not know it, but that’s your fault.
Can you have games which deceive the player about what the best possible score is? Sure. The original Donkey Kong promises that you can play indefinitely; but go too high and the game will always crash. The upper bound is not where one thought it was. Or there are political games in which one tries to prevent 9/11 (IIRC); of course, the game must sooner or later defeat you, like those old arcade games.
What would it mean for a game to have scores players couldn’t reach? If in Mario, there is code to paint a picture of a 1-UP on a corner of the screen surrounded by unbreakable blocks, then in what sense is the player missing out on 1k points (or whatever). If a cut scene depicts a hostage dying, then how ‘could’ I have saved it? What if I can choose between a cut scene depicting hostage A dying, and hostage B? What if it’s in-game, and there’s a timer or rescuing A triggers the death of B?
Or what if there is a trove of 1 million points coded in, but the only access is to type in a true contradiction? Would the world record holder for Mario really be missing 1 million points off his score just because he can’t come up with one? (Yes, there is a number equal to his score+1 million; but there’s an infinite number of integers. What makes score+1 million special? All the lower number are special because it’s possible to manipulate a given blob of code to display characters we interpret as those lower numbers; but we can’t get it to emit any images of higher numbers, and that’s that.)
If you choose one sub-game, then there’s optimal play for that; if you switch between them, then there’s still optimal play, it’s just you need to weight it if there is no canonical ultimate score (just like with utilities).
If there aren’t any scores or sense of progress at all, I question whether it’s a game at all, or whether it merely bears a Wittgensteinian family resemblance. If you and I push around piled pieces on a Go board just for the pleasure of watching the piles build and collapse and form swirling patterns, we’re doing something entertaining (maybe) but who would call it a game? To call life itself a game is to either commit a tired weak metaphor, or to drain the word game of all meaning.
I recently had an idea that seemed interesting enough to post here: “Shut Up and Multiply!”, the video game
The basic idea of this game is that before each level you are told some probabilities, and then when the level starts you need to use these probabilities in real time to achieve the best expected outcome in a given situation.
The first example I thought of is a level where people are drowning, and you need to choose who to save first, or possibly which method to use to try to save them, in order to maximize the total number of people saved.
Different levels could have different scenarios and objectives.
You are given time to examine the probabilities before the level starts, but once it starts you need to make your decisions in real-time.
Another twist: You see the actual outcome of your actions, randomly generated by the probability formulas. However, you aren’t scored based on the actual outcome, but instead you are scored on the expected outcome of your actions, using the expected utility formula. I originally intended this to prevent people from getting a high score just by luck, or to prevent low scores caused by bad luck. Though I later realized that this doesn’t actually fix the problem—you can still play repeatedly until, by luck, you happen to guess high-scoring actions. Still, I think it would be a good idea to show both scores.
Other random ideas:
The first few levels should be a tutorial. Showing how to do the calculations in order to maximize your expected score. Or there could be a separate turotial mode. Or maybe the game itself is a bad idea, but the tutorial might still be useful.
During each level you need to make your decisions as quickly as possible—the longer you wait the worse you score. Though maybe only some levels should be like this.
Later levels require more options to choose from, and more complex scenarios.
As much content as possible should be generated randomly, to prevent the game from being the same if you play it again.
Maybe the player could also be scored based on some calculations they do before the level starts? Or just integrate this with the tutorial?
And most importantly: Specifically design the game so that the player must learn to overcome some of the standard biases, in order to maximize their score. We should try to work in as many of these biases as possible. And also plenty of generally useful advice for working with probabilities.
So, now that I posted this idea, I’ll let you decide what, if anything, we should do with this idea.
First, is this a stupid idea, that couldn’t possibly work as described?
Or is it a good idea, but a low priority, compared to the other projects we’re working on?
Should this be a group project? Does anyone volunteer to lead the project? Does anyone want to take on the project entirely on their own? Or should I lead the project, or work on it on my own?
What language would be best to implement this in? Flash? Java? PHP? Python? Something else?
I still haven’t earned much karma on this site (only 1 point actually, when I originally posted this). Mainly because I don’t expect to have anything original to say. And so I’m posting this here as a comment in the Open Thread, rather than making an actual post of it. If this comment gets enough upvotes for me to be able to make this into its own post, I plan to do so, unless someone objects. Or if anyone would like to take over the idea, please feel free to do so. I don’t care about credit, and generally prefer to avoid it. Possibly by the flawed reasoning “credit = responsibility = blame”, which I suppose might deserve a post of its own.
I do competitive lifeguarding (possibly the world’s weirdest recreational activity) and there is actually an event like this, called Priority Assessment or PA. Your team walks in and finds an area of the pool with a bunch of people drowning (for a team of 4 rescuers, usually it’s about 12 victims.) The scoresheets are set up so that you get more points for rescuing the victims who are more likely to survive–i.e. non-swimmers and injured swimmers have a much higher point multiplier than unconscious, submerged victims. PA involves a lot of strategy–it’s not always the teams of fast swimmers that win, although that helps. There is an optimal strategy, which has to be worked out in advance because it’s a two-minute event.
I’ve been thinking about educational games as well. The main problem, it seems to me, is that trying to make learning fun for someone who isn’t already interested and motivated is a waste of time because you’re just trying to hide the teaching under a sugarcoating of computer game, and that never works. On the other hand trying to make learning fun for someone who is already interested and motivated is pointless, because they already want to learn and the game just adds needless hassle like completing levels in order to reach the next piece of knowledge, or whatever game mechanic you’re using. It’s a pity, because I think of the way games like Portal build up complex puzzles from simpler ones and use the level itself to ask the player leading questions, like a kind of visual/spatial socratic method, and I think there must be a way to use that to teach, espiecally mathematics where visual/spatial metaphors could easily translate into mathematical metaphors… but I just can’t come up with a concrete version of the idea that wouldn’t be boring or stupid.
Lately I’ve been thinking that the fastest way to get to grips with a new subject is probably just to memorise big chunks of information without trying to understand it, using techniques like a memory palace and spaced learning programs like Mnemosyne and Anki, then think about what you’ve learned later, and insight might strike you. This would be espiecally effective if you combined it with a social precommitment to teach your knowledge to someone else, or to take part in a competitive quiz.
If this is true and you aren’t too much of an outlier, it would go a decent way to explain the failure of a good chunk of educational theory in the past few centuries.
The basic idea works for me, but I think Tom’s simplifying it. It’s not about “add to Anki with 0% understanding” vs “gain 100% understanding when first learning”; instead, it’s more like “add to Anki with 80% understanding” vs “gain 90% understanding, having had to spend several hours for the extra 10%.”
Far too many people tend to get hung up over that one thing in a chapter that they can’t understand. More often than not, it’s something they could understand perfectly if they just said “meh” and read a few pages ahead, but no; they just stay stuck on that one spot, thinking “wtf is this??!!”
Also, far too many people read books word-by-word when they could get essentially the same amount of information by skimming over the pages. Anki helps here, as it forces you to extract the relevant pieces of information from the text, (or at least stuff that looks important) instead of letting you comfortably wade through a wall of text and believe you’ve understood it.
(It seems, on first glance, that these two paragraphs contradict each other, but they actually don’t. The third one is talking about stuff that looks easy but actually isn’t, the second one’s talking about stuff that looks difficult, but wouldn’t if you’d just read ahead.)
Also, you generally don’t have to wait until later for insight into whatever you’ve failed to understand… More often than not, insight strikes even as you’re adding the cards.
http://www.gwern.net/Spaced%20repetition#abstraction
I’d suggest making it so you get scored by the number of people you save, but the game is long enough that luck doesn’t make a difference.
I wonder if it would be a good idea to give it multiple scores. For example, lives saved, life-years saved, quality-adjusted life-years saved, etc.. This way, you won’t have as many problems with people disagreeing with the scoring system.
Alternately, you could just have it so you could change the scoring mode in the options part. It would also act somewhat as a difficulty setting. It would get harder when you have to weigh a destitute child, who will live longer, vs. a middle-class adult, who will be happier.
Reading that, it occurred to me that in all the computer games I’ve played, it is possible to totally succeed. Kill every monster, pick up every piece of treasure, solve every puzzle, gain the perfect score. (I have not actually played all that many, but I also read some of the games press and it seems to be the usual case.) Even the old arcade games fit the pattern: you can’t win them, only endure as long as possible, but until then your goal is to impeccably handle every challenge that comes at you.
Are there any games in which this is deliberately made impossible? For example, PeerInfinity’s suggestion of trying to save drowning people, in a scenario that makes it impossible to save all of them (and which is varied every time you play the game, so you can’t simply search out the optimal walk-through). Military operations in a city, fighting door-to-door, with innocent civilians everywhere whose lives matter in game terms.
I’m not sure the suggestion of a game in which one cannot ‘get the top score’ makes sense. It seems contradictory - ‘is there an optimal path through the game which is not the optimal path through the game?’
Can you have games where the ‘path’ to a top score, the optimal play, varies from game to game? Sure. Not every game carries it to quite the extent of Nethack, but most do it to some extent. Non-random games like go or chess are generally the exception, and they can be trivially randomized. But each specific game can be seen as ultimately deterministic: given the output of the random number generator this time, the ideal path is such-and-such. You, the player, may not know it, but that’s your fault.
Can you have games which deceive the player about what the best possible score is? Sure. The original Donkey Kong promises that you can play indefinitely; but go too high and the game will always crash. The upper bound is not where one thought it was. Or there are political games in which one tries to prevent 9/11 (IIRC); of course, the game must sooner or later defeat you, like those old arcade games.
What would it mean for a game to have scores players couldn’t reach? If in Mario, there is code to paint a picture of a 1-UP on a corner of the screen surrounded by unbreakable blocks, then in what sense is the player missing out on 1k points (or whatever). If a cut scene depicts a hostage dying, then how ‘could’ I have saved it? What if I can choose between a cut scene depicting hostage A dying, and hostage B? What if it’s in-game, and there’s a timer or rescuing A triggers the death of B?
Or what if there is a trove of 1 million points coded in, but the only access is to type in a true contradiction? Would the world record holder for Mario really be missing 1 million points off his score just because he can’t come up with one? (Yes, there is a number equal to his score+1 million; but there’s an infinite number of integers. What makes score+1 million special? All the lower number are special because it’s possible to manipulate a given blob of code to display characters we interpret as those lower numbers; but we can’t get it to emit any images of higher numbers, and that’s that.)
It can make sense if the game does not have a one-dimensional score. World Of Warcraft, The Sims, Second Life, D&D… Life itself, for that matter.
If you choose one sub-game, then there’s optimal play for that; if you switch between them, then there’s still optimal play, it’s just you need to weight it if there is no canonical ultimate score (just like with utilities).
If there aren’t any scores or sense of progress at all, I question whether it’s a game at all, or whether it merely bears a Wittgensteinian family resemblance. If you and I push around piled pieces on a Go board just for the pleasure of watching the piles build and collapse and form swirling patterns, we’re doing something entertaining (maybe) but who would call it a game? To call life itself a game is to either commit a tired weak metaphor, or to drain the word game of all meaning.