Minsky, in his paper, posits that there is a system of censors in the brain, to prevent incorrect action. He says that the sensation of humor is produced by the creation of a new censor, or the reinforcing of an existing one. It’s not just learning; it’s the learning of a particular kind of knowledge. It’s possible for the steering subsystem to reward phenomena in the learning subsystem, right? In this case, the steering subsystem doesn’t need to understand what the content of the censor is, just that there is one. And it rewards the learning subsystem with a little jolt of reinforcement. We call this kind of reinforcement “humor”.
Anyway, this theory seemed plausible to me, for several reasons (I don’t know how much of this is in the paper, and how much I came up with since then.)
First, I think censors can be localized to the prefrontal cortex, or at least the frontal lobe. Prefrontal damage or hypoxia makes you both rude and unsafe, suggesting that safety rules and politeness rules go in the same place, and that place is above the eyeballs.
Second, optimality theory is a well-regarded theory of phonology (speech sound selection) where all possible pronunciations of a sentence advance through a list of rules until only one is left, which is then pronounced. The theory of censors is an extension of this mechanism to action selection. I’m not claiming that it’s all of action selection; clearly we need positive, reward-anticipating mechanisms too. But I think there’s a place for a multitude of censors that are constantly watching, preventing you from making a boo-boo, in either action or reasoning.
Third, I’ve looked at a lot of jokes, and it’s easy to see what the censor-worthy behavior is for any given joke. Of course that might just be confirmation bias on my part.
Fourth, it provides an explanation that explains why jokes stop getting funny due to repetition—once the censor is established, there’s no need for reinforcement. That’s why bathroom humor is hilarious when you’re a little kid, but not when you’ve become a grownup and no longer need reinforcement for the “no potty talk” censor.
Fifth, an explanation in terms of surprise or physiological arousal is over-general. Here’s an example of a story: “I was crossing the street in the middle of the block and forgot to look left. Suddenly I realized there was a car careening toward me with screeching brakes. Fortunately the car managed to stop before I was struck, and everything was fine.” It’s got surprise, it’s got arousal, it’s got sudden relief. But it’s not funny.
First off, I’m very sympathetic to a SoM-like picture where we learn a giant collection of abstract patterns of the form “When your thought pattern-matches to X, then it’s a bad idea to do Thing Y which you would otherwise do.” (Among other patterns.) That seems similar to what you’re calling “censors”, right?
I don’t particularly associate learning those patterns with laughter though. Like, I think back to when I was a kid practicing Olympiad-style math problems in high school: Early on I learned from experience “If you see a problem of this general form, try applying the Cauchy-Schwarz inequality”, and then somewhat later I would learn from experience an exception to that rule: “If you see a problem of this more specific form, then you’ll be inclined to try Cauchy-Schwarz, but don’t bother, it won’t work.” No laughing was involved during the process of learning that; it probably just looked like a kid sitting quietly in his room doing deliberate practice. Lots more examples like that: I know from experience hundreds of questions that I shouldn’t bother googling because I won’t find anything, and dozens of topics that I shouldn’t mention in the vicinity of certain relatives because they get very sensitive, and computer icons that I shouldn’t click in certain situations, and on and on. I don’t think any laughter was involved in learning or reinforcing any of those “censors”.
If the “creating or reinforcing censors” pattern isn’t sufficient for humor, is it necessary? Actually, maybe it is!! I wasn’t thinking this way last week when I wrote this post, but I was commenting here that maybe it’s impossible for a single conscious thought to create both Ingredient (A) an Ingredient (B). So instead, humor (and conversational-laughter) require thinking an Ingredient-(A) thought and an Ingredient-(B) thought in immediate succession, and maybe even flipping back and forth a couple times within the space of a second or two. And maybe the only way for that to happen in practice is that you’re thinking about a single “thing” but flipping back and forth between two different frames / analogies for that thing, one of which (call it “Frame 1”) triggers Ingredient (A) and the other of which (call it “Frame 2”) triggers Ingredient (B).
What does that have to do with the “censor” pattern? Well, during this process, your brain is incidentally learning / reinforcing the higher-level pattern that: “if something pattern-matches to Frame 1, then it’s possible to flip that thing to Frame 2, and vice-versa”. So in the future, you could be thinking a thought that seems to be a good plan, but this higher-level pattern jumps in and re-frames / re-analogizes it in a way that makes it feel like an aversive / bad plan. So if you’re laughing from humor or conversation, you’re incidentally learning or reinforcing a censor, I claim.
Fifth, an explanation in terms of surprise or physiological arousal is over-general. Here’s an example of a story: “I was crossing the street in the middle of the block and forgot to look left. Suddenly I realized there was a car careening toward me with screeching brakes. Fortunately the car managed to stop before I was struck, and everything was fine.” It’s got surprise, it’s got arousal, it’s got sudden relief. But it’s not funny.
When I think of that scenario, I feel like there’s no Ingredient (B). Or if there is, it’s not close enough in time to Ingredient (A) for the two to temporally overlap in the hypothalamus / brainstem. Like, I imagine almost getting killed jaywalking but the car stops inches from my legs. Standing safe on the sidewalk five seconds later, consciously I’d say “I am safe now”, but emotionally I don’t feel safe, or relaxed, at all. I would continue to feel extremely on-edge / jumpy / traumatized for at least minutes and possibly hours after that. Maybe the feeling gradually fades away and (B) gradually replaces it, but I would never have a juxtaposition of simultaneous (A) & (B) in my hypothalamus, I think.
Minsky, in his paper, posits that there is a system of censors in the brain, to prevent incorrect action. He says that the sensation of humor is produced by the creation of a new censor, or the reinforcing of an existing one. It’s not just learning; it’s the learning of a particular kind of knowledge. It’s possible for the steering subsystem to reward phenomena in the learning subsystem, right? In this case, the steering subsystem doesn’t need to understand what the content of the censor is, just that there is one. And it rewards the learning subsystem with a little jolt of reinforcement. We call this kind of reinforcement “humor”.
Anyway, this theory seemed plausible to me, for several reasons (I don’t know how much of this is in the paper, and how much I came up with since then.)
First, I think censors can be localized to the prefrontal cortex, or at least the frontal lobe. Prefrontal damage or hypoxia makes you both rude and unsafe, suggesting that safety rules and politeness rules go in the same place, and that place is above the eyeballs.
Second, optimality theory is a well-regarded theory of phonology (speech sound selection) where all possible pronunciations of a sentence advance through a list of rules until only one is left, which is then pronounced. The theory of censors is an extension of this mechanism to action selection. I’m not claiming that it’s all of action selection; clearly we need positive, reward-anticipating mechanisms too. But I think there’s a place for a multitude of censors that are constantly watching, preventing you from making a boo-boo, in either action or reasoning.
Third, I’ve looked at a lot of jokes, and it’s easy to see what the censor-worthy behavior is for any given joke. Of course that might just be confirmation bias on my part.
Fourth, it provides an explanation that explains why jokes stop getting funny due to repetition—once the censor is established, there’s no need for reinforcement. That’s why bathroom humor is hilarious when you’re a little kid, but not when you’ve become a grownup and no longer need reinforcement for the “no potty talk” censor.
Fifth, an explanation in terms of surprise or physiological arousal is over-general. Here’s an example of a story: “I was crossing the street in the middle of the block and forgot to look left. Suddenly I realized there was a car careening toward me with screeching brakes. Fortunately the car managed to stop before I was struck, and everything was fine.” It’s got surprise, it’s got arousal, it’s got sudden relief. But it’s not funny.
Thanks, that was another really helpful comment.
First off, I’m very sympathetic to a SoM-like picture where we learn a giant collection of abstract patterns of the form “When your thought pattern-matches to X, then it’s a bad idea to do Thing Y which you would otherwise do.” (Among other patterns.) That seems similar to what you’re calling “censors”, right?
I don’t particularly associate learning those patterns with laughter though. Like, I think back to when I was a kid practicing Olympiad-style math problems in high school: Early on I learned from experience “If you see a problem of this general form, try applying the Cauchy-Schwarz inequality”, and then somewhat later I would learn from experience an exception to that rule: “If you see a problem of this more specific form, then you’ll be inclined to try Cauchy-Schwarz, but don’t bother, it won’t work.” No laughing was involved during the process of learning that; it probably just looked like a kid sitting quietly in his room doing deliberate practice. Lots more examples like that: I know from experience hundreds of questions that I shouldn’t bother googling because I won’t find anything, and dozens of topics that I shouldn’t mention in the vicinity of certain relatives because they get very sensitive, and computer icons that I shouldn’t click in certain situations, and on and on. I don’t think any laughter was involved in learning or reinforcing any of those “censors”.
If the “creating or reinforcing censors” pattern isn’t sufficient for humor, is it necessary? Actually, maybe it is!! I wasn’t thinking this way last week when I wrote this post, but I was commenting here that maybe it’s impossible for a single conscious thought to create both Ingredient (A) an Ingredient (B). So instead, humor (and conversational-laughter) require thinking an Ingredient-(A) thought and an Ingredient-(B) thought in immediate succession, and maybe even flipping back and forth a couple times within the space of a second or two. And maybe the only way for that to happen in practice is that you’re thinking about a single “thing” but flipping back and forth between two different frames / analogies for that thing, one of which (call it “Frame 1”) triggers Ingredient (A) and the other of which (call it “Frame 2”) triggers Ingredient (B).
What does that have to do with the “censor” pattern? Well, during this process, your brain is incidentally learning / reinforcing the higher-level pattern that: “if something pattern-matches to Frame 1, then it’s possible to flip that thing to Frame 2, and vice-versa”. So in the future, you could be thinking a thought that seems to be a good plan, but this higher-level pattern jumps in and re-frames / re-analogizes it in a way that makes it feel like an aversive / bad plan. So if you’re laughing from humor or conversation, you’re incidentally learning or reinforcing a censor, I claim.
When I think of that scenario, I feel like there’s no Ingredient (B). Or if there is, it’s not close enough in time to Ingredient (A) for the two to temporally overlap in the hypothalamus / brainstem. Like, I imagine almost getting killed jaywalking but the car stops inches from my legs. Standing safe on the sidewalk five seconds later, consciously I’d say “I am safe now”, but emotionally I don’t feel safe, or relaxed, at all. I would continue to feel extremely on-edge / jumpy / traumatized for at least minutes and possibly hours after that. Maybe the feeling gradually fades away and (B) gradually replaces it, but I would never have a juxtaposition of simultaneous (A) & (B) in my hypothalamus, I think.