[yotamarker]

what triggers humor ? how can an AI recognize humor ?

[Don Patrick]

I never made a humour detection algorithm, but I did slap together a sarcasm detection algorithm, which is closely related. I've hightlighted two other computational humour projects in this blog article, the second of which had a pretty good grasp of the factors involved:
https://artistdetective.wordpress.com/2015/09/07/computers-can-not-understand-sarcasm

Personally I'd say the key factor would be to detect the degree of word sense ambiguity, but AI can't even do that well enough yet for normal purposes of disambiguation.

[HS]

Part of it may be an expansion on the double entendre principle. Where the more complete and reasonable your description of a situation is, the stronger your confidence becomes in your interpretation of the situation. The stronger your confidence is, the more surprising an alternate interpretation becomes. Surprise may be a reward mechanism for productive avenues of inquiry.

[yotamarker]

why is this so funny ?



how would an AI find it funny ?

[yotamarker]

[yotamarker]

would it have been funny, had mag not looked for her trapper keeper ?

[HS]

Social transgression is inherently risky, but not as logically obvious as a tiger, maybe our DNA has it figured out more than anything else. You know how the MythBusters would get laughing fits after witnessing the explosions they set up? Many practical joke videos are based on the same principle. They are intricate physical puns acted into existence, where first the conditions for anxiety are set up, then they are revealed to fit into largely harmless contexts, and your ability to recognize alternate interpretations gets rewarded, creating the conditions for mirth. Of course it can go the other way too, I guess that's what dark humor is.

[ivan.moony]

I remember a documentary where they filmed people on a speedy roller coaster. They were all having grins on their face, occasionally screaming in fear on more demanding segments of the route. The grins on their faces were almost identical to smiling. So I believe, a good laugh has something to do with a fear. It could be that a laugh is a feeling that happens after mastering problems and overcoming fears that become irrational once when we know how to handle their causes. Basically, a good laugh could be the same feeling as fear, with a difference that in a case of laugh we know nothing bad will happen, while in fear we experience unwanted anxiety in the same time.

[Art]

I believe that a good number of humans realize what humor is whether it be pratfalls, jokes, slapstick, puns, or innuendos, his question remains, "...How can an AI recognize humor?"
Some scripts I've seen are geared toward the obvious 'Haha' so the computer knows that the user thinks or said something funny. The same question could be asked of sarcasm or disgust or anger, etc.

For a computer/AI to know the meaning of a word does not mean that it also understands the usage, context or knowledge within the word. That is perhaps a very large hurdle for the AI to grasp and use properly. Time will tell.

[WriterOfMinds]

Here are a couple categories of humor that I thought of off the top of my head:

1. Failure. Watching other people (or yourself) mess up is funny. This is especially true if it's a failure at some low-skill task where you would have expected success ... abnormal clumsiness, outrageously dumb choices, misunderstandings, etc. There's a whole website (Fail Blog) specifically dedicated to this type of humor. If the person failing is someone you care about, then the consequences of the failure have to be fairly minor, or other emotions will start overriding the amusement.

Fart jokes can be an example of this, because letting loose a fart when you don't want to is a minor failure. It won't actually hurt anybody or subject you to serious social consequences ... but boy, is it embarrassing.

2. The absurd. This covers some of the reasons we laugh at small children and animals. E.g. it's funny when cats insist on sitting in boxes. Why? Because we wouldn't do it. They seem to be doing it for no reason. It's alien, it doesn't make sense, but it's not a "fail" because they're clearly doing it on purpose. The absurd gets funnier the farther it strays from what we would think of as "dignified." Cat sitting in box? Funny. Cat going through contortions to stuff himself into a box that's way too small? Hilarious.

The absurd is norm-breaking, but not necessarily transgressive or dangerous.

To detect these, an AI would need enough common-sense knowledge to know when somebody is failing to achieve a goal, guess the likelihood of the failure and the severity of the outcome, and know what qualifies as normal/sensible by human standards (so as to detect the absurd by negation).