This paper does not draw much on existing theories of humor. Rather, it is an exploration into humor based on those aspects of emotion for which we have already established a computational platform. The goal of this research, then, is to allow us to specify a small, computable, form of humor suitable for use in a believable agent environment.
Our thesis is as follows: Using the Affective Reasoner (hereafter AR) we have demonstrated that it is possible to capture certain important elements of stories by representing the underlying human emotions that are present in the stories (See Story-morphing in the Affective Reasoning paradigm: Generating stories semi-automatically for use with ``emotionally intelligent'' multimedia agents, Elliott, et al., Autonomous Agents 1998) Some stories are also funny stories. Certain kinds of funny stories share similarities in their emotion representation. By extracting this emotion representation, and using story-morphing techniques explored in earlier work, we should be able to generate new stories, or at least new forms of existing stories, that are also funny. Additionally, we should also be able to answer some preliminary questions about the nature of certain structural components of humor by varying this structure under computational control.
The motivation for this work is that interaction with human-like agents can be made more pleasurable if the agents are able to sometimes amuse us with humorous invention. This may be particularly important for agent-helpers such as those in automated tutoring systems: In documenting the persona effect, [James Lester, et al., The Persona Effect: Affective Impact of Animated Pedagogical Agents , Proceedings of CHI '97, Conference on Human Factors and Computer Systems] Lester and his group have shown that engagement has been an important component in developing quality interaction between users and tutoring systems. The strategic use of humor is one clear way that agents can be more engaging.
To illustrate how our emotion classification techniques apply we will examine several humorous stories, break them down into a coarse emotion representation as stories --- independent of the fact that they are humorous, compare the similarities in emotion representation between them, and suggest ways in which non-humorous stories can be morphed into humorous ones using the emotion theory.
It is important to note that for the purposes of humor generation, and construction of emotion fabric underlying funny stories, it is not necessary that this be the correct interpretation. Any correct interpretation will do as long as new stories generated from the delineated structural components are sound.
In keeping with our work on the emotion fabric of stories we will use only stories believed to be true.
Our approach to story representation uses several points of leverage. This leverage is based on the ideas that human emotion can be faithfully represented at a descriptive level suitable for story representation; that the communication of human emotion is often indirect, and inaccurate; and that humans have a natural tendency to project plausible interpretations onto social interactions and observed emotion expressions. Following are the most salient points of this leverage on which we hope to capitalize.
First, the underlying descriptive emotion theory that we have used in the Affective Reasoning project, based largely on the Ortony, Clore, and Collins model, has been exercised, and extended, over the years in ways useful for describing human emotion narratives. That is, where the theory has not been sufficient to describe some scenario it has had to be extended; where it has had a component that has not been necessary in describing scenarios it has been reduced. In this ongoing process we have described portions of about six hundred different scenarios. Through this transcription process we have developed a somewhat canonical form for the representation of the human emotion component of stories.
Second, in previous work we have given many examples illustrating that humans can have multiple, even conflicting, emotions stemming from the same emotion-eliciting events (e.g., see Elliott, The Affective Reasoner, 1992) For example, suppose that one unexpectedly comes by desperately needed money, but it is inherited because a much admired, and loved, favorite uncle dies. News of the uncle's death might simultaneously lead to distress over the loss of a loved one, joy over newly-inherited wealth, relief over escaping from creditors, anger that death came to the family, guilt over feeling happy about the benefits derived from the uncle's death, remorse over not having spent more time with the uncle recently, etc. What this means for story-morphing is that, with respect to interpretations of emotion eliciting situations, we are not so burdened with the ``foolish consistency'' that is the hobgoblin of plot structures.
Third, humans seem inclined to project not only anthropomorphic qualities to most everything, but also social qualities to anything that has agent-like attributes and interacts in ways perceivable as supporting a community. Given a set of agents engaged in dialog, or characters interacting in a script, people naturally attribute human-like motivations to the agents/characters for their actions. We humorously refer to scripted presentations of our personality-rich agents as a multimedia Rorschach test -- but there is a strong element of truth in this line of thinking.
Fourth, facial expressions, music, and inflections in spoken text seem to trigger explanatory mechanisms --- that there is some sort of cognitive dissonance that is quelled only when observers abductively attribute narrative qualities to sequenced expressions, and other social cues, on which they can hang the complex fabric of the interaction.
CATEGORY LABEL AND EMOTION TYPE
|Well-Being||appraisal of a situation as an event||joy: pleased about an event
distress: displeased about an event
|Fortunes-of-Others||presumed value of a situation as an
event affecting another
|happy-for: pleased about an event desirable for another
gloating: pleased about an event undesirable for another
resentment: displeased about an event desirable for another
jealousy*: resentment over a desired mutually exclusive goal
envy*: resentment over a desired non-exclusive goal
sorry-for: displeased about an event undesirable for another
|Prospect-based||appraisal of a situation as a prospective
|hope: pleased about a prospective desirable event
fear: displeased about a prospective undesirable event
|Confirmation||appraisal of a situation as confirming
or disconfirming an expectation
|satisfaction: pleased about a confirmed desirable event
relief: pleased about a disconfirmed undesirable event
fears-confirmed: displeased about a confirmed undesirable event
disappointment: displeased about a disconfirmed desirable event
|Attribution||appraisal of a situation as an accountable
act of some agent
|pride: approving of one's own act
admiration: approving of another's act
shame: disapproving of one's own act
reproach: disapproving of another's act
|Attraction||appraisal of a situation as containing
an attractive or unattractive object
|liking: finding an object appealing
disliking: finding an object unappealing
|compound emotions||gratitude: admiration+joy
|compound emotion extensions||love:admiration+liking
Syntax note: In general we will use parentheses, (), for function arguments, and square braces, , for parenthetical comments where appropriate in the examples.
The parallel structure of these two stories is relatively obvious, and we offer one analysis here. We can find direct mappings for the following labelings within each of the stories. Step-by-step, using the existing emotion theory underlying the AR, we have the following:
P is held by Agent-C and Agent-A.
Act-1: Agent-C violates P; E1-a-shame(Agent-C,P); E1-b-mixed-emotions(Agent-A,P,G,R1) [E1-b-i-gloating(Agent-A,P,R1)/E1-b-ii-anger(Agent-A,[at]Agent-C,G,P)]
Act-2: E2-shame(Agent-C,P), intensity is increased by Act-2(Agent-A); relationship [Agent-C --> R2-adversarial --> Agent-A, with respect to P] is established if not already present.
Act-3: Agent-A violates P; E3-a-shame(Agent-A); E3-b-gloating(Agent-C over Agent-A)
Each of these cases can be handled, ad hoc , without changing the computational framework in the AR. First, that one is an authority figure is in this case really only with respect to the principle under consideration: the senior graduate student is an authority on sending return mail; the Dean is an authority on meetings. Any time a person feels it their job to criticize others for the violation of a principle, for the purposes of this type of humor we can consider them to consider themselves an authority on that principle.
In the case of adversarial relationships being established with respect to certain principles, this is straightforward. In the AR, relationships have been more-or-less arbitrarily established (although dynamically set) anyway. For the purpose of studying humor then, setting the relationships so that an agent is in an adversarial relationship with another agent need not be specific to a certain principle, but can continue to be, instead, global, thus encompassing that principle. From a future-design plausibility standpoint, it would seem that all fortunes-of-others-emotions dependent on that particular principle, and that agent, could simply use whatever setting of the adversarial relationship was desired, ignoring the setting at other times. In this way, should V's guns be stolen, K would gloat, but should V's television be stolen K would feel pity.
Here we can also see anger replacing shame, that is, instead of E1-a-shame(Agent-C,P) we have E1-a-anger(Agent-C,P2,G2) where G2 is some goal thwarted by Agent-A as a consequence of Agent-C having violated Agent-A's principle P, and P2 is Agent-C's principle, "Do not thwart my goals on the basis of principles I do not believe in."
For example many non-Christians gloated over the downfall of Jimmy Swaggart, even though they themselves might not have held the principles he was embarrassed to have violated. However, they felt that through the political arena their own goals were being blocked through Swaggart's actions based on their own violation of Swaggart's principles.
There are many examples that have a direct parallel to the above situations, and which are funny. We can learn more by looking at funny stories which touch on the gray areas of such representation. To this end we present two more stories, which although they are different on the surface, nonetheless seem to be funny because they share the same underlying emotion structure.
Finally, after a week we managed to get the motorcycles running somewhat reliably. We rode them out on a tour of the hills, vainly attempting to be "impressive" in a way that we simply had not been able to capture for all our efforts. Finally, stopping in the gravel parking lot of a woodland area we looked at each other defeated. "It" just had not happened. We were hopeless.
In frustration Richard said, "Watch this!" and attempted to spin a doughnut with his back wheel. Because of the gravel his front wheel slipped as well, the bike went flying out from under him, and he landed face down in the rocky dust. He got up and we inspected the damage. On his bike, the gear shift was crazily twisted around the engine, his brake and clutch grips were curlicued like munchkin shoes from the Wizard of Oz around his handle bars, and his fender was ripped loose. Richard had not fared much better: both palms were bloody and imbedded with small stones from the parking lot, and his face was scraped.
"Wow," I said.
"Now that really WAS cool!" Richard said.
We both burst out laughing. We laughed so hard that we cried, and our stomachs began to hurt. That we could reach such heights of Nerd-dom was really one of the funniest things I have ever encountered.
This vignette is interesting because although on the surface it is different from the two above, from an emotion-humor perspective it might be seen as similar. Here is one analysis suggesting that a common structure with the preceding examples is plausible.
Act-1 = Being nerds over the course of a week.
P = Not being nerds.
Agent-C = Richard-and-me [Nerds].
Act-2 = Inflicting the standard of Not-nerdness
Agent-A = Richard-and-me [Judges of Nerds].
R1 = Adversarial relationships with ourselves -- conflict over wanting to not be nerds, but being nerds. That is, that it is not acceptable [Agent-A] to simply relax and be nerds [Agent-C] indicates that there is motivation, P, for calling us to task for our nerdness.
R2 = Adversarial relationship from Agent-C [nerds] to the part that believes we should not be nerds Agent-A [believing should be not nerd] resulting in the increasing stress from continuing to lose the "not nerd" battle.
G = Have a moment of Not-nerdness through ripping off a cool doughnut
with the motorcycle; thwarted.
Act-3: Through perspective change, Agent-A [Judges of Nerds] is seen to have violated P. Agent-C [Nerds] can enjoy this at their expense.
Additionally, had we not cared about our Nerdness, we might have continued blissfully to be Nerds and the story would not have been funny (except perhaps in a different way -- see the Professor G story below).
Next let us examine a fourth story where the principle, P, at first seems hidden.
The faculty judges came in, serious minded, and formidable. I went over all my preparations one last time, and everything seemed in order. Then, minutes before my defense began, a team of workmen unloaded equipment from a truck in the parking lot right next to our building, started up a diesel compressor, and began to dismantle the sidewalk with jackhammers, eight feet from the window of the defense room. I was mortified. It had taken months to get the committee in one room at one time. It was not going to be possible to reschedule the event. The show had to go on.
Because the talk was on emotion theory I hastily drew up an alternate emotion-scenario diagram that was based on dissertation defenses and bulldozers. When I presented it as one of the first slides in my talk -- pointing out the futility of all my efforts to prepare so carefully -- it was funny. We all laughed about it."
Agent-A = Me as authority on being prepared.
Agent-C = Me as preparer of defense.
P = I will handle all situations that arise during the defense through expert preparation.
Act-1: Not handling all contingencies.
One important concept is in looking carefully at how, in our theory, we delineate tragedy from humor. Clearly there is a connection between these two, and a sometimes delicate balance exists between having a situation fall into one category or the other. Much humor involves emotions that can easily turn to sorrow, pity, chagrin, and other negative emotions. What might we say about this here?
First, there are a number of ways to turn the above scenarios into situations that not only are not funny, but are not pleasant either. Or, we can leave them as stories, that are funny for the same reason, but change the degree or character of that funniness. Looking at this can give us clues into constraints that we may have to place on the synthesis of such situations.
For example, if the Dean is highly respected, and the principle of Being on time is shared by all, then being uncomfortable with the Dean's tardiness would prohibit humor -- unless of course he was able to remove that discomfort by laughing at himself.
The use of traditional comic tension and release is compatible with our approach, and in fact may give us some new insight into its tweaking. We can, for example, increase the tension of this story, building to a greater comic release, if we tell it such that the chorus gloats over the authority before the authority discovers that she or he has violated the principle.
To wit: if the Dean is seen to be hurrying across the street, and chugging up the stairs, while the faculty is enjoying themselves at his expense knowing that he will be late, this might increase the humor. If the clock in the meeting room is set back so that the Dean believes he is on time, it may become funnier while we wait for his discovery. If a faculty member leaves the room to purposely arrive even later and is dressed down by the unknowing Dean, we add additional tension. In each case the projected embarrassment is increased, while the real embarrassment has been delayed.
Similarly if the senior graduate student is allowed further pontification without knowing of his own gaff, it becomes funnier.
Suppose we just increased the importance of my Ph.D. defense so that desire was very high. if I did not pass my wife was going to divorce me, I was going to lose my inheritance (hah!), and my advisor was going to retire and leave for the jungles of Borneo on a permanent spiritual quest. In this case the bad break of having jackhammers threaten my ability to pass the exam would be seen only as sad, not funny. Those in the audience might feel badly on my behalf, had they been made aware of the details, but would likely not have found it humorous.
If the victim is not embarrassed it can be funny for a different reason which we might call "oddball humor," best delivered deadpan. This is not something with which our reasoning system, that knows almost nothing about the world other than emotions, would do very well. However, in each of the examples given above this no understanding of how the world works, outside of reasoning about emotions, seems to be required.
If I had not been bothered by the jackhammers, and had given the talk without seeming to even notice them, this might have struck some of the chorus members as odd , and hence funny. We can argue however, that the original quality of the humor is lost because one of the essential emotions -- embarrassment in this case -- is missing, and only serendipitously replaced by another form of humor.
Here is a different example. I met a sprightly old AI professor, "Professor G.", when interviewing for graduate schools. He came by while I was sitting on a sofa in the Computer Science office, and we had an animated discussion about various AI topics. Suddenly he jumped up and said, "Here, I have an extra copy of the paper, I'll get it for you." I followed him around the corner where he leaned hard against a door after unlocking it, and managed to open it only about six inches. He reached inside and retrieved a stack of papers from which he gave me my copy. Upon inquiry he told me that, yes, it was his office, no there was nothing wrong with the door -- it was just that it was "a little messy" inside, a stack of boxes had fallen against the door, and as far as he could remember he had not been in his office for about a year.
The professor was not the least bit concerned about this, and the whole thing struck me as very odd, and very funny.
This is an original instance of "oddball humor." Can we morph it into the previous kind of humor we have been discussing? Yes, but only by adding the necessary components. To wit: Professor G. argues at lunch that he does not for a minute buy the proposition of eccentric professors. As far as he concerned this is just a "lack of professionalism." Later he is "discovered," chagrinned, in his attempt to retrieve something from his office. This is funny, but now for because for the "Citizen's Arrest" reason. As discussed above, with this type of humor we can increase the intensity of the humor by tweaking some of the variables we have associated with emotion intensity: If Professor G. works hard at avoiding discovery, his embarrassment is increased.
When there is no resentment present, or implied resentment because of social status, then the balance between pity for the victim, and gloating (or relief) is altered, and the situation might not be funny. In the case of the Ph.D. defense, if we remove the sense, on the part of the chorus, of the highly principled environment, wherein Ph.D. students have all manner of principles which they are made to uphold, the situation becomes not funny.
The Ph.D. defense does not appear, on the surface, to have the element of resentment on the part of the chorus members. But let us look at this more closely. The chorus, each of them an academic, and former graduate student, has experienced the somewhat stodgy nature of academia, which has a tradition, dating back almost to the dark ages, of holding neophyte scholars up to scrutiny with respect to all manner of principles. By making fun of myself for preparing so carefully for the academic presentation only to be defeated by jackhammers I set myself up for us all to laugh at as the bearer of those principles we have each had to endure. If this seems far-fetched, let us consider this: would an audience of plumbers or high-school coaches have found this to be humorous? Probably not.
In the case of the motorcycles the question arises as to whether or not there is (a) shame originally present for the chorus members, and (b) shame present for the victims. In both cases this is easy to mark because frustration was clearly present, and within the emotion theory we are using for stories frustration is a form of anger. Anger in turn is not possible without someone being held accountable for the violation of a principle. It is the violation of this principle, that of "not being nerds" which led to shame, or embarrassment. This was present during the week preceding the humorous incident, and was also present at the time of dumping the motorcycle in the gravel parking lot.
In the Ph.D. defense, this situation does not appear to have initial shame, but I wonder -- is it funnier because the chorus has each experienced the highly principled academic atmosphere of which we are making fun? The point is that it was the silliness of my preparations which were funny. This could be made more clear if we consider that had I prepared even more exotically -- that is, tried to beat the system -- say by bringing in some sort of low-volume multimedia examples only to discover that they were useless next to the Jackhammer.
That these are instances of situational humor, and not dependent on the timing, and cleverness, of much humor is clear. Even if any four of these stories are told in reverse order, though diminished, they are still funny, and are still stories. This suggests that this type of humor is not dependent much on the sequencing, or on an element of surprise:
"The Dean came late to a faculty meeting and was embarrassed about it. We all thought this was pretty funny because just two days before he had lectured the rest of us about being on time."
"A senior graduate student mistakenly sent mail to all of the people in computer science by using "R" instead of "r". The funny part of this was that in his message he was berating a first-year student for doing exactly the same thing.
"Richard dumped his motorcycle in the parking lot, confirming beyond all hope of redemption that we were hopeless nerds. Thinking about how hard we had worked to be "cool" with these motorcycles, and all the effort we had put into the endeavor, we just burst out laughing at how ridiculous we were.
"At my dissertation defense there were jackhammers and a compressor running right outside of the window. What made this funny was that in my anxiety I had prepared for most every other contingency that could arise."
Situations are funny in a different way if the victim can laugh as well -- but is not necessary. The victim can even be angry, or sad, and it can still be funny.
Conversely, when humor of this "citizen's arrest" type is deemed, ad hoc, to be present, will we always find such AR tokens present?
In addition to the emotion theory elements, what other tags do we require? This is important because while our AR agents do a good job at manipulating emotion content, they have greatly impoverished representation of other elements in the world, and NO understanding of what those elements represent. Hence each of these additionally elements must simply be treated as labels, without any inference manipulation.
Here we presented only a minimal sample of the kind of humor that can be represented through emotion theory. A number of additional structures for humorous situations have also been explored.
Importantly, by reducing each human scenario to just Affective Reasoner elements, we able to make a decent case that situational humor which is dependent on the desires and principles of the characters (and explictly not on and understanding of the world) is computable within an already existing framework.