Daydreaming in Humans and Computers

DAYDREAMING IN HUMANS AND COMPUTERS

Erik T. Mueller Michael G. Dyer

Artificial Intelligence Laboratory Computer Science Department University of California Los Angeles, CA 90024

Abstract

This paper examines daydreaming: spontaneously re­ calling or imagining personal or vicarious experiences in the past or future. The following functions of daydreaming, for both humans and computers, are discussed: support for processes of creativity, future planning and rehearsal, learning from successes and failures, emotion modification, and motiva­ tion. A computational theory of daydreaming is being imple­ mented and tested in a computer program called DA YDREA­ MER. A prototype version of DAYDREAMER which produces several daydreams (in English) is currently running.

I. INTRODUCTION .

An intriguing aspect of humans is that we spend much time engaged in thought not directly related to the current si­ tuation or environment. This type of thought , usually called daydreaming, involves recalling past personal or vicarious ex­ periences and imagining what might have happened in the past or what might happen in the future. On the face of it, daydreaming may seem like a useless distraction from a task being performed. However, since daydreaming is so pervasive in human experience (Pope, 1978) (Singer, 1975), we must ask ourselves what purposes it might serve, both for humans and intelligent computer systems.

We hypothesize several important roles of daydream­ ing in human and computer cognition. These roles will be dis­ cussed in the context of the following episode and daydreams which might result from such an episode:·

NUART-EPISODE

Last night I went alone to the Nuart Theater to see a film. This actreu whom I've always had a cru·h on also came alone and happened to ·It down near me. I recognized her and started a conver1atlon. At one point, she mentioned the obvlou·ly well-known director of the film we were about to see and I didn't know the name. I wu embarru·ed. I finally uked her If ·he'd !Ike to go have a drink after the film. She eald she had other plan·.

DAYDREAM!

I feel embarru·ed tor not knowing who the director wu. I should've pretended to know who the director wu. In the fu­ ture I'll read up on the film before I go.

DAYDREAM2

I'm disappointed that she didn't want to go out with me. She might change her mind If I were to uk her out again. I would call her up, but I don't have her telephone number. I should've uked for her telephone number when I had the chance. I Imagine that I find out her telephone number, call her up, and ·he accept. my offer tor a date. I reiJI'et not hav­ Ing uked for her telephone number. But ·he'· a movie star and out of my league.

The work reported here was made possible in part by a grant from the Keck Foundation. The first author was also supported in part by an Atlantic Richfield Doctoral Fellowship and the second author by an IBM Faculty Development Award.

· These daydreams are a composite of several daydreaming protocols we obtained from subjects via immediate retrospec­ tive reports (see Varendonck, 1921, pp. 215-216; Ericsson & Simon, 1984).

DAYDREAM3

I'm disappointed that she didn't want to go out with me. I Imagine that ·he accepted my offer and we ·oon become a pair. When she hu to do a film In France, I leave my Job and travel there with her. As we continue to travel, my career Buffers. I become unhappy and feel unfulfilled. She lo·e· In­ terest In me, because I have nothing to offer her. It'· good that I didn't get Involved with her, becau·e It would've led to dl·uter. I feel le·· disappointed that she turned me down.

DAYDREAM4

I'm angry that she didn't want to go out with me. I Imagine that I pur·ue an acting career and become a star even more famou· than ·he 1·. She remembers meetlns me a long time ago In a movie theater and calla me up. I'm glad ·he admits ·he wu wrong about me. I go out with her, but now ·he hu to compete with many other women tor my attention. I even­ tually dump her.

ll. SUPPORT FOR CREATIVITY

A relationship between daydreaming and creativity has often been suggested; e.g., see (Freud, 1908). We believe that the following attributes of daydreaming make this mode of thought useful as a form of 1upport for creativity:

Relazation of con1traint1: Daydreaming explores possibilities which would normally not have been pursued. Some of these possibilities, while perhaps unlikely at first glance, may lead to a new and useful solution to a problem. This is similar to the idea of "brainstorming" (Osborn, 1953).

Fortuitou1 analogy recognition: While daydreaming about one thing, it is possible to stumble into a solution to another prob­ lem. That is, daydreaming often includes the serendipitous recognition of analogies among problems. For example, sup­ pose that the daydreamer is leafing through an alumni direc­ tory from the college he attended and notices that the tele­ phone number and address of a famous astronaut is listed. He may then experience the following daydream:

DAYDREAMS

I thought famous people didn't list their telephone numbers. I am reminded of the actre·· I met. It I could find out where the actreBB went to college, and get an alumni directory from that college, I might be able to find out her telephone number. I'll have to ftnd an article about her somewhere.

Revi1ion: Daydreaming occurs in the context of a self­ modifying episodic memory (called a "dynamic memory" by Schank, 1982). Each time a problem is examined, new infor­ mation may be available that will enable a better, different, or more creative solution. For example, in DA YDREAM2 the daydreamer does not figure out how to find out the movie star's telephone number, while in DA YDREAM5 after receiv­ ing new information he is able to move one step closer to a solution to this problem. The daydreamer now has more ideas about how to contact the movie star, and thus future at­ tempts have more possibilities from which to proceed.

Motivation: Daydreaming provides the motivation to create in the first place. For example, a daydream which anticipates aesthetic satisfaction or praise from others upon a successful creation provides a strong motivating factor to produce that creation.

ln8piration: Finding an initial solution to a problem through daydreaming may inspire one to flesh out that solution. DAY­ DREAM5 is an example of how daydreaming may spur new interest in a problem.

Daydreaming is often plot-oriented and thus is ideally suited to story invention. For example, the plots of DAY­ DREAM3 and DAYDREAM4 resemble those of several recent motion pictures. Daydreaming can provide initial ideas for a story and, once the story is begun, can suggest continuations and further possibilities.

m. FUTURE PLANNING AND REHEARSAL

Daydreaming is often concerned with planning for the future and rehear.al (Singer, 197 5 ). Daydreaming may gen­ erate future plans which are incorporated into memory. For example, in DA YDREAM5 the daydreamer plans to find and then read a magazine article about the actress. The next time he is near a newsstand or library, he may remember this plan and look for an article about the actress.

Daydreaming often consists of the exploration of hy­ pot hetical future scenarios. This allows the consequences of possible future actions in possible future situations to be con­ sidered in advance. In this way, advantageous actions may be employed over less advantageous ones. For example, in DAY­ DREAM3 the daydreamer considers the results of going out with the actress, which in this case he deems to be negative. Simulating hypothetical future scenarios through daydreaming thus assists in decision-making and also provides an adaptive safety mechanism to avoid undesirable courses of action.

Daydreaming aids not only in making life choices, but also in choosing among detailed alternative methods for carry­ ing out a plan in the face of various contingencies. For exam­ ple, before a job interview one may anticipate questions of the mterviewer in order to prepare a good response for each ques­ tion. Rehearsing future actions through daydreaming may help increase the accessibility of planned actions and the skill with which they can be performed.

Instead of shutting a computer down, its free time can be spent daydreaming about future tasks. Although day­ dreaming cannot guarantee the best behavior in all situations, systems with daydreaming will always have an efficiency ad­ vantage over systems without daydreaming.

IV. LEARNING FROM EXPERIENCE

Daydreaming about a recalled failure or success is often concerned with the generation of alternatives which might have resulted in a different outcome. For example, DA YDREAM1 contains the alternative of having pretended to know the famous director. Each alternative that is generated highlights a reason for the failure or success. Heider (1958) has noted that one may learn by avoiding that to which failure is attributed or by strengthening that to which success is attributed. Through the abstraction of generated alterna­ tives to planning strategies for use in future similar situations, daydreaming enables learning from failure6 and 6ucceue6. For example, in DA YDREAM1 the daydreamer may learn to read up on films before going to see them or to pretend to know something he does not know in order to maintain a po­ sitive image. In DAYDREAM2, the daydreamer may learn to ask for a person P's telephone number while talking to P if he anticipates wishing to contact P in the future and there is no other way of contacting P in the future. The next time the daydreamer finds himself in a similar situation, he will prob­ ably remember to ask for the telephone number.

Learning from an experience is not necessarily com­ plete after that experience. Nor does learning from an experi­ ence take place at one sitting afterwards. Rather, learning from past experiences is an ongoing process carried out through daydreaming. Often, this involves reinterpretation in light of experiences and knowledge that have been acquired in the meantime. This relates to the revi6ion attribute of day­ dreaming previously discussed.

V. EMOTION MODIFICATION AND MOTIVATION

Daydreaming is often concerned with the reduction or elimination of negative emotions resulting from a recalled failure experience. For example, in DA YDREAM3 the failure is rationalized and the disappointment reduced through the gen­ eration of a daydream demonstrating that going out with movie stars is undesirable. In DA YDREAM4, the daydreamer gains a form of revenge against the actress, resulting m a posi­ tive emotion. Thus a function of daydreaming is to make the daydreamer feel better. In effect, failures result in a form of ·cognitive dissonance· (Festinger, 1957) which must somehow be reduced. However, sometimes an increase, rather than de­ crease, of negative emotion results from daydreaming about a failure . An example occurs in DA YDREAM2 when the day­ dreamer regrets not having asked for the movie star's tele­ phone number. In general, daydreaming results in modification of emotional state since emotions may result from imagined as well as real events.

From a purely computational point of view, it may seem pointless to incorporate emotions of any type into a sys­ tem. As long as computers are driven by human-specified tasks this may well be the case. Independent agents, however, must find motivation6 to drive the creation and relative strength of their goals. We believe that emotions provide this motivational force. Suppose, for example, that one loses a wal­ let. In a planning system, the loss of possession of an object would create a goal on the part of the planner to repossess that object, which would in turn cause the planner to select a plan for achieving this goal. With people, however, the situa­ tion is more complex. What if one does not feel upset over the loss of the wallet! In this case it does not seem likely that effort will be expended in searching for it. That is, the activa­ tion and strength of the goal is based on the intensity of the emotional impact of an experience. Without the need to reduce the negative emotions of fear or anger, there would be no motivation to plan avoidance or retaliation. Without feel­ ings of satisfaction arising from imagined goal accomplish­ ment, there would be no motivation to achieve goals. Thus emotions both set and reflect the relative importance of goals and the amount of effort expended in planning for them.

VI. COMPUTATIONAL DAYDREAMING THEORY

The DAYDREAMER program is being constructed in order to implement and test a computational theory of day­ dreaming. DAYDREAMER takes as input simple situational descriptions, such as accidentally meetmg a movie star, and produces as output 1) actions that it would perform in given situations, and 2) daydreams, all in English. DAYDREAMER learns as it daydreams by indexing daydreams, planning stra­ tegies, and future plans into memory for future use.

DAYDREAMER is composed of:

• an emotion component, in which daydreams initiate, and are initiated by, emotional states arising from goal outcomes (Dyer, 1983) and personality traits,
• a rela:zed planning mechani6m,
• a per~onality model which provides the goals and con­ trol goal· which guide relaxed planning,
• a dynamic epi6odic memory of experiences used by re­ laxed planning, and
• domain knowledge of interpersonal relations and com­ mon everyday occurrences.

Control goals are postulated to help direct daydream­ ing. Four such goals are common in the transcripts we have examined:

Rationalization, activated by negative emotions resulting from a recalled failure, involves generating reasons for why an out­ come is satisfactory to the daydreamer in order to reduce the negative emotions and maintain self esteem.

Revenge, activated by emotions of anger resulting from a re­ called failure caused by another, involves imagining retalia­ tions against that person in order to generate positive emo­ tions.

Failure/Succeu Reverul, activated upon recall of a goal failure or success, involves altering reality by imagining scenarios in which failures were prevented or in which successes failed to come about in order to learn future plan­ ning strategies.

Preparation, activated upon recall of an event planned for the future, involves generating hypothetical future scenarios in order to learn planning strategies and/or specific actions to be used in possible future situations.

Once a control goal is activated, the relaxed planner selects and executes a plan to achieve that control goal. Plans for achieving control goals may often be captured by Plot Un­ its (Lehnert, 1982). One plan for rationalization is expressed by the MIXED BLESSING Plot Unit, which represents the abstract situation of a goal success leading to the failure of another goal. This plan is employed in DA YDREAM3 in order to achieve the rationalization control goal activated by the emotion of disappointment resulting from being rejected by the actress. Here the MIXED BLESSING Plot Unit is realized as a fantasy in which dating the actress leads to eventual failure of the relationship and the daydreamer's career. Another plan for rationalization is captured by the SUCCESS BORN OF ADVERSITY Plot Unit, which represents the abstract situation of a goal failure leading inadvertently to the success of another goal. Here we could imagine a daydream in which, as a result or being turned down, the daydreamer ima­ gines drowning his sorrows at a bar and by chance meeting a more beautiful actress.

Realization of plans for achieving control goals ex­ pressed as Plot Units may be simplified through reminding of appropriate episodes indexed by those Plot Units. For exam­ ple, in DA YDREAM4 a revenge control goal is first activated by anger toward the actress. A plan for revenge is captured by the RETALIATION Plot Unit, which represents the abstract situation of a person A achieving a positive outcome by caus­ ing a negative outcome for B, where B had originally caused a negative outcome for A. Next, the RETALIATION Plot Unit along with other indices such as "rejection" might retrieve a vicarious episode such as the following:

Sarah applied to Yale for graduate school but was rejected so ·he went somewhere else. Years later she won the Nobel Prbe. Then when Yale tried to recruit her as a professor, ·he ·pltefully turned them down.

This episode may now be adapted through analogy to the si­ tuation of DA YDREAM4 in order to achieve the active re­ venge control goal.

VD. CURRENT STATUS AND FUTURE WORK

A prototype of the DAYDREAMER program has been constructed using GATE {Mueller & Zernik, 1984), an in­ tegrated set of graphical artificial intelligence development tools for the T language {Rees, Adams, & Meehan, 1984 ), a di­ alect of Scheme running on Apollo Domain workstations. DAYDREAMER currently 1) participates in NUART­ EPISODE by receiving and performing actions in response to input phrases such as 'You accidentally meet your favorite ac­ tress at the Nuart Theater' and 'She turns you down', 2) indexes the episode into memory under the DENIED RE­ QUEST Plot Unit, 3) generates versions of DA YDREAM1, DA YDREAM2, DA YDREAM3, and DA YDREAM4 in simplified English, indexing the daydreams into memory via Plot Units and emotions, and indexing into memory the planning stra­ tegy formed during DA YDREAM2, i.e., not to forget to ask for someone's telephone number, and 4) participates in another episode in which it demonstrates that it has learned the above planning strategy. The computational theory of daydreaming which we are developing and the DAYDREAMER program are described in greater detail in {Mueller & Dyer, 1985).

Future plans call for the extension of DAYDREAMER to enable it to 1) participate in a larger variety of input ep­ isodes, 2) generate more daydreams in response to each ep­ isode, 3) incorporate many daydreams, planning strategies, and future plans into memory, and 4) demonstrate the use of recalled daydreams, planning: strategies, and future plans in future internal (daydreaming) and external behavior. We in­ tend for DAYDREAMER to be able to daydream continuous­ ly, stopping only to receive new experiences.

VTII. CONCLUSIONS

The functions of daydreaming and a c.omputational theory of daydreaming are being implemented and tested in the DAYDREAMER computer program. We have argued that, far from being a useless epiphenomenon, daydreaming serves an important cognitive role in supporting creativity, fu­ ture planning and rehearsal , learning from failures and successes, and emotion modification and motivation . Truly in­ telligent computers should not be idle when left unemployed by their users, but daydreaming like ourselves.