ditors
| tware Studies,
Expressive Processing
Digital Fictions, Computer Games,
and Software Studies
Noah Wardrip-Fruin
‘The MIT Press
Cambridge, Massachusetts
London, Englandts have bow
vackgrounds, sich
Chapter 5
The Tale-Spin Effect
The “Metanovel”
In fall 1974, James Mechan was a graduate student at Yale
University. He had an idea in mind for his dissertation
topic, but didn’t know how to pursue it, The topic had
been suggested to him by Alan Petlis—one of the most
famous figures in US. computer science, who had been
named chair of Yale’s department a few years before—
on the first day they met. But Perlis didn't know how to
move forward with the idea, either. In the preface to his
dissertation, Mechan describes the idea this way:
A metanovel is a computer progeam that tells
stories that only a computer could tell, stories of
such complexity of detail that oaly a computer
could handle, stories with more flexbilty—even
reversibilty—of events and characters than a
human could manage. A metanovel time-shating
system tells a story to many people at once, 20
‘wo of whom read the same thing, because they
have cach expressed different interests in the
events and characters they want to heat about, and
because they may each desire a different style of
storytelling, And yet, among all these readers, these
is bar one story—the Metanovel itself—and each
reader is only following those threads of the story
that interest hie. (1976, i)
Mechan’s dissertation dida’t set out to create a metanovel
but rather 1o make progress toward the possibility. As he
notes, “If I've been successful, we't a little closer than
we were” In the decades since, his work has become
a landmatk—the first project cited in nearly every
discussion of story generation,of Neekan's cigs,
ter aspects of ich
focus onthe mismatch
between te bear
of Kens system
and tries the
behavior of eran
storytellers,
Computer systems had certainly produced stories
as output before. The most famous of these was the
“qutomatic novel writer” produced by Sheldon Klein and
his collaborators (1971) at the University of Wisconsin at
Madison, But the ficld has generally followed Mechan’s
titique of Klein's system: it was made up of explicit
chunks of action, with a path through these chunks
selected randomly? So, for example, Klein's system
includes a “rule for people arriving at George's living
00m” (27), Mechan’s system, on the other hand, contains
general rules for character movement (and reasons for
movement) that operate no matter which spaces are
available—living rooms, bedrooms, caves, or meadows.
More broadly, this points to the two main types of
systems for creating variable fictions. Klein's system,
like quest flags and Choose Your Oam Adventure books, is
composed of explicit chunks of story content. If one of
these is changed or removed, not much else changes. At
most, if some story chunks are only accessible by passing
through others, when a chunk is removed other chunks
may become impossible to reach, Mechan’s system, on.
the other hand, models story as a relatively fine-grained
set of processes and data that are used to generate story
events. In the case of TaleSpin, this is accomplished
by creating a simulated world, processes for behavior
in and of the world, and characters and objects that
populate the world. As a result, changing one aspect of
the simulation (such as 2 rule for character movement
of the existence of a particular object) can lead to wildly
different TaleSpin fictions. This kind of flexible model
has become « defining characteristic for the pursuit of
story generation.
16
Basis for tt
‘The breakthr
forward towa
scruffy actif
and Robert A
fall 1974, and
processing s¢
weekly battle ¢
victims” such,
Many of
field of com;
previous work
For example, i
language to at
and dictionary
instead propo
of semantic n
expressions.
Information Pra
serve as an dn
from any one
by translating
representation
an internal me
AL projects, |
and inference
by Schank an
Shin followed
dependency ¢
But under
than simply 1
context. “MasThe Tale-Spin Effect
duced stories Basis for the Model
| these was the 2 ‘The breakthrough that allowed Mechan to see a path
\don Klein and forward toward the metanovel was exposure to the
£ Wisconsin at scruffy artificial intelligence (AT) ideas of Roger Schank
vwved Mechan's j and Robert Abelson. Schank had just arrived at Yale in
| up of explicit = fall 1974, and Meehan enrolled in his natural language
| these chunks processing seminar—which Meehan described as “a
Klein's. system ‘weekly battc of ideas” featuring favorite “guest speakec/
Seorge’s living victims” such 2s Abelson (1976, i).
| hand, contains _ Many of Schank’s theories were developed in the
| id reasons for a field of computational linguistics, where he argued that
[ch spaces are a previous work was fundamentally flawed in its approach.
| or mendows 3 For example, in the area of computer translation from one
| main types of language to another, projects based on syntactic parsing
dlein's system, ‘and dictionary-style substination had largely failed. Schank
| ture books, is > instead proposed a language-independent representation
| teat I one of = of semantic meaning, which he called concpfual dopendensy
| se changes. At a expressions. As Schank (19752) desesibes themin Concyptual
| le by passing ; Information Procesing, such semantic representations could
| other chunks a serve as an interlingua, so that the translation of a statement
ats system, on from any one language to another could be accomplished
ly fine-grained by translating that statement to and from the semantic
‘generate story : representation. Such a representation could also serve as
| accomplished = an internal meaning sepresentation—or data format—for
for behavior - AT projects, helping in processes such as paraphrasing
4 objects that : and inference making, as demonstrated in systems built
one aspect of ‘ by Schank and his students (see sidebae: Yale AI), Talk
ter movernent : Spin followed this lead, operating in terms of conceptual
lead to wildly dependency expressions.
exible model 4 But understanding the meaning of sentences, rather
he pursuit of than simply their structure, requires understanding their
context. “Mary hit John” probably means different things,Expressive Processing
Yale Al
James Mechan’s Tafe-Spin isthe only project | discuss in detail that was developed
at Yale University's Artificial Intelligence (A) laboratory. However, the Yale Al Project
‘and the worldview of Roger Schank, its charismatic leader, were central to the
evelopment ofthe strand of work that connects 2 number of my central examples.
‘This Is especialy true of Tale-Spin, Michael Lebowitz’s Universe, Scott Turner's
‘Minstrel, and the ideology mode! for Michael Mateas et a's Terminal Time,
In 1987 Stephen Slade published an article looking back on a dozen years of
projects following Schani’s arrival at Yalo, In reviewing the largest body of work,
that described as “cognitive modeling," Slade notes how it largely continued in the
mold set by the Margie system, developed by Schank and his students during his
time at Stanford University (before Yale). Specifically, Slade notes Margie's:
Task Orientation, An Al program should address a specific, real-world task.
‘The rogram should model something that a person actually does, rather than
an artificial abstraction of inteligent behavior. Margie's tasks included reading,
paraphrase, and translation.
Psychological Process Model. The Margie program was @ cognitive simustion
Not onty did it ry to perform tasks that people perform, but i tried to simulete
‘the manner in which the human mind works. By comparison, a computer chass
program which exhaustively searches ahead several moves may be abe to play a
‘ine game of chess, bu itis unrealistic to consider such a program a model of the
‘way in which a person plays chess.
Canonical Representation of Knowledge. The heart of the Margie ystom was the
‘conceptual dependency knowledge representation system. CD provided a means
(of representing actions and states in a canonical, language independent fashion,
‘A concept represented using the dazen CD primitives might be expressed in any
‘number of ways in any numberof languages... (68)
All three of these Schank-driven priorities are clearly evident in Tale-Spin, as this
‘chapter's discussion makes clear. They also deeply inform the backaround of the
other Yale-related projects | consider, as well as influencing and reflecting the
broader Al research of the era, especially that which could be termed “scruffy.” With
these goals in mind, the motivations for cartain systern-design choices should be |
‘easier to understand as | discuss them in the pages ahead,
118
if Jobn is M:
at Mary’s bla
of projects
together the
stories with «
same task.
that would bi
primary goal
the causal 1
step was at
for certain ro
or catching «
succeeded in
accidents) by
sentences of
an internal sc
But Tale
scripts are “s
great story 1
for theorizing
elements of bs
around Abels
central Al ct
place just as
results looke
“The Tale
status of Sch
behavior In
were data. T
patterns 10 «
stories beingdeveloped
AI Project
tral to the
examples,
ft Tumer's
8
years of
¥¥ of work
ued in the
during his
's:
| ¢ task
cr than
cading,
lation
imulate
rohess
> play a
ofthe
as the
‘mean
ashion
sin any
tn, 28 this
und of the
ecting the
vy.” with
should be
The Tale-Spin Effect
if John is Mary’s sparring partner or if John is sitting
at Mary's blackjack table. This fact motivated a series
of projects directed by Schank that attempted to bring
together theorizing about how humans understand
stories with the building of computer systems for the
same task. The systems were viewed as experiments
that would help refine and validate the theories, and a
primary goal was developing theories for understanding
Ne fest
the causal relationships between clements
step was a theory of sripts—that human knowledge
for certain routine situations (eg, going to a restaurant
of catching a bus) exists as stereotyped sequences of
common events. Computer systems were built that
succeeded in interpreting simple stories (eg, of auto
accidents)
sentences of the stories and corresponding clements of
yy determining relationships between the
aa internal seript,
‘But Tab-Spin lagely avoids septs. As Meehan puts it,
sctipts are “so developed that they're uninteresting: not
great story material” (1976, 213), Sctipts only allowed
for theorizing and system building about the least novel
clements of hummaa life. The next step was to build systems
around Abelson’s and Schank’s developing versions of the
central AI concepts of plans and goal ‘This was taking
place just as Schank artived at Yale, and to Meehan the
sesults looked like mach better story material?
‘The TaleSpin system produces stories by changing the
status of Schank’s and Abelson’s theories about planning
Dehavior. In story-understanding systems, these theories,
were data, They were used by the system processes as
patterns to compare against the behavior described in
stories being interpreted. But in TaleSpin they become
[ 2. stam aces
Neha’ pinion of
‘he elatie interest
va of tres
about cris ves
those abut pars
inside Computer|
esta
Mery sores
(parity te more
intresting ona) rt
oat ene ew
Notes continued at end
oft capi.
|Expressive Processing
process, used to create a world in which characters behave E permanently
according to the theories. As Meehan puts i: copy of i
TaleSpin includes a sisoulator of the real world ; Ongoing,
Turn it on and watch all the people. The purpose : -was made po
| ofthe simulator is 10 model rational behavior; 4 fictions prod
the people are supposed to act like real people. :
| (1976, 107) foem the ba
| 4 Spin, wich
| Most TaleSpin stories don’t featare “people” thet are 3 dissertation 5
| human characters, however. Instead, the people are TalesSpin's pe
largely bitds, bears, bees, foxes, and so on. This was about its dat
| Schank’s suggestioa, inspired by, as he puts it, “the fact q : science disev
tha [had lie is a he sim eo Tas taking up sytem 2
{stories} and trying to see how fl] did it?” Iealso served to For man
compensate, Mechan writes, for the fact that “we weren't a » TalesSpin fice
‘going to get very sophisticated or elaborate outpat fom : never learne
this program’ (2006). In other words, the decision to tell 3 q tolearn abou
‘animal stories is an example of the common AT approach . both with d
‘of defining a mrioword in which the program will operate. : berween tho:
Jast as some Al researchers demonstrated progeams that 3 only savy its
could understand the physics of a simplified “blocks a After ea
wold” (allowing them to ignore issues such as the relative : even more d
strength of material), TaleSpin operated in a simplified : foracopy of
animal world resembling the settings for Aesop's fables Tale Spin (M
(shaving complexity both from the possibilies of che But this wa
‘world and the potential motivations of the charseters). i interested m
| : formersoade
| Tale-Spin Today of Californi
Today Tak-Spinis one of the most widely discussed digital a pleasant cc
fictions ever produced. It is not only a couchstone for and he looks
computer science accounts of story generation but also 3 traces of Tal
broadly cited in writing about digital literature and the s “After sor
|
future of fiction, Atthe same time, TalsSpinitself seemed : locating a e¢
120acters behave
al world:
+ purpose
behavior;
we weren't
output from
scision to tll
Al approach
sill operate.
rogyams that
| fed “blocks
| as the relative
1a simplified
cesop’s fables
ites of the
‘haracters)
cussed digital
uchstone for
tion but also
ture and the
itself seemed
The Tale-Spin Effect
permanently lost during most of this conversation —unti
a copy of its code was unearthed in 2008.
Ongoing discussion of this seemingly lost software
‘was made possible by two facts. First, many examples of
fictions produced by TalsSpin ase in circulation. These
form the basis of most humanistic discussions of Tale-
Spin, which vend to be dismissive. Second, Meehan’s
dissertation gives detailed accounts of the operations of
TaleSpi’s processes, along, with significant information
abouts data. This serves as the basis for most computer
science discussions of TaleSpin, which tend to treat the
system as worthy of serious engagement.
For many years I was only exposed to examples of
TaleSpin fictions, which I found so uninteresting that I
never learned more about the system, But once I began
to learn about TaleSpin’s processes I became fascinated—
both with the system itself, and the differing attitudes
between those who understood the system and those who
only saw its output,
‘After reading Meehan’ dissertation, began to look for
even more detailed sources of information and especially
foracopy of the software itself Thad previously seen Miro
‘TaleSpin (Meehan 1981), a smaller, pedagogical version,
But this was so simplified that it lacked most of what
interested me about TaleSpin. Through Walt Scacchi, 2
versity
formerstudent of Mechan's feombis time atthe Us
of California at Irvine, I got an email introduction. Thad
a pleasant correspondence with Mechan, now at Google,
‘and he looked through his garage on my behalf—secking.
traces of TaleSpin—but his intial search came up empty,
‘After some further searching, I gave up hope of ever
locating a copy of Tal-Sin? But thea, in Match 2008, I
121
[ 5. mene stage
‘ny search bepan
whan seed fom
Meahan abot snot
ieraton ofthe sytem,
ore crix than
the "mero" version,
rot fo tho 1987
Sacto |
cab at he Boston
‘Computer Museu
Unfortunately.
Notes contin a end
of tis chapter.Expressive Processing
4 Though tobe
sorpubesly acct,
‘o my tnonledge
| Weintauen only
} csriutes te
verati coe or
the Doct sri Hs
Dubro on re
tea syst, hanes,
wor opi enough
to make independent
reimplenenition
relatively easy.
received an email from Meehan that began:
Never say never. I'm in the provess of moving to
a smaller place (in San Francisco), which means
going through every box and deciding what to
Keep and what not to keep. Lo and behold, I found
2 ptint-out of the complete source of TakSpin
It’s about 4K lines of MLISP, printed on fan-fold
paper. Theres even a cross-reference table at the
end. (2008)
Hesent me the listing, which gives acomplete specification
of Tale Spin’s processes (Mechan 1980). Unfortunately, he
did not turn up a listing for any of the accompanying data—
files, However, this still opens the possibility for us to learn
again from TakSpin, as we do today from Eliga/ Doctor.
Jn particular, it gives us the opportunity for knowledge
through personal interaction, programmatic substitution
and deformation, and examination of the code, rather
than simply from reports of its processes ‘The best first
step is probably to follow Warren Sack's lead and translate
the code from MLISP into Common LISP, as he did with
‘Micro TaleSpin (Meehan and Sack 1992)
‘My discussion hete emerges from my study of
Mechan's 1976 dissertation and 1980 source code. As it
happens, some of the most interesting lessons I perceive
atise from a kind of inversion of the Hlize effect.
Rather than a surface illusion of process complexity
and intelligence, TaleSpin creates a surface illusion of
process simplicity and arbitrary action. Te is far from
alone in this, As a result, I believe understanding the
more general Tale.Spin effect can provide insights for
both authors and intespreters of digital fictions, as
well as those who seek to understand computational
systems broadly.
A Tale-s
TalesSpin, as
storytelling
audience to
would, ‘hil
production
dissertation
story, about
to illusteate
Creating
Tale Spin 2
and the re
results in 2
of facts fo
the
story told ¢
imulat
for the bes
gives the b
of story ch
bear, create
exists, mar
and lets be
Dear is in 1
Acchur, an
created (al
a meadow
other's list
other exist
location,
This kei
conceptualThe Tale-Spin Effect
‘ A Tale-Spin Story Fee he tie mode,
| noving to 4 TedeSpin as described in Mecha disertation, ha three | Mevanwwieé
[ame storyteling modes, Two modes areinteractive, asking the | Ss is sh
la Yioond 1 audience to make decisions about features of the story | as'TheFox ad te
| Tab Spin ‘world, while one mode “fixes” the world to assure the | Hw ard “The Ant
|» fan-fold " production of particular stories’ Chapter 11 of Mechints MMe”
bana S dissertation gives a detailed account of an interactive
| . story, about a hungey bear named Arthur, that I will use
| specification S to illustrate Tale Spin’s operations and their backgrounds,
ortanately, he :
apanying data : Creating a World
| rustolearn ‘TaleSpin asks what characters to use for the story,
| Bliza/ Doctor. a and the reader chooses a bear and bird. Each choice
|r knowledge e results in a “picture procedure,” creating the basic set
|: substitution : of facts for that character that will be used to make up
| code, sather the simulation but won't necessarily appear in the final
| rhe best first 4 story told to the audience. For example, the procedure
and translate for the bear creates a bear, chooses the bear's gender,
wshe did wich 4 gives the beat a name, adds the bear's same to the list
i e Of story characters, chooses a height and weight for the
jay study of : bear, cteates a cave, creates a mountain in which the cave
vecode. Asit ; exists, marks the cave as being the home of, the bear,
ans [perceive s and lets both the bear and the system know that the
Elia effect. ; beat is in the cave. In Meehan’s story the bear is named
5 complexity Asthus, and by similar means a bird named George is
¢ illusion of | created (along with a nest, a maple tree for the nest, and
is far from : a meadow for the tree). Each character is added to the
standing the other's lit of acquaintances, so they know that each
insights for : other exist, and they each have an idea of the other's
fictions, as - location.
smputational . ‘This kind of information is represented in Schank’s
conceptual dependency expressions. While Schank’sExpressive Processing
6. Tis example and
the ane above are
adapt rom Meshar's
dissertation (1976,
195-196)—substiting
oer charac wit
those inthis soy
published accounts at the time represented conceptual
dependency expressions using complex diagrams with
multiple forms of connectors, software such as Tale Spin
used more machine-readable formats. In Tale-Spin, the
conceptual dependency expression for “Arthur is in his
Cave” could be represented as:
((ACTOR *ARTHURBEAR® IS (*LOG*
VAL (‘PROX* PART *CAVE*O))))
Bat much of TalsSpin operates on moze complex
conceptual dependency expressions, as the coming pages
will reveal. Meehan writes his own English-language
representations of these, and I will follow his lead here.
Otherwise, even relatively simple ideas such as “Arthur
knows that George thinks he’s an idiot” would sequire
parsing conceptual dependency expressions that look
like this: €
((CON ((CON (ACTOR *ARTHURBEAR®
IS (*SMART* VAL (-9)))
1S (MLOC* VAL (*CP* PART
*GEORGEBIRD")))
IS (MLOC* VAL (*CP* PART *ARTHURBEAR*))))
Though from here my discussion of ‘TaleSpin will
take place using English ranslations of concepmual
dependency expressions, looking a bit more at the above
examples will help to clatify the structure of these
expressions, If one looks at the sepresentation of the
idea that Arthur isin his cave, the structure is similar to
that for the idea that Arthur is an idiot (his Smart has the
value of “9” on a scale that runs from 10 to -10). MLoe
is a Mental Location, and in the two cases above itis the
124
CB or Cons
wwe might 1
from the ow
sentence lik
that isin
Bear's intel
‘The wo
and inferen
involving sc
dependency
The MLoe|
events —the
a persons 2
play import
ate most ca
relatively ec
a body par
object), Fg
Grasp (phy
sound), and
on somethi
abstract:
Alirans (Ch
t0 an object
Motivatir
Remrning t
any miscel
“flower,” “
worm, and
is currently
about the wed conceptual
diagrams with
ch as TaleSpin
TalesSpin, the
cethur is in his
nore complex
coming pages
\plish-language
> his lead here
ich as “Arthur
would require
ons that look
RBEAR*))))
TaleSpin will
3f conceptual
re at the above
ture of these
tation of the
re is similar to
5 Smart has the
Ito 10). MLoe
above it is the
The Tale-Spin Effect
CP, or Conscious Processor, of one of the characters. So
‘we might more literally translate the expression above,
from the outermost parentheses to the innermost, with a
sentence like, “I isin the consciousness of Arthur Bear
tacit isin the consciousness of Geonge Bird that Arthur
Bear’ intelligence is at a very low level?”
‘The world of TalesSpin changes based on events
and inferences from these events. All events are acts,
involving some actor taking some action, In conceptual
dependency theory, there are only eleven primitive acts,
‘The MLse facts above would likely grow from MTrans
events—those that transfer information, either within
a person’s memory or between people. Primitive events
play important roles in TaleSpin, and the remaining ten
axe most casily described in two groups. Seven acts ase
rclatively concrete: Prope! (apply force to}, Mase (move
a body part), Ingat (take something inside an animate
object), Expl (force something out of an animate objec),
Grasp (physically grasp an object), Speak (produce a
sound), and Attend (direct sense organ toward of focus it
on something), The last three acts, like MTTrans are more
abstract: PTrans (change the location of something),
ATinans (change some abstract relationship with respect
to an object), and MBuild (create or combine thoughts)”
Motivating Action
Returning to Acthus’s story, the audience is asked whether
aay miscellaneous items should be" created: “berries.”
“lower;” “river?” and “worrn.” The audience chooses
worm, and it is created along with the ground where it
is currently located. The audience is asked who knows
about the worm, and chooses Arthur Beat. TalesSpin now
7.This summary
‘rams Shank’
Concept internation
Processing (1788,
41-4|
|
|
Expressive Processing
hhas enough information to ask who the main character of
the story will be:
THIS IS A STORY ABOUT
1: GEORGE BIRD 2 ARTHUR BEAR
‘The audieace chooses Arthur, and then is asked about
his problem:
HIS PROBLEM IS THAT HE (S
J;HUNGRY 2: TIRED 3: THIRSTY 4 HORNY.
‘The world of Tak-Spin revolves around solving problems,
pasticulatly those represented by four “sigma-states”
Signa Hunger, SigmaThirt, Signa-Rest, and Signa-Sex.
‘These, as one might imagine, represent being hungry,
thirsty, tired, and—in the system's terminology—horny,
respectively. ‘These problems are solved using Schanke’s
and Abelson’s approach to planning.
Schank characterizes planning this way in his paper
“Using Knowledge to Understand”:
A Plan is the name of a desired action
whose realization may be 2 simple action (f]
conceptualization involving a primitive ACD).
However, if itis realized that some state blocks the
doing of that action, the plan may be translated
into @ deltact to change the state that impedes
the desired ACT. Thus, a Plan has attached to it a
‘gfoup of deltacts with tests for selecting between
them. (19758, 119)
So, for example, if—-by unlikely circumstance—Arthur
forms a plan to eat when he already has honey in front
of his mouth, all he has to do is the primitive act of
Ingesting the honey. But if Arthur does not already have
honey in front of his mouth, the plan will be translated
126
into a deliact w
some other foo
one that invol
hand. If he do
some food the
idea of deltact
comes from A
as an attempt t
building blocks
‘causing a state
some later acti
steps from the
1975, 5). Sch
‘what he calls
to achieving th
For TaleSp
that correspon
acts’ PTrans,
is used when
something to
‘changing one’
Neg Proxis use
Delta Know is
state of some
procedure fo:
the procedure
Dao Final
to acquite sor
Ip TaksSp
acts depend
to do someth
together intoThe Tale-Spin Effect
character of into a deltact with the goal of getting honey (or perhaps (6, Yestan cals ar
a some other food) in front of his mouth. He may choose | detracts in his
a a ae Sa | ei, te
. cone that involves moving some food there with his | Sysw ani sem
hand, If he does not have any food in his hand, getting | cat tum abc in
‘ne 1978 puns
some food there may become a state-change goal, This
| food there may be state-change goal. This | Wa Were
asked about idea of deltacts (or as Meehan calls chem, “delsa-acts”) | ageon 1875). wil
i comes from Abelson." He describes his work on them | alms, depending
: as.an stempt “develop a set of intention primitives ag | ™¥ns wat
4: HORNY building blocks for plans. Each primitive isan act package \————~
| : causing a state change. Each state change helps enable
| ag problems, some later action, such that there is a chain or lattice of
| igema-states” a steps from the initial states to the goal state” (Abelson
| Signa Sexe a 1975, 5). Schank embellished the idea of delacts with
j cing hungry, ; what he calls “planboxes.” These are specific approaches
‘ogy-—horny, to achieving the state change.
jing Schank’s a For TalsSpin, Mechan implemented three delta-acts
| E that cortespond to the primitive conceptual dependency
in his paper | acts PThans, MTrans, and ATrans ‘The first, Deite-Pros
is used when someone wants to change the state of
something to make it neat something else, which includes
|
action
ction (la 2 changing one’s own state to be near something, (Dello
Tas a NegProcis used to get something away from somewhere.)
translated : Dae Know is used when someone wants to change the
impedes x state of someone to know something. Meehan calls his
foe procedure for the communicative version of this ‘Telland
: the procedure for wanting to find something out oneself
vox Finally, Delta-Contralis used when someone wants
nce—Arthur, 2 to acquire something,
>ney in front : In TaleSpin, many of the planboxes of these delta-
nitive act of a acts depend on somehow persuading another character
already have to do something, These common planboxes are gathered
bbe translated together into a Persuade package, which includes: simply
127Expressive Processing
askings linking the action to a goal thatthe other character
is presumed to have, and enabling the action; bargaining;
and threatening, Like all planboxes, each of these has
preconditions—a character does not ask another to
simply tell her something if she believes that the other
intensely dislikes her. Similarly, 2 character does not
threaten a character she likes 2 lot.
Making Plans
When the audience chooses hunger as Arthur's
problem, Tale-Spin generates the conceptwal dependency
expression “Arthur knows that he s hungry.” Thisis then
asserted. To understand what this means, it is heipfal to
look at figure 5.1, which illustrates the conteol structuse
of TaleSpin's simulation
‘One thing that assestion does is add information to
memory. TaleSpin’s memory contains information about
the structure of the physical world, the structure of
sclationships between characters, facts about particular
characters, and so on. Every fact in memory is indexed
based on who believes it, with true things indexed to the
Figure 5:1. Goal stutireof Tabs sus ater eshans Tae 1
128
system, Ev
(erue at son
hhangey” is:
believed by
When ¢
the fact bei
negative sig
conceptual
knows tha
expression
to the asse
problem so
The nar
hunger is 5
eat honey «
if Acthur k
or knows
80 S-Fhuage
honey—an
Dons fi
starts check
it checks ta
checks to s
structure. I
Why do
already par
of reasons,
is alzeady
then it mak
cheek for d
filed goals
a characterThe Tale-Spin Effect
other character system, Every fact is also either currently true or “old”
| on; bargaining : (erwe nt some time in the past) In this case, “Archus is
1 of these has Iangey” is addled to memory and indexed as something.
sk another to : believed by Arthur,
that the other When this fact is added to memory, it results in
eter docs not the fact being passed to the inference mechanism. For
negative sigma-states, the inference mechanism creates
‘conceptual dependency expressions of the form “Arthur
knows that Arthur intends not to be hungry” This
as Arthu’s = expression represents, of course, a gach When itis passed
| al dependency to the assertion mechanism, the result is to invoke the
| ye" Thisis then, problem solver, which begins the process of planning,
itis helpful to ; ‘The name of the TaleSpin proceduse for satisfying
trol structure unger is S-Hanger (For Siga- Hunger). In TalesSpin, bears
cat honey and berries. J-Hurger checks memory to see
formation to if Arthur knows that he owns some honey ot berries,
rmation about ‘or knows where some honey or berries are. He doesa't,
steucture of $0 S.Hunge selects a bear food at random—in this case
rout patticular 4 honey—and invokes DGont (Delte-Contro) to get some.
ory is indexed DCont forms the goal “Arthur has some honey” and
indexed to the starts checking the preconditions for its planboxes, First
it checks to see if it is already true (it is not). Then it
checks to see if this goal is already part of Arthur's goal
structure, Its not, 0 itis added
Why does DCont check to see if getting honey is
already part of Arthue’s goal structure? For a number
of reasons, the most obvious of which is because iF
a its alzeady his goal (or a subgoal toward a higher goal),
then it makes little sense to add it. But another reason to
check for the goal’s existence is that TaleSpin also keeps
filed goals, and the reasons for their failure, as part of
a character's goal structure. Before this was added, TaleExpressive Processing
Spin created stories with a certain surreal quality. Here's
an example:
Joe Bear was hungry. He asked Irving Bird where
some honey was. Irving refused to tell him, so Joe
offered to bring him a worm if he'd tell him where
some honey was. Irving agreed. But Joe dide’t
know where any worms were, so he asked Irving,
who refused t0 say. So Joe offered to bring him a
worm if he'd tell him where a worm was, Irving
agreed. But Joc didn’t know where any worms
‘were, so he asked Irving, who refused to say. So Joe
offered to bring him a worm if he'd tell him where
a worm was. (Meehan 1976, 129-130)
In fact, these errorproduced stories—which Mechan
called “misspun tales”—are TaleSpi’s most famous
ourpurs, much more widely reprinted than anything
produced by the completed system. There are a number
of seasons for this. Fist is the amusement value. Another
is that these stories, hand translated from conceptual
dependency expressions by Mechan, have much more
fluid prose than stories produced by TalesSpin's companion
saturallanguage genezation system, Mumble, Third, these
stofies also give 2 hint, simply though their structure,
of some of the mechanics of the underlying TalSpin
processes. Unfortunately, as I will discuss later, those who
attempted to interpret these outputs without any deeper
understanding of the Tab-Spin system often reached
tather misleading conclusions,
Developing Relationships
Arthur is trying to get some honey, which is a Dalte
Control problem. In TalSpin, the first precondition for
DGont’ planboxes is to know the location of the thing
to be controlled, Since Arthur does not know where
130
any honey is,
“Arthur know
trae, and its
is added. The
is a standard
checking a w:
one of these
of DKwow is
fon these mai
bur Archu’s
moves to the
of the Persua
A ftiend is so
positive value
Bird's opinio:
that Arthur
have only on
up on the gi
agent to find
Relations
to Schank’s
his work oa
relationship
state to men
mode of ask
DOES ARI
GEORGE €
+ ALOT
TalesSpin ca
states that 4
one, is: DosThe Tale-Spin Effect
i
be
waligy Heres , any honey is, DKnow’ i called. DXnow forms the goal [9 tearing
“‘Asthur knows where some honey is.” Its not already) Mestar's dissertation
| " : (197,50), The
ied where j true, and its not yet part of the goal structure, soit | fvnoraf his
m, oie ‘ is added. The first planbox of DKnow is used if there | porion of chapter
oe dite is a standard method for finding the information: like | 185 (bid, 170)
eles ts stp,
2d Icving, : checking a watch of clock to learn the time. ‘There isate
ee ‘one of these for honey in Tale-Spin, so the next planbox
Pwo of DKwow is to see if there’s someone who is an expert
ay. So Joe a. fon these matters to ask? Bears are experts at honey,
Ea iieeed bbue Arthur's the only bear he knows about, so TaleSpin
| moves to the next planbox, which is to use the methods
lich Meehan of the Perstade package to convince a friend to tell you,
xost famous A friend is someone that you think relates to you with a
an anything: positive value for affection. Arthur's opinion of George
we a number Bird’s opinion of him is unknown. If the audience says
due. Another that Arthur doesn’ think they'te friends, Arthur will
» conceptual . have only one DKnow planbox left before he has to give
much more up on the goal of having some honey: Pasadeing an
Scompanion agent to find out the answer and tell him.
| ‘Third, these Relationship and personality ‘states ate extensions
jie structure, a to Schank’s and Abelsoa’s work created by Meehan for
ing, TalbSpin a his work on Tale Spin (1976, 40): Testing memory For a
2 those who: ; relationship state is done by a procedure that adds the
tany deeper 3 state to memory if it isn’t there, Since TalesSpin is in the
en reached : mode of asking the audience it types
DOES ARTHUR BEAR THINK THAT
GEORGE BIRD LIKES HIM?
his a Det 1: ALOT 2: ALITTLE .3:NOT MUCH 4:NOT AT ALL
condition for = ‘Tale-Spin can simultaneously maintain four different
of the thing states that are similar to this in memory. The first, this
know where S cong, is: Does Actiur think that George tikes him? Thei
Expressive Processing
others are: Does Arthur think that he likes George
(ies does Arthur like George)? Does George think
that Arthur likes him? Does George think that he likes
Acthur (Je., does George like Arthur)? All are used i
making different types of plans.
‘The audience says that Arthur thinks George likes
him 2 lor, 80 Persnade starts working through its planboxes
with the goal of Arthur getting George to tell him where
some honey is, in pursuit of the higher goal of getting
some honey. The frst planbox is to simply ask, but it has
further preconditions. So TolsSpin asks if Arthur feels
deceptive toward George (the audience answers: aot at
all, if Axthur feels competitive toward George (not atl),
if Arthur likes George (@ lod), and if Aetiar thinks that
George is trying to deceive him (not 2t al).
Finally, Arthur Bear has a plan to know where some
honey is, so he can contro it, so he can eat it, so he can be
less hungry: he'll ask George Bird.
Speculations and Lies
From here things begin to go poorly for Arthus, It turns
out that George is a deceptive bird and will deliberately
tell Arthur about an alternative possible world that isn’t
the one of this TalesSpin story. That is, he’s going to lie,
and Tale-Spin is going to create a parallel world structure
to support this lie, representing the fact that Arthur
believes it
First, Arthur wants to Te George his sequest. Tell
has two planboxes: do it yourself, or get an agent to
do it. Arthur starts with the first one, which requires
that he be near George. This is a Delta-Prox problem,
resulting in the formation of the goal “Archur is neat
132
George,” and
going, Since
thinks he kn¢
need to start
more of the
Acthue’s mot
P Trans moves
to near Geor
include Artht
has arrived
there, Before
Speak to tet
of characters
TaleSpin's fat
Heney An
bank whe
Henry sli
tall for
Mechan adde
location, caus
not so lucky
Henry A\
iver ban
sitting, He
drowned.
In this versic
performed a
them both in
Bill, butgeav
how gravity
dependency
ReturningThe Tale-Spin Effect
ikes George George,” and the appropriate planboxes of DPrax start
Jeorge think : going, Since Arthur knows where he is (the cave) and
| that he likes 3 thinks he knows where George is (his nest), there’s no
1 1 are used in a need to start up DKnow planboxes. DLink then creates
| 2 more of the wold, so that there’s a connection between,
| George likes 2 ‘Arthur's mountain and George’s meadow. Then Do-
| its planboxes i Trans snoves Arthur along the ground (bears cannot fly)
him where 4 to neat George's maple tree. The inferences from this
| al of getting a include Arthur and George both knowing that Arthae
| isk, but it has has arrived there, and Arthur knowing that George is
| Arthur feels there. Before these sorts of inferences, characters had to
| swers: not at 2 Speak to let others know of their presence. This failure
| ge (aot at all), a of characters to notice each other created another of
ur thinks that : Tale-Spin’s famous misspun tales:
1 Henry Ant was thirsty. He walked over to the river
where some é bank where his good friend Bill Bird was sitting,
: Henry sipped and fell in the river. He was unable
sso he can be
to call for help. He drowned. (1976, 128)
| ‘Meehan added noticing as an inference from changes in
e location, causing Bill to save Henty’ life. But another was
‘thur, Ie curns 4 not so lucky:
|
eeeeae Henry Ant was thirsty. He walked oversto the
| ld tha io 4 river bank where his good friend Bill Bird was
| sitting, Henry slipped and fell in the river, Gravity
|
+ going to lie
eo drowned. (129)
oeld struceure
| hat Acthur In this version of Tale-Spin, gravity was a character that
i 4 pesformed a Grup on Henry Ant and thea PTRansed
| request. Tell them both into the river. Henry got rescued by his friend
| an agent 10 . Bil, bot evi had no way nt. This was fied by changing
| hich requires E how gravity operates in terms of Schank’s conceptual
| rex problem, : dependency expressions —Prapeling Henty instead,
{ etnur is neat a Returning to our story, once everyone knows Arthur
133Expressive Processing
and George are near each other, Do-MTransis called and
Arthur asks George to tell him where some honey is,
resulting in a call to the procedure for handling favors:
Request. Now there is reason to ask the audience how
George views his selationship with Arthur. Tale Spin asks
if he likes Arthur (answer: little), if he trusts Arthur (a
little), if he thinks he dominates Arthur (not at all) if he
feels indebted to Arthur (aot ata), and if he thinks that
Arthur trusts him (a litte)
Here wie come to the main “psychological action” of
this Tak-Spin story. This action is enabled by the Fact that
character planning can make use of Tale Spins inference
mechanisms. George Bird uses these mechanisms to
speculate about the potential impact of answering
Arthur's request.
“The first inference is an obvious one: Artur will think
that George has told him there is honey somewhere.
Second, from what George believes about theirseationship,
George can make the inference that Arthur will believe that
hie believes there is honey in this place. Also, based on the
sclationship, he infers that Arthur wil believe there is honey
in that place. Finally, George can infer, based on this, that
Arne will make a plan to eat the honey.
George then takes all of these inferences and, in turn,
generates all the infezences from them, We might say that
he imagines worlds in which each of these things is true
But he doesn't find what he's looking for in any of them
‘As Mechan puts it, he “was looking for an inference that
said he was going to be happy (or sad), so that he could
decide in favor of (or against) Arthur's request” (1976,
183). So in consultation with the audience, it is decided
that George will lie to Arthur,
134
Wheathis
tells Arthur i
but some of
the systems
Bee, who ow
sedwood tree
is part of a1
the system m
and honey at
‘When he tell
be asked hos
“a little?” an
are added to
Another T
Now Arthur
‘As Meehan «
Do-MTI
renurns &
returns 1
‘The goa
Arthur n
honey—
which re
toward 3
DCONT
4976, 1
Nothing go
asked how hi
reference w
Tale Spin
honey. Arth
found, but
tests—he'si The Tale-Spin Effect
When this happens, theentire structure of what George
tells Arthur is created in the course of the Do-MTrans,
but some of it is only indexed to character memory, not
| ing favors:
| ence how a the system's memory. This leads to the creation of Ivan
| eSpin ass a Bee, who owns honey, which is ina beehive, which i in a
| s Arthur @ redwood ire, which s surrounded by some ground, which
| eal) i he 4 is part of a valle. ‘The physical locations ace indexed to
thinks that the system memory theyee seal But at first the hive, bee,
and honey are only in Geomge’s memory, and not believed.
| action” of When he tells Arthur about them, the audience needs to
hae face that z be asked how much Arthur trusts George. The answer is
*s inference Ee “a litle? and the inference maker determines that they
aanisms 10 ate added to Arthurs beliefs about the worl,
answering
; Another Trick
ar will chink: a Now Arthur thinks he knows where to find some honey.
somewhere. . ‘As Meehan notes,
-clationship, , DO-MTRANS retums to REQUEST which
believe that 3 returns £0 the first call of DO-MTRANS which
oo returns (0 TELL which returns to PERSUADE.
eed cn The goal of the persuasion bas been satisfied—
exeishoney 4 Archur now thinks he knows where there’s some
oy this, that 2 hhoney—so PERSUADE returns to DKNOW
| s which retuens to DCONT. ... All this has gone
toward achieving only she rst precondition of
snd, in tae0, 3 DCONT, finding out where some honey was,
ght say that . (1976, 184-185)
ings is tue. Nothing goes well for Arthur from here. The audience is
ay of them. asked how he feels toward Ivan (real feelings, in Arthurs
| erence that reference world, even if toward a fictitious being in this
| at he could : TaleSpin world), and he decides to ask Ivan for some
test” (1976, hhoney. Aithur travels to where he believes Ivan can be
eis decided 4 found, but when he gets there, Ta fails its precondition
| - tests—he’s not “close” to Ivan, because Tvan's not thereExpressive Processing
Axthur infers that George has lied, and proceeds to
distrust him, like him somewhat less, and believe George
is trying to deceive him. However, Arthur hasn't stopped
Dekoving that Ivan exists, only that he knows where Ivan
cxists, As it curs out, there’s no one to ask where Ivan
really is except George.
So Arthur heads back to George’s tee. Given that
failed goal, he
tsies the next Persuade planbox: bargaining, He infers that
George Would probably like 2 worm, and the initial serup
determined that Arthur knew the location of a worm. So
asking George where Ivan is still exists as
Arthur offers George a worm if he will tll him where to
find Ivan. George promises to do so (a possible world),
while making the decision to trick Arthur (@ different
possible world, which will only come to exist if triggered
by the worm’s arrival). After Arthur succeeds in getting.
the worm and bringing it to George, the bird eats it and
makes fan of him. With this planbox having failed, the
next isto threaten. But the audience, when asked, asserts
that Arthur feels 2 litle dominated by George, so he
won't ty it.
Having failed to find a way to ask Ivan for honey
directly, Arthur is moved by the next planbox of Tellto
try to get someone else to ask Ivan to give him honey.
He tries with George—there’s no one else—but this ails
even more pitifully than his ast attempt. Things continue
to unravel until TaleSpin decides there's nothing more
that Arthur knows to do to try to address his hunger
His options are particularly limited because thete’s
10 way for him to go looking for food—if he cannot
convince someone else to tell him, thete's no way to find
out, And as we remember from the setup for the story,
136
0 one knov
Aethur and
the story e¢
soon as the
problem, he
of them ha
Tale-Sp
“That was @
sore than
in this boo
undertaking
story syste
levels of ¢
Blige Dect
illustrates b
representa!
operational
But its
our TabSp
of fiction.
behavior
story. Ia a
in the gen
is present.
the audien
comput of
such sho:
return late
say thatThe Tale-Spin Effect
i
proceeds to i ‘no one knows about any bear food in this world, Between
ieve George Arthur and George they only know about one worm. So
sa’t stopped 4 the story could not have turned out any other way. As
where Ivan i ‘soon as the audience decided that hunger was Arthut’s
where Ivan, 4 problem, he was doomed. He made maay plas, yet none
of them had a chance of working: The end.
Given that
| led goal, he Tale-Spin’s Fiction
| infers that = ‘That was a significant amount of detail about Tals Spin,
| ini scrap more than T will offer about any other system described
| aworm, So 3 in this book. I hope it gave some sense of the Gpe of
sim where to a undertaking involved in creating even a first-generation
| sible world), story system, There's much more going on—at the
| (a difference levels of character and stozy—than in something like
| sif triggered Eliga/ Doctor or a standard computer RPG. Further, it
[sin going iustatesow-a compute ter that sc to generate
eats it and teptesentations of human behavior can be built as an
|g filed, the ‘operationalization of theories about human behavior,
| sked, asserts But it’s also worth noting that the story produced in
[porge, so he ‘our Tale-Spin example wase’t a particularly strong instance
| of fiction, While TalsSpin creates character behavior, this
| a for honey E behavior doesn't necessarily take the shape of a traditional
| ox of Tall a story, In addition, much of the action that takes place
him hones. a in the generation of such a story (as with our example)
| uc this fails is present only at the process level and imperceptible to
ings continue the audience. An audience member presented with the
| othing more output of this TalesSpin story would have seen something
| shunges ‘mach shorter and simpler. This is something to which I'l
use there’s return later. For now, I want to consider what it means to
| £ he cannot say that Tale Spin produces fiction at al.
| way to find
[oe the sonsExpressive Processing
10. Rye partway | Possible Worlds we plan (9
ote aca Van, ‘A shorthand definition of atin might be a listing of hhave great’
ieee ene its most familiar forms: “novels and short stories.” or logic more
DolseIolowedty | peshaps “novels, short stories, theater, television, and versions of
1 ulate movies.” Yet such definitions won't be much help to us impossible:
| Thoms Pare in if we are interested in thinking about emexping forms. of a numb
| te rst chapter of ‘Ac one point, the novel was an innovative new form of seturned
| [atoacene | fiction and would have been left out of such a definition well after dl
sie picture ‘Now there are varions innovative digital forms of fiction A simp!
[Wocsommesucay) (meting In order to think about new forms, we need a to fiction 1
af scar. | more principled definition of fiction. our world,
— In response to this, a numberof digital media theorists alternative
have begua to work with definitions of fiction emerging there are a
from possible worlds theory and modal logic. These ate this forrmal
among a wider group of literary approaches desived from berween fi
philosophy and linguistics, and this particular strand began, 2 in referens
to establish its own identity in the 1970s. In Possible Werks complex |
Artificial Inselliggns, and Narrative Theory, Marie-Laure Ryan alternative
(1992) traces the lineage of this work to the late-1970s both fcto
essays of 2 number of writers. 2 nonfiction
In philosophy the nation of possible worlds is often the people
traced back to Gottfried Wilhelm Leibniz, whose position 4 10 these is
is pethaps most widely remembered as that devoutly held the exeatic
by De. Pangloss in Voltaie’s Candide. Leibniz argued that : necessary
there are an infinity of possible worlds in the mind of : Ryan ¢
God—and that one of them, the best possible, is the one 2 simply the
in which we live. Leibnia’s position is engaged with the but the re
most elevated concerns of philosophy and theology, but 4 indexicalt
that is not the only place we find possible wotlds. Closer ae to refer 10
to home, we imagine possible worlds when we speculate 3 as “newual
(‘Twonderif the hunch special will be Lemongrass Tofu”), not only ¢
‘when we wish (‘Lhope Jane gets that promotion”), when : system of
138The Tale-Spin Effect
| 4 wwe plan (“Weill get there before the doors open, so we
| a listing of q hhave great seats”), and s0 on. Possible worlds, and modal
| stories.” of 3 logic mote general, aze tools for thinking throagh
| sision, and vvessions of such questions as: What is possible? What is
| help to us S impossible? What is necessary or contingent? The work
| ging forms. a ‘of a number of thinkers has, especially since the 1960s,
| ew form of e returned philosophers to consideration of these topics
| a definition, a ‘well after the time of Leibniz,
| as of fiction. : ‘A simple attempt at applying possible worlds theory
| swe need a to fiction might propose that nonfiction texts refer to
| out world, the real world, while fictional texts refer to
| dia theorists alteenative possible worlds. As Ryan points out, though,
| 5a emerging there are a number of problems with this. For example,
|e These ae this formulation does not provide a way of distinguishing
| Nezived from becween fiction, errors, and lies—all are statements made
[ strand began : in reference to alternative possible work’s. At a more
| ossible Works, complex level, there is also the problem that farther
| Laure Ryan, 5 alternative possible worlds are continually embedded into
| re late-1970s both fiction snd nonfiction. For instance, both fietion and
| nonfiction desesibe the untealized wishes and plans of
| ands is often : the people who appear in them. Ryan’s work responds
| rose position - to these issues by identifying a further element—beyond
| tevoutly held e the creation of an alternative possible world—that is
| 2 argued that ce necessary for fiction,
the mind of 5 Ryan considers the constituent move of fiction, not
sles the one simply the creation of an alternative possible world
ged with the but the reeamferiag of discourse to that world—so that
theology, but . indexical terms such as “here” and “now” are understood
| rosids, Closer ] to refer to the alternative possible world, and terms such
wwe speculate as “actual” become indexical. Further, for Ryan fiction
rgrass Toh”), a not only creates an alternative possible world but also a
otion”), when system of seality, a universe. This is necessary becuse
139Expressive Processing
the alternative world of a fiction may also have many
alternative possible worlds emanating from it, and each
of them may have further alternative possible worlds (as
‘when one character speculates as to the plans of another
character). Obviously, parallels with the operations of
‘Tale Spin were already at work in the description above,
Digital Fictions
Beyond the field of story genetation (an interest I share
‘with Ryai), a mumber of authors have used possible worlds
theory as a way of grappling with emerging forms of
fiction in digital media, For example, Jill Walker Rettberg,
ited catlier for her discussion of quests,.is a digital
‘media theorist interested in many forms of fiction with
a strong textual component (including email narratives
and weblog fictions) as well as some without text 2s a
primary component. In Fiction and Interaction: How Clicking
4 Mouse Can Make You Part of a Fictional World, Reterg,
(2003) examines intetactive digital fictions that somehow
include the user as a character in their alternative possible
‘world, She introduces the term “ontological interaction”
in the course of desctibing how users ate included in
these worlds, building particularly on the notion from
‘Thomas Pavel’s book of “ontological fusion” (between
‘our actual selves and fictional selves when engaging with
fictional worlds) and Kendall Walton's theory of how we
use fictional representations as props in a game of make-
believe. (She cites Walton's 1993 Minsis as Make-Beleve)
Walker notes that this approach provides a vocabulary for
discussing the common generation of fictional experience
in works “as disparate as installation artworks, interactive
narratives and computer games” (2003, 31).
140
Ina cel
theorist Jes
primary ob
in fiona:
that is an
understand
kinds of w
and how th
that game
are incom
world outs
on some k
‘world part
and other
of the wor
most litera
are, in Jou
the things
in terms ¢
worlds in
be explain
as the ma
‘without a
toward rei
going far
games cue
including:
box, and 1
and ramos
Given
and digital
possible wThe Tale-Spin Effect
Ina related vein, HaifReal—by game developer and
| have many 4
and ee | theorist Jesper Juul (2005)—takes computer games as the
2 worlds (as a ptimaty objects of study. Given this, Juul is not interested.
of another 4 in fiion as a category of artworks but rather as something
erations of . that is an element of many games. He is interested in
an above. ‘understanding how games project fictional woslds: what
Kinds of worlds, how players are cued to imagine them,
and how those worlds relate to the games’ rules, Juul notes
srest I share that game fictions, like those of traditional literature,
sible worlds are incomplete. Just as in Hamit we know litle of the
| g forms of world outside the castle and its immediate vicinity—but
‘er Rettberg, - con some level assume itis embedded in a fully detiled
is 2 digital 2 ‘world partially filled in by knowledge from ont own world
fiction with . and other texts —so in the game HafLfé we know litle
il nartatives e of the world outside the complex where it is set. Unlike
nut text as @ mostliterary fetions, however, the workds of many games
How Clicking : are, in Juul’ terminology, “incoherent” (which is one of
1d, Rettherg the things that limits Juul’ interest in discussing games
at somehow in terms of arrativ, 2s opposed to fiction). These are
tive possible wwotkls in which significant events take place that cannot
interaction” = be explained without discussing the game rules, such
| included in as the many games that feature multiple and extra lives
notion from. without any element of the game fiction that points
1” (between : toward reincarnation. Juul desctibes a number of ways,
agaging with : going far beyond traditional novelistic techniques, that
of how we : ‘games cue players to imagine possible fictional worlds,
ne of make- including: graphics; sound; text; cut-scenes; game title,
Make-Belewe) : bos, and mammal; haptics; ules; player actions and time;
veabuilary for ‘ and rumors (133-138).
alexperience Given that authors considering the concepts of fiction
3, interactive and digital media from differing directions have found the
possible worlds approach to fiction fruitful, how can it
1411. Ort putt omer
vagy, ais eves
and thes poraya
help one understand TaleSpin, as well as story generation
more generally? First, of cousse, it can clarify that Tale
Spi’s operations produce Sections. The psychological
action noted during George's speculations about Aethur
is revealed as precisely the stuff of fiction when seen
theough the lens of possible worlds theories. Second,
this approsch ‘provides a way to think about fictions
that is purely in terms of simcture This is important
because TalsSpin, like most story-genesation projects,
is concerrfed with producing the events of stoties—and
very litle with how those events are presented to the
audience. In literary theory this is sometimes referred
to as the distinction between “story” and “discourse”
(os, in some more specialized contexts, “fabula” and
“sjuzhet”).!" In. fact, Mechan was so litte concerned
with the presentation of Tale Spin ictions that Mumblo—
+ the naturallanguage generator used to turn TaleSpin's
conceptual dependency expressions into English—was
cobbled together in single day. (As I will discuss shortly,
however, the problems with TaleSpia’s ourput cun deeper
than Munblés stilted constructions)
Most important, seeing fiction in this light ean help
us identify the core operational logic at work in TalesSpin
planbox-based planning This method of planning, as
outlined by Schank and Abelson, operates by projecting
potential behaviors that change the state of the world—
possible worlds-—-which launch further projections, which
jn turn launch farther projections, finding a “chain of
lattice” of worlds that may lead from the current world
to one in which a goal is satisfied. The TalesSpin system
certainly contains other operational logies, such as those
‘governing movernent, character relationships, and so on.
142
But TalesSpi
that charact
‘happens onl
fleshed out
‘Tak Spin doe
selationships
feshes out
required by
contribute t
logic of pls
the profusic
fictions and
The Tale
Ta Tealo Ca
‘0 characte
garden. Pol
through sym
picees on a
vastempite.
building wit
model of th
constructed
d
of Euphem
In Eesli, ¢
hhovses—m:
solstice 2
and agency-
se taken dc
is continual
of the sta
precisely —iThe Tale-Spin Effect
generation But Tale Spin does not begin with a complete virtual world
yy that Tale that characters can move across. Instead, movement
| yehological happens only when required by plans, and the world is
out Arthur i fleshed out only to the degece required by plans. Similarly,
| when seen a TaleSpin does not generate a complete set of interpersonal
J es. Second, q relationships when chazacters ate exeated but rathet only
| at fictions Ee fleshes out the connections between characters that are
| important required by plans. So while other operational logies may
[on projects, contribute tothe Fictional works of Tali, it the
tories —and logic of planbox-based planning that is central, It creates,
| oted to the . the profusion of imagined worlds that define TalSpii’s
| es referred fictions and drive the work of the system's other logics,
“discourse” !
| Jabula” and s The Tale-Spin Effect
concerned, - In Italo Calvino's Inusibe Gitts (Calvino and Weaver 1974),
at Mrmble— two characters named Kublai Khan and Marco Polo sit i.
| a Tale Spins ; garden, Polo tells Klan-—sometimes in words, sometimes
aglish—was : through symbols, and sometimes through the relation of
| cuss shortly, 5 pieces on a chessboard—of cities he has visited within the
tun deeper : ‘vast empire. Here are afew In the middle of Fedora isa metal
building with a crystal globe in every room, each containing a
ght can help model of the city as it might have been in a possible future,
in Tale Spine S constructed at a different stage of its history. At every
planning, as 2 solstice and equinox, around thie fires of the marketplace
iy projecting 2 of Euphemia, there is trade not in goods but in memories.
the world— In Ersilia, the inhabitants stretch strings between all the
ctions, which houses—marking relationships of blood, trade, authority
a “chain ot : and agency—until one can no longer pass, all but the strings
asrent world Z ad Exsilia is built again elsewhere. Thekla
Spin systera is continually under construction, following the blueprint
uch as those a of the stars, while Andria already reflects the heavens
sand s0 on. precisely —in every street, building, job, and ceremons-—butExpressive Processing
those who live there must carefully weigh each change to the
city, given the changes it will produce in the heavens, Polo
and Khan each propose a model city, from which all others
can be deduced. They look through atlas pages that contain
ot only all the cities of Khan's empire but also all those
that will one day come to exist (Patis, Mexico City) and all
imaginary lands (Utopia, New Atlantis)
Tris not hard to picture Tale Spin as an addition to
this list of imaginary lands. It is the place made up of
nothing but plans within plans within plans-‘Phe people
have no emotions, except those that help select between
possible plans. They have no memories, except of plans
uundecway, plans that have failed, and the locations of
things they may use in plans, And these locations—the
very geographies of this imaginary place—come to exist
only as needed by their plans.
Like one of Calvino’s cities, TaleSpin is an alien
place. And yet, each is alien because some element that
wwe recognize of our own lives becomes the defining
clement, practically the only element, of the people and
landscape. On some level we do trade in memories lke the
inhabitants of Euphemia, clot free passage with networks
of connection lke the inhabitants of Esilia, and, like the
inhabitants of TaleSpin, make Chinese boxes of plans
within plans that at times obsess us so that nothing else
seems to exist,
Nevertheless, to those who consider only the output
of Tale-Spin fctions, al of this is truly invisible. Take, for
example, George Bird wying to decide how ro answer
Arthur Bear’ request to tell him where to find honey.
‘Meehan doesn’t provide us with the Mumbe output from
this fiction, but a story with a similar situation in it is
144
reprinted in
is the Mum
speculation:
such like t
Tom aske
there wer
Wilma we
‘The em
undoubted
story, if on
of TaleSpi
‘contain ace
or elaborat
all this psyc
only at the
made rando
Rather
structures,
from conce
fictional en
Once upd
was a be
inthe be
beeniv.
rest ina
Tom kre
that Mag
was ih
a bush.
Betty kn
And soon,The Tale-Spin Effect
| ange to the reprinted in an appendix co Mechan’s dissertation, Here
| eavens, Polo is the Mumble output of a seties of considerations and
| ich all others 3 speculations that, given TaleSin’s structures, is probably
| shat contain 3 ‘much like those in Arthur's and George’s story:
‘so all those 4 Tom asked Wilma whether Wilma would telt Tom where
| City) and all there were some berries if Tom gave Wilma a worrn,
‘iim was inctined to ie to Tom, (Meehan 1976, 232)
addition to
made up of : “The empty space between those two sentences is
‘The people : undoubtedly one of the most interesting parts of this
lect betweeo story, if only we could see it from the “Interior” view
| epr of plans of TaleSpins operations. But Murble stories never
locations of E contain accounts of characters’ multilevel speculations
sations—the e of claborate considerations of potential plans. Instead,
| come to exist = all this psychological action is, as above, elided. Looking
| only at the surface, the decision might as well have been
j- is an alien : made sindomly.
| clement that 4 Rather than TaleSpin’s most interesting story
the defining a structures, Mumbleoutputs plodding, detailed information
| e people and : from conceptual dependency expressions that have litle
rories like the : fictional energy, as with the beginning of this same story: .
with necworks (Once upon a time Betty Bear lived in a cave. There
» and, like the ; was @ beehive in an apple tree, Maggie Bee lived
ves of plans in the beehive, There was some Honey in Maggie's
nothing else - beehive. Tom Smith was ina chair. Thera was 2
ie rest in a redwood tree. Wilma Bird lived in the nest
‘Tom knew that Wilma was in her nest, Tom knew
ay the output
ena 7 that Maggie was in her beehive. Tom knew that Betty
eases was in her cave, There were some cranbertes near
a bush, There was a worm near 3 patch of ground,
0 find honey ne
Betty knew thatthe cranberries were nea the bush.
‘output from
uation in it is And 30 on, including painstaking reports of travel such asExpressive Processing
|
|
|
|
146
“Tom walked from the chair actoss a living zoom down in general
a hall via some stats down a hall down a hall through a in creating
valley across a meadow to the ground by the redwood es by comput
trce.” 1¢3 no wonder that most humanists have seen Tale 4 interesting
Spin as only worthy of ridicule. q andience ex
Further, unlike the open-ended textual interaction of a later chapte
| Elia/ Doctor, the interaction allowed by Tale-Spin takes 4 and what I ¢
| the form of highly restricted menu selection. No amount
| of play with che system will have the result of play with Tale-Sp.
Elza) Décor, insight into the processes at work. Instead, 4 ‘The Tab-Sp
| it would be impossible, through play, to pierce the boxing. Tate Spin as
surface of a TalsSpin story to see the more interesting 4 ‘was positios
fiction taking place through its processes, . 4 ‘As Meehan
| ‘One can look at it this way: The Blize effect creates 7 structures 0
| a surface illusion of system complexity—which play (if ae the most eff
allowed) dispels. The Tale Spin effect, om the other hand, If this were