Chris Crawford on Interactive Storytelling (33 page)

This approach suggests that perhaps something could be accomplished if the problem of interactive storytelling is turned inside out. Instead of trying to model the laws of drama, perform lots of clever processing, and then convert the results into language just before presenting it to players, why not make language itself the core of the process? The belief here— and it’s a belief, not a proof—is that language itself contains the core elements of drama.

The notion of designing the language to define the reality is tried and true. As far back as the early seventeenth century, philosophers realized that natural language was too sloppy to enable the kind of precise reasoning they wanted. This realization gave rise to a number of speculations about designing a philosopher’s language. It never got off the ground, but mathematics met part of the need for a precise language. Since then, all manner of special-case notational systems have been developed for governing certain classes of thought: mathematics, music, chemistry, nuclear physics, quantum mechanics, and so on. There’s the classic fictional example of designing a language to govern thought: NewSpeak, the language appearing in the novel
1984
, was designed to ensure that all citizens had good thoughts because there were no words for bad thoughts.

So I’ll ask you to grant, for the purposes of this chapter only, the possibility that a properly designed language could somehow contain the elements needed for interactive storytelling. This possibility leads to the real question: Can such a language be designed?

Because language and our perception of reality are intimately intertwined, it stands to reason that the size of any language reflects the complexity of our
perception of reality. This is the real reason that so many attempts to make computers understand natural language have failed. It’s not that computer scientists can’t express in algorithmic form the many types and structures of language; no, the killer problem is bringing enough knowledge of the real world into linguistic calculations.

Here’s an example: “Johnny, I want you to stop crying and take this pill right now!” What does this sentence mean? A typical computer program would assume that the speaker is demanding Johnny take possession of the pill, but any human knows that the speaker is requiring Johnny to
swallow
the pill. That’s because the verb “take” has a special meaning when used with pills. If you don’t know this tiny detail about reality, you can’t understand the sentence. It also helps to know that children cry when they’re reluctant to take their medicine and are usually referred to by diminutives (Johnny) rather than their formal names (John).

Here’s an even trickier example: “She flew into a rage when she noticed the lipstick on his collar.” You have no problem understanding exactly what that sentence means. But consider all the social information you brought to bear in figuring it out: the relationship between men and women, concepts of faithfulness in marriage, how lipstick can get onto a collar, and so forth. How’s a poor computer to keep up with all that knowledge?

Therefore, the dramatic reality that’s covered must be tightly defined to make a dramatic language usable. You’ll have to dispense with one of the storyteller’s favorite tricks: the complexly implicit realization. Lipstick on the collar is a perfect example of the kind of knowledge that no dramatic
sublanguage
(a vocabulary that’s much smaller than everyday working languages) could address.

The impossibility of using complexly implicit realizations might lead some storytellers to dismiss the potential of dramatic sublanguages. Lots of stories include elements like the lipstick on the collar—without these elements, how can you build good stories? My answer is that you can still build stories, but they can’t share this feature with more complexly crafted stories. That doesn’t mean stories built with dramatic sublanguages are crippled; it means they’re different. In the early years of cinema, there was no sound track; some old fogies insisted that without sound, storytelling was impossible. Early cinema artists worked out a new visual vocabulary, but they
never
really solved the problem of dialogue. Instead, they created stories that required little dialogue. Interactive storytelling
requires something similar. Fortunately, you don’t have to create an entire language as expressively powerful as, say, English or Chinese; you can get away with much smaller sublanguages.

To be computable, the vocabulary of a dramatic sublanguage must be small— smaller than the average person’s working vocabulary (about 5,000 words). Fortunately, a vocabulary of only a few hundred words should be adequate. Even more fortunately, there are numerous lists of the most common words in a variety of languages
¹
. Here are the 100 most common English words:

1. the
   
2. be
   
3. of
   
4. and
   
5. a
   
6. to
   
7. in
   
8. he
   
9. have
   
10. it
    
11. that
    
12. for
    
13. they
    
14. I
    
15. with
    
16. as
    
17. not
    
18. on
    
19. she
    
20. at
    
21. by
    
22. this
    
23. we
    
24. you
    
25. do
    
26. but
    
27. from
    
28. or
    
29. which
    
30. one
    
31. would
    
32. all
    
33. will
    
34. there
    
35. say
    
36. who
    
37. make
    
38. when
    
39. can
    
40. more
    
41. if
    
42. no
    
43. man
    
44. out
    
45. other
    
46. so
    
47. what
    
48. time
    
49. up
    
50. go
    
51. about
    
52. than
    
53. into
    
54. could
    
55. state
    
56. only
    
57. new
    
58. year
    
59. some
    
60. take
    
61. come
    
62. these
    
63. know
    
64. see
    
65. use
    
66. get
    
67. like
    
68. then
    
69. first
    
70. any
    
71. work
    
72. now
    
73. may
    
74. such
    
75. give
    
76. over
    
77. think
    
78. most
    
79. even
    
80. find
    
81. day
    
82. also
    
83. after
    
84. way
    
85. many
    
86. must
    
87. look
    
88. before
    
89. great
    
90. back
    
91. through
    
92. long
    
93. where
    
94. much
    
95. should
    
96. well
    
97. people
    
98. down
    
99. own
    
100. just
     

Here are the English words in the frequency range 2,100–2,200:

2,100 sometime

2,101 applaud

2,102 underneath

2,103 hello

2,104 pretense

2,105 descent

2,106 conquer

2,107 framework

2,108 confidential

2,109 adoption

2,110 disgust

2,111 waist

2,112 momentary

2,113 receipt

2,114 pearl

2,115 ray

2,116 lazy

2,117 limb

2,118 grammatical

2,119 beast

2,120 monkey

2,121 jewel

2,122 persuasion

2,123 obedience

2,124 sock

2,125 vowel

2,126 hammer

2,127 inn

2,128 chimney

2,129 dissatisfaction

2,130 annoyance

2,131 ornament

2,132 honesty

2,133 outward

2,134 sharpen

2,135 handkerchief

2,136 greed

2,137 heavenly

2,138 thirst

2,139 niece

2,140 spill

2,141 loaf

2,142 wheat

2,143 worm

2,144 secrecy

2,145 rude

2,146 heighten

2,147 flatten

2,148 loosen

2,149 cheese

2,150 rivalry

2,151 royalty

2,152 discontent

2,153 complication

2,154 right

2,155 indoor

2,156 lower

2,157 actress

2,158 congratulation

2,159 ounce

2,160 fry

2,161 everlasting

2,162 goat

2,163 ink

2,164 disappearance

2,165 reproduction

2,166 thicken

2,167 avoidance

2,168 spoon

2,169 strap

2,170 deceive

2,171 lengthen

2,172 revenge

2,173 correction

2,174 descendant

2,175 hesitation

2,176 spade

2,177 basin

2,178 weed

2,179 omission

2,180 old-fashioned

2,181 bicycle

2,182 breadth

2,183 photography

2,184 coward

2,185 mat

2,186 rejoice

2,187 cheat

2,188 congratulate

2,189 discomfort

2,190 enclosure

2,191 attentive

2,192 paw

2,193 overflow

2,194 dissatisfy

2,195 multiplication

2,196 whichever

2,197 tidy

2,198 bribe

2,199 mend

Note two observations about these lists. First, many words in the first list are function words, such as articles and prepositions, which can often be discarded in a sublanguage. Second, many words in the second list aren’t truly necessary for interactive storytelling. You could build a perfectly good sublanguage for interactive storytelling without using such words as “handkerchief,” “photography,” or “ray.” This means your vocabulary can be smaller than 2,000 words—a large but achievable goal.

Moreover, your vocabulary can be extended by a variety of means. All languages sport a set of prefixes and suffixes that extend the meaning of a root word in many directions. Here are just a few of the common prefixes and suffixes commonly used in English:

bak
er

fond
ness

nation
hood