[posted to rec.arts.int-fiction and alt.hypertext in August 1993]

            Issues in Interactive Fiction
                   Jorn Barger

I've been collecting my thoughts for a while, with an eye toward re-writing 
(with you all, as a group?) the 'theory' part of the raif-faq, filling in 
some more detail, and adding some of the threads we've explored this year.

How many different sorts of interactivity can we identify, in fiction, at 
this early stage of IF evolution?

One clear and concrete paradigm for IF is the choose-your-own-path style of 
book-design, where each short 'chapter' ends with a choice of directions 
you might go next (traditionally only two, for convenience).  If the book 
is written so as to make the reader the main character, these may be 
physical directions, like "If you choose to go north, turn to page 61" or 
actions like "If you choose to eat the cupcake, go to page 62".  (One might 
give the reader more godlike control over events, however: "If you want to 
know what was going on in Rangoon at this moment, turn to page 63" or "If 
you want to know what Bobo was thinking at this moment, turn to page 64" or 
"If you want a plague of locusts to descend, turn to page 65" ;^)

So a 'chapter' may represent not just a new place, but a new state of 
affairs in the same place, possibly involving the passage of time.  ("The 
clock on the wall now reads two a.m.  The guard is asleep at his desk.")

I think it's accurate to say that every Infocom-style text adventure 
*could* be re-published this way, in book form (which I'll call "book-
tech"), but almost all their chapters would then be just a sentence or two, 
like "You eat the cupcake.  It's very tasty." and after each such short 
chapter you'd have to be offered every possible combination of command-
words.  Almost-identical chapter-text would have to be repeated countless 
times with slightly different menus at the end. (If you could end a chapter 
with "Go back to where you just came from" that would help a lot!  But 
that's pretty advanced, for book-tech...)

A number of important consequences for IF theory follow from this single, 
simplest design element, *plot branching*:

1. Loss of authorial control:  the author now has to anticipate every 
possible reader path, and craft the layout of chapters so that every path 
offers *a reading experience the author can be proud of*.

2. The 'finished-it' paradox:  The first time you reach 'The End' you will 
not have finished reading all the author's text.  Are you then expected to 
go back and exhaust every alternate path, to 'get your money's worth' or to 
express your pleasure in her craft?  Can an author keep this interesting?  
(With book-tech, you can just reread it cover to cover, although you'll 
miss some of the subtler meanings this way.)

3. The multi-book paradox:  Should other paths offer greatly differing 
stories?  There's a limit to how much detail you can put into each 
alternate path!  With book-tech, for example, *choosing a point-of-view* at 
the start will necessarily imply choosing one of two entirely distinct 
sub-books.  On a computer, text-modification rules might make this more 
practical.

4. Dead ends:  The usual strategy for dealing with the multi-book paradox 
is to have a single happy ending and any number of dead ends, that usually 
involve the main character's (in fact) ending up dead.  Where book-tech 
might require the reader to use a bookmark to backtrack at this point, 
adventure games usually offer a save-game feature, or a password to begin 
again at an earlier point, or an Undo facility.  "Loom" is an exception, 
designed with no dead ends.

5. Added value:  There needs to be some added value from the interactive 
elements.  Some possibilities:  a) challenge of solving a puzzle  b) 
increased psychological involvement  c) learning rules of alternate world  
d) seeing the consequences of alternate lifechoices  e) beauty of chance
recombinations

6. Hide-the-happy-ending: A good IF designer will cleverly disguise the 
branches so that most players guess wrong at least a few times at each 
point.  But too many wrong guesses will be frustrating, and too few will be 
unchallenging.  This might be thought of as 'engineering an optimal 
literary path'.

7. Looping, and nonsequential episodes:  a path may return you to the exact 
same chapter.  And a single chapter may have *several* choices that lead 
back to it.  One can try to predict what order these loops will be tried 
in, but the text itself can't make assumptions about this.  Some sets of 
chapters might be freely experienced in any order-- David Graves calls this 
'browse around'.  ("If you want to visit the art gallery, go to page 67.")  
In a book, such chapters must be optional to the plotting, because you'll 
be able to skip them altogether.  (In this way they strongly resemble 
hypertext.)  On a computer, they can supply behind-the-scenes changes that 
eventually have a visible payoff.

8. Hidden information:  an optional chapter may provide information that 
allows one to guess the 'right' path at some later point.  (This is a 
cliche in computer adventures.)  On replays, though, these chapters can be 
safely skipped.

9. Linear sequences: Some chapters may have only a single choice at the 
end.  This makes most sense on a computer that's displaying graphics of 
each scene.  Graves calls this 'user-paced sequencing'.

10. Merging paths vs shared text:  with book-tech, if you can get to a 
given chapter by two different paths, nothing on those paths can possibly 
make any difference to the future course of events.  This is therefore a 
much more trivial sort of interactivity.  (Semantically, this might often 
involve what AI-folk call 'plan repair'.  One choice was the right plan, 
the other was wrong, but can be quickly corrected.)  Computers offer an 
improvement where multiple paths may share some text but not all text, so 
that only the 'right' path reveals the possibilities hidden for the others. 
Graves might class this as 'progressive disclosure'. (?)

11. Menus vs typing: on a computer you can allow the reader to type 
anything, or you can offer a menu of options.  In a book you're stuck with 
menus.  Using menus makes solving the puzzle a *lot* easier, which is 
usually not what you want. ("If you choose to look in the wastebasket, turn 
to page 66...")  With typing, though, where the possible inputs are almost 
infinite, most of the player's guesses will necessarily lead to boring 
dead-ends.

12. Literal menus vs tokenized menus: if the menu offers "If you choose to 
tell the creepy guy to bug off, turn to page 67" and page 67 says "You tell 
the creepy guy to bug off" ...that's boringly repetitive.  One solution is 
to offer 'tokens' in the menu, like "warm response" "cool response" "cold 
response", which are expanded with more humor or interest in the following 
text, instead of simply repeating the menu choice.

=----------=-    ,!.    --=----=----=----=----=----=----=----=----=----=----=
Jorn Barger      j't    Anon-ftp to genesis.mcs.com in mcsnet.users/jorn for:
  <:^)^:<    K=-=:: -=->   Finnegans Wake, artificial intelligence, Ascii-TV,
 .::.:.::..   "=i.: [-'   fractal-thicket indexing, semantic-topology theory,
jorn@mcs.com   /;:":.\     DecentWrite, MiniTech, nant/nart, flame theory &c!
=----------=  ;}'   '(,  -=----=----=----=----=----=----=----=----=----=----=