NPC Design/Best of Three

2 views
Skip to first unread message

ems...@mindspring.com

unread,
Nov 16, 2001, 3:57:36 PM11/16/01
to
Thanks very much to everyone who played Best of Three and gave me
feedback about how it worked for them. This was a very difficult game
to write; a good deal of that difficulty lay not in devising the code
(which was heinous, but at least I knew what effect I wanted to
achieve) or the dialogue (which was easy enough), but trying to guess
how to make a character who would have the proactivity, versatility,
and goal-seeking behavior of a human being, while still giving the
player enough of a chance to participate in shaping the conversation.

That it apparently failed, at least in some respects, is much less
important to me than *how* it failed: from my point of view, the
underlying library finally begins to approximate an acceptable tool
for NPC creation (though I think it would be pointless even to attempt
porting it back to the z-machine, unfortunately), and the trick now is
to learn how to incorporate conversation of this type into the flow of
game play. The message seems to be that if people have options
available all the time, they'll feel railroaded and led by the nose
(or at least CYOA'd); if the menu runs out too often (a la Pytho) or
if there's obscure ASK/TELLness (a la Galatea), they get stuck.
Presumably somewhere in here is a middle road that lets the
conversation flow sometimes, but sometimes come to a bottleneck of
enforced decisionmaking? Not sure. More testing needed. (Returning
to the laboratory to stitch together the new monster...)


Whether you thought the NPC was obnoxious is another question: I wrote
Helen a fair number of rude responses that you could only get to by
using >TOPIC, because I wanted her default characterization to be
somewhat wimpy and allow the player to guide her into more assertive
behavior. So I'm not a 100-percent Grant-endorser either, and frankly
I'm rather fond of the ending where he realizes what a jerk he's been.


In any case. Those who have asked about the mechanics and background
of the game may be interested in:

http://emshort.home.mindspring.com/Bo3About.htm

and

http://emshort.home.mindspring.com/NPC3.htm

(yes, this is the same old NPC essay, but with some modifications and
the addition of a couple of sections. Those who have already read the
essay will find the new sections indicated as such in the outline.)

Cheers--

ES

Daniel Barkalow

unread,
Nov 16, 2001, 8:55:24 PM11/16/01
to
What bothered me about Best of Three was that I found I had no idea how to
explore the space of the game. On replaying it, the conversation
took a somewhat different set of turns, and I had no real idea of what I'd
done differently or why it had the effect it did. There was plenty of
local coherency in the conversation, but the larger-scale map didn't make
sense. As a result, I was left with the feeling that there were other
endings I hadn't found, but no idea of how to find them.

The conversation system really ought to tell you what topic you're on, in
case you want to come back later, and also let you switch back to the
topic you were on previously, to let you fight getting
sidetracked. Possibly some sort of "topic inventory" of things the PC has
obviously thought of to talk about would be nice. Also, I think it would
be useful to, rather than having the NPC just give you more chances to say
something of your choice, allow the player to say things without giving up
the conversational turn. I.e., instead of just being able to choose
"1" from the menu, be able to choose "1" or "1+", where the latter would
give you a chance to say more without stopping; of course, you could be
interrupted, but normally wouldn't be.

I don't think "think about" shouldn't take a turn. While some people
obviously can't think with their mouth open, it's disconcerting for this
to be a game mechanic. There were lots of times I would have liked to
consider my options with some advance knowledge of what I'd have to say
about them. If you're trying to be attentive, you barely get a chance to
consider your backstory at all (aside from what you might say when things
come up), and that's annoying, since you've spent the whole afternoon
there and also wouldn't have to think hard anyway.

-Iabervon
*This .sig unintentionally changed*

OKB -- not okblacke

unread,
Nov 16, 2001, 10:59:57 PM11/16/01
to
Daniel Barkalow iabe...@iabervon.org wrote:
>What bothered me about Best of Three was that I found I had no idea how to
>explore the space of the game. On replaying it, the conversation
>took a somewhat different set of turns, and I had no real idea of what I'd
>done differently or why it had the effect it did. There was plenty of
>local coherency in the conversation, but the larger-scale map didn't make
>sense. As a result, I was left with the feeling that there were other
>endings I hadn't found, but no idea of how to find them

In one sense, I agree with you, but in another sense, this was one of the
few things I really liked about this game. I really like the idea of hiding
the branch points so well that the player cannot tell what branch he is on
(without playing many times and comparing the various sequences), or indeed
whether he has branched at all. Best of Three did this well.

The problem, for me, was that the branches weren't distinct from one
another. I played the game twice and played it half-through 2 more times. In
each case, what I got was a big conversation, which ran the gamut from the dull
to the pretentiously grand, between two people who seemed to me like stuck-up
jerks. It wasn't even that one play-through gave me a dull conversation and
another gave me a pretentiously grand one; each playing gave me essentially the
same elements in a different order -- and (and this is what ticked me off more
than it probably should have) I was making a conscious effort to do things as
differently as I could.

So basically, I like the idea of hidden branches, but the branches in
Best of Three seemed not only hidden but indistinguishable.

--OKB (Bren...@aol.com) -- no relation to okblacke

"Do not follow where the path may lead;
go, instead, where there is no path, and leave a trail."
--Author Unknown

Eytan Zweig

unread,
Nov 17, 2001, 4:36:44 AM11/17/01
to
>I don't think "think about" shouldn't take a turn. While some people
>obviously can't think with their mouth open, it's disconcerting for this
>to be a game mechanic. There were lots of times I would have liked to
>consider my options with some advance knowledge of what I'd have to say
>about them. If you're trying to be attentive, you barely get a chance to
>consider your backstory at all (aside from what you might say when things
>come up), and that's annoying, since you've spent the whole afternoon
>there and also wouldn't have to think hard anyway.
>

Er, I think you mean that you DO think that "think about" shouldn't take a
turn, in which case I absolutely agree. Sure, in real life, it's possible to
get lost in thought during a conversation and this takes time. But here, the
"think about" was more of a glossary of backstory terms which would be
familiar to the character without deep thought, rather than a true "think
about" feature.

Eytan

> -Iabervon
>*This .sig unintentionally changed*

------------------------------------------------------------
Get your FREE web-based e-mail and newsgroup access at:
http://MailAndNews.com

Create a new mailbox, or access your existing IMAP4 or
POP3 mailbox from anywhere with just a web browser.
------------------------------------------------------------

Jon Ingold

unread,
Nov 17, 2001, 11:05:55 AM11/17/01
to
My own opinion on it:-

I think the stroke of genius in the current system is the TOPIC command. The
ASK/TELL system of Galatea is loosely speaking a very good one, because it
allows for & forces the player to *think* within the setting of the game in
order to provoke some conversation. CYOA games tend not to be immersive
because you can stare at the various options and bash the numbers 1 thru 4
till something happens; it's difficult to find a goal and it's difficult to
feel connected. However, with ASK/TELL, the player needs to consider the
questions: What does this NPC know about? What is this NPC "interested" in?
What will engage the conversation?

But the joy of the menu is that you can see what you're saying, and
moreover, you can choose different approaches on it (ASK, TELL, are often to
vague to feel any 'control' over). Shall I be positive on this topic, or
negative? Shall I dismiss it?

So you have the TOPIC command - the player thinks, immerses himself in order
to find something which the game will engage on. There is a trade-off,
between what the game wants to do and what the player wants to do. The
player's thought-process will drag them into the game world. Topics of
interest, revelations (however incidental) will feel earned and therefore
*special* to the player; like all the smells and sounds in So Far which made
the world that much richer.

Now - I guess you know all this, Emily, as you came up with the thing. ;)
But I think the important issue is emphasis - the system needs to rely, I
would say, more heavily on the TOPIC rather than the extant menu. I found on
playing it that most of my topic choices produced nothing new. So - when I
followed the conversation as it was, the experience was very enjoyable -
when I tried to bring myself to bear upon it, it ignored me.

I don't know how difficult is to cover the possibilities of TOPIC and assess
the various states of current conversation too (I would suspect the answer
is "very very"), but that I think is where the work is needed. Rather than a
new interface - the current one is very nice indeed - it needs a shift
towards, well, I guess re-obfuscation.

[But then, I like puzzles, so I guess I would say that.]

Thoughts?

Jon

[Incidentally - I played it twice: the first time, trying, desperately, to
do something, get somewhere, move the plot along a bit, provoke a reaction,
and the game ended and I was surprised. Then, I played it again, this time
realising the conversation was the *point* of the thing, and not a
interlude, and I enjoyed it thoroughly, and was a lot more engaged. For some
reason, whereas with Galatea I realised that "the play's the thing", here I
didn't; perhaps it was badly flagged, or perhaps I was just dense - which is
not unlikely. Anyway, it could be fairly argued my desire for a greater span
of topics is outside the point of the game.]


OKB -- not okblacke

unread,
Nov 17, 2001, 12:00:51 PM11/17/01
to
"Jon Ingold" ji...@cam.ac.uk wrote:
>CYOA games tend not to be immersive
>because you can stare at the various options and bash the numbers 1 thru 4
>till something happens; it's difficult to find a goal and it's difficult to
>feel connected. However, with ASK/TELL, the player needs to consider the
>questions: What does this NPC know about? What is this NPC "interested" in?
>What will engage the conversation?

I think the problem you mention with ASK/TELL is carried over into the
topic/menu system. In fact, it is one of the fundamental reasons why I dislike
ASK/TELL. (The other reason is that I can't tell what the wording of my
dialogue will be, but the menu system does solve that.) Basically, I don't
like having to rack my brain for topics to ask the character about, and to me
whether the topics are communicated via ASK/TELL or an explicit TOPIC command
is not significant.

Something about Best of Three in particular, though, that made it not
work for me, was what I mentioned in another post: the various branches were
too similar. In other words, from my perspective, talking about Grant's
parents, asking Grant about college, and chewing the fat about old times were
all essentially equivalent. As such, if I were to write a game about these
same two people having a conversation, I would be more likely to abandon the
TOPIC command and (where necessary) expand the menu to 5 or 6 choices rather
than just 3 or 4.

I guess what I'm saying is that I LIKE it when the various options on a
single menu relate to different topics. I would prefer to see "The butler did
it!", "Who do you think did it?", "Nice tie.", "Kiss me, you fool.", "I hear
John was arrested yesterday.", and "Have you ever been to the 13th floor?" all
on one menu, rather than have to topic-switch to see them all.

>There is a trade-off,
>between what the game wants to do and what the player wants to do. The
>player's thought-process will drag them into the game world. Topics of
>interest, revelations (however incidental) will feel earned and therefore
>*special* to the player; like all the smells and sounds in So Far which made
>the world that much richer.

Hrm, to me it's just the opposite. Unless every topic I think of is
provided for, it breaks mimesis. In a real-life conversation ANY topic is fair
play, although many may not lead anywhere. Having to guess at which topics
this particular game cares about is unappealing.

Now, obviously we can't have an NPC that responds to all the infinite
topics of conversation that are possible in real life -- and that's why I like
menus. They make it clear that THESE topics, and no others, are the
conversational domain right now, thus relieving the player of the burden of
having to feel out the boundaries of the NPCs knowledge, which is a separate
task from that of actually extracting information from them.

I.R.

unread,
Nov 17, 2001, 12:32:39 PM11/17/01
to
bren...@aol.comRemove (OKB -- not okblacke) wrote in message news:<20011116225957...@mb-mv.aol.com>...

> It wasn't even that one play-through gave me a dull conversation and
> another gave me a pretentiously grand one; each playing gave me essentially the
> same elements in a different order -- and (and this is what ticked me off more
> than it probably should have) I was making a conscious effort to do things as
> differently as I could.

I concur. This is also a problem I have encountered with the "City of
Secrets" trailer: you choose options, but this often fails to make any
difference - same elements pop up, only in different order. This
defeats the very purpose of using menus for conversation (steering the
game in a certain direction): there are branches, but, as has been
pointed out, they hardly differ.

-I.R.

ems...@mindspring.com

unread,
Nov 17, 2001, 3:09:04 PM11/17/01
to
Eytan Zweig <eyt...@MailAndNews.com> wrote in message news:<3BF8...@MailAndNews.com>...

> >I don't think "think about" shouldn't take a turn. While some people
> >obviously can't think with their mouth open, it's disconcerting for this
> >to be a game mechanic. There were lots of times I would have liked to
> >consider my options with some advance knowledge of what I'd have to say
> >about them. If you're trying to be attentive, you barely get a chance to
> >consider your backstory at all (aside from what you might say when things
> >come up), and that's annoying, since you've spent the whole afternoon
> >there and also wouldn't have to think hard anyway.

A number of people have said that they don't like this feature at all
and that they consider it a cop-out for "real exposition," so I'm not
sure whether it is likely to reappear in future games in any case. My
intention was to provide exposition that you could experience
optionally, so that it wouldn't be clogging up the text in replay
games where you already know all the background yadayada. Maybe I
shouldn't have worried about that: one's eye tends to skim efficiently
on down text one has already read, so repeating it may not be such a
cardinal sin.

In any case, this is not really part of the conversation system, but a
detachable feature.



> Er, I think you mean that you DO think that "think about" shouldn't take a
> turn, in which case I absolutely agree. Sure, in real life, it's possible to
> get lost in thought during a conversation and this takes time. But here, the
> "think about" was more of a glossary of backstory terms which would be
> familiar to the character without deep thought, rather than a true "think
> about" feature.

What would a true "think about" feature do?

ES

ems...@mindspring.com

unread,
Nov 17, 2001, 4:18:42 PM11/17/01
to
Daniel Barkalow <iabe...@iabervon.org> wrote in message news:<Pine.LNX.4.21.011116...@iabervon.org>...

> What bothered me about Best of Three was that I found I had no idea how to
> explore the space of the game. On replaying it, the conversation
> took a somewhat different set of turns, and I had no real idea of what I'd
> done differently or why it had the effect it did. There was plenty of
> local coherency in the conversation, but the larger-scale map didn't make
> sense. As a result, I was left with the feeling that there were other
> endings I hadn't found, but no idea of how to find them.
>
> The conversation system really ought to tell you what topic you're on, in
> case you want to come back later, and also let you switch back to the
> topic you were on previously, to let you fight getting
> sidetracked. Possibly some sort of "topic inventory" of things the PC has
> obviously thought of to talk about would be nice. Also, I think it would
> be useful to, rather than having the NPC just give you more chances to say
> something of your choice, allow the player to say things without giving up
> the conversational turn. I.e., instead of just being able to choose
> "1" from the menu, be able to choose "1" or "1+", where the latter would
> give you a chance to say more without stopping; of course, you could be
> interrupted, but normally wouldn't be.

This is an interesting idea, and one that would, in fact, be not
especially hard to implement, though it would make it impossible to
have the conversation menu clickable. (Whether there's any great
advantage there I don't know. I personally never use the clicking
ability, but I figured, okay, we have Glulx, we have a menu, might as
well make it as friendly to various types of input as possible.)

The thing is, of course, that conversation dynamics don't break down
easily into discrete temporal chunks any more than, say, the burning
of a log of wood. There's a progression; small pauses signify things
to one's conversational partner, but one would prefer not to have to
type >WAIT on every occasion when it is all right for the NPC to take
the lead. I've occasionally thought about an experiment in realtime
conversation, where you'd have options, but if you didn't select them
immediately, the NPC would react to your silence and move on. The
irritations of such a scheme are obvious -- you'd have to be a
reasonably fast reader, for one thing, and the options menu would be
constantly morphing.

Anyway, I guess that the objection I see with the 1+ scheme is that it
adds complexity (in the number of possibilities I'd have to code for;
in the number of options the player would have to take into account),
without necessarily adding that much of an advantage. The problem
here is that players are finding it too easy to play through the game
on autopilot. So the trick is how to >force< them to make choices
more often (preferably ones involving thoughtful interaction, not the
selection of something obvious off the menu), not how to >allow< them
to choose more.

[BrenBarn and Jon Ingold seem to have argued eloquently for both sides
of the ease-for-player vs. challenge-for-player debate. Personally
I'm in Jon's camp, obviously; and if there aren't a satisfying number
of topics to which the NPC will respond, then I need to write more
topics. Simple as that. I passionately dislike the idea of giving
the player a complete list of topics to which the NPC will respond,
because that reduces the character, again, to a box whose dimensions
are known, or a machine with a really large number of buttons that can
be pressed in any order. And the IF world has plenty of those
already.

A lot of the time I spent on Bo3 was originally intended for writing
and expanding the dialogue, but wound up hijacked for working on the
library itself. As a result, there isn't as great a diversity of
content as I'd hoped, which doubtless contributed to many of
BrenBarn's objections.]

ES

Adam Thornton

unread,
Nov 17, 2001, 5:03:44 PM11/17/01
to
In article <a69830de.01111...@posting.google.com>,

ems...@mindspring.com <ems...@mindspring.com> wrote:
>The thing is, of course, that conversation dynamics don't break down
>easily into discrete temporal chunks any more than, say, the burning
>of a log of wood.

"Speaking of which," continues Emily, "have you looked at my Peacock
library?"

Adam

Eytan Zweig

unread,
Nov 17, 2001, 6:31:00 PM11/17/01
to
> > Er, I think you mean that you DO think that "think about" shouldn't take
a
> > turn, in which case I absolutely agree. Sure, in real life, it's
possible to
> > get lost in thought during a conversation and this takes time. But here,
the
> > "think about" was more of a glossary of backstory terms which would be
> > familiar to the character without deep thought, rather than a true
"think
> > about" feature.
>
> What would a true "think about" feature do?
>

Well, what I'd like to see in a "think about" system is that thinking about
things should (at least sometimes) change the way you feel about them - for
instance, say that you have an NPC conversation game, where the PC is having
a disussion with his friend Sue

Sue: Mary just told me that she's getting married to David
1. Wow, I didn't even know they were dating.
2. Really? He deserves better than her.
> Think about mary's wedding
Slowly, Sue's news about Mary's impending wedding to David sinks in. While
you never liked Mary, you come to realize that she might indeed be the
perfect mate to David - after all, they not only share the same interests,
but he always wanted a woman who could mother him - and after all, that's
exactly what annoys you most about Mary.
1. Wow, I didn't even know they were dating.
2. You know what? I think that match just might work.

I'm assuming that the thinking does not move the turn forward, in this
example. Maybe it should. Also, you'd have to keep track of context -
thinking about the marriage before you hear about it should give you a very
different response. Sometimes, of course, thinking about things should be
counter-productive - maybe the "gut response" you had before was actually
the right thing to say?

Eytan

> ES


Sean T Barrett

unread,
Nov 17, 2001, 7:14:11 PM11/17/01
to
ems...@mindspring.com <ems...@mindspring.com> wrote:
>but one would prefer not to have to
>type >WAIT on every occasion when it is all right for the NPC to take
>the lead.

This, in fact, is how most of the conversation in The Weapon
worked, and a lot of people (well, a lot of betatesters)
complained about the experience. Of course, The Weapon isn't
a game about conversation at all, and it was intended to
tempt you into playing with stuff while listening, *exactly
as the PC would have*, but sadly it didn't really work.

>I've occasionally thought about an experiment in realtime
>conversation, where you'd have options, but if you didn't select them
>immediately, the NPC would react to your silence and move on.

This is almost how Doug Sharp's King of Chicago worked. There
were only small snippets of text to read, and there were
generally only two choices, and if you didn't pick one fast
enough, one was chosen for you.

But KoC was intended to feel like a participatory movie,
and, at least in non-multimedia IF, I suspect it's simply
the wrong direction to go--in much the same way that all
the subtle improvisation and planning in turn-based
hack-n-slash "RPG" Nethack turned into frantic button-mashing
in the obiously-derivative real-time "RPG" Diablo. (Which
is not to say that Diablo isn't fun, at least for some
people, but it appeals on a different level and to a
rather different set of skills.)

In other words, the indirect experience of text seems to go
hand-in-hand with non-real-time interaction, IMO.

>Simple as that. I passionately dislike the idea of giving
>the player a complete list of topics to which the NPC will respond,
>because that reduces the character, again, to a box whose dimensions
>are known, or a machine with a really large number of buttons that can
>be pressed in any order. And the IF world has plenty of those
>already.

I think arguably if you nest menus deep enough you can
regain that level of complexity; providing a list of topics
to choose from the size of a dictionary doesn't really
make the game any more trivial. But this requires having
an awful lot of topics available. Far better is when
the set of choices are synthesized in a combinatorial
fashion, as in traditional IF "verb noun" pairs, but I
don't see any reasonable way to apply that model to
conversation topics.

One of the Lotech comp games, the archaeologist one
whose name escapes me now, demonstrated reasonably well
how you can still get a relatively open-ended feeling
while restricted entirely to small menus--both by
keeping an underlying state that wasn't tied to the
menus, and by using sub-menus to expand out the details
of options.

To a certain extent I'm devil's-advocating here, though.

SeanB

OKB -- not okblacke

unread,
Nov 17, 2001, 7:22:09 PM11/17/01
to
ems...@mindspring.com wrote:
>>Also, I think it would
>> be useful to, rather than having the NPC just give you more chances to say
>> something of your choice, allow the player to say things without giving up
>> the conversational turn. I.e., instead of just being able to choose
>> "1" from the menu, be able to choose "1" or "1+", where the latter would
>> give you a chance to say more without stopping; of course, you could be
>> interrupted, but normally wouldn't be.
>
>This is an interesting idea, and one that would, in fact, be not
>especially hard to implement, though it would make it impossible to
>have the conversation menu clickable. (Whether there's any great
>advantage there I don't know. I personally never use the clicking
>ability, but I figured, okay, we have Glulx, we have a menu, might as
>well make it as friendly to various types of input as possible.)

Well, you could have clicking on the main text of the dialogue in the mneu
be the same as "1", and, say, clicking on an ellipsis at the end be "1+". So
the menu might show: "1: Then there was the time we all had our taxes done by a
platypus. . . ." If you click on the text itself, you say it and that's that,
but if you click on the ellipsis, you say it and get a chance to say more.
(Admittedly, this is not very intuitive; you'd have to alert the player to this
in an introductory text or something.)

>The thing is, of course, that conversation dynamics don't break down
>easily into discrete temporal chunks any more than, say, the burning
>of a log of wood. There's a progression; small pauses signify things
>to one's conversational partner, but one would prefer not to have to
>type >WAIT on every occasion when it is all right for the NPC to take
>the lead. I've occasionally thought about an experiment in realtime
>conversation, where you'd have options, but if you didn't select them
>immediately, the NPC would react to your silence and move on. The
>irritations of such a scheme are obvious -- you'd have to be a
>reasonably fast reader, for one thing, and the options menu would be
>constantly morphing.

Heh, yes. I can just see a game where the statusline shows the time,
accurate to one hundredth of a second, and you type things like >WAIT FOR HALF
A SECOND or >SAY "LOOK OUT!" IN .1 SECONDS or >SAY "I'M TIRED" IN 2 SECONDS.

But seriously, folks, I agree that a real-time conversation would be
interesting, but probably not the best thing for an IF game.

>The problem
>here is that players are finding it too easy to play through the game
>on autopilot. So the trick is how to >force< them to make choices
>more often (preferably ones involving thoughtful interaction, not the
>selection of something obvious off the menu), not how to >allow< them
>to choose more.

I think the trick is to give them options that they WANT to choose, rather
than bullying them into choosing one "against their will" -- and this is
probably more a function of good character development than of interface
design. In other words, you can have glitzy menus and whiz-bang jiggery-pokery
up on the screen, but if the various choices are banal, the player sits there
going "I have to say THAT?" (I should know -- I've written at least one game
with less-than-gripping dialogue on the menu.) But if the dialogue options are
diverse and interesting, the player will be eager to choose one to see what the
result will be.

>Simple as that. I passionately dislike the idea of giving
>the player a complete list of topics to which the NPC will respond,
>because that reduces the character, again, to a box whose dimensions
>are known, or a machine with a really large number of buttons that can
>be pressed in any order. And the IF world has plenty of those
>already.

I think the crux of this "the NPC is a button box" question is that the
player only knows the dimensions of the box at that instant. As soon as a menu
option is chosen (or, with the right coding, as soon as something else
happens), the menu can change. If the conversation is a tree, the player can
only see the branches immediately ahead of him, and has no way of knowing what
lies further on. The "lawnmower" argument that, by using UNDO, the player can
run the NPC into the ground applies to ASK/TELL as well. >ASK NPC ABOUT X.
>UNDO. >ASK NPC ABOUT Y. >UNDO. >ASK NPC ABOUT Z. >UNDO. And so forth.
Regardless of the mechanism, the player can only travel up the tree as many
turns as his UNDO capability (i.e., if you take 2 turns choosing menu options
and can only undo 1 turn, you cannot go back to the original menu and take a
different path).

>As a result, there isn't as great a diversity of
>content as I'd hoped, which doubtless contributed to many of
>BrenBarn's objections.

I must admit I am having to think carefully about which of my complaints
are against the topic system in general and which arise specifically from Best
of Three.

It is, I think, possible to devise a compromise system, wherein the menus
alone provide access to all nodes of the conversation tree, but the TOPIC
command allows more rapid travel. So that, for example, you could (in the
conversation) get from history to quantum physics via botany and sex, or you
could simply type TOPIC QUANTUM.

Jon Ingold

unread,
Nov 18, 2001, 9:35:05 AM11/18/01
to
> whether the topics are communicated via ASK/TELL or an explicit TOPIC
command
> is not significant.

Well, yeah, this is what I was saying. The ASK/TELL topic-selection is in my
view a _positive_ thing; marred by the vagueness of it. Which is solved by
the TOPIC system.

> Hrm, to me it's just the opposite. Unless every topic I think of is
> provided for, it breaks mimesis.

Er, really? That's sort of equivalent to saying: Unless every article I want
to read in the Hitchhiker's Guide isn't there, it breaks mimesis; or unless
every action I ever think of performing is specifically dealt with. I think
in trying to create a NPC-dialogue of the sophistication Emily's going for,
she's raised the bar incredibly high for herself, but we still play the game
within the context of it being a game; and just as I don't expect it to
recognise "GO NORTH SWIFTLY" I don't expect it to recognise everything I
think of. It certainly won't throw me out of the game-world. [Unless,
perhaps, the topic is something totally obvious (eg. "HIMSELF", "MYSELF"),
and there's a continuum of obviousness to consider].

> Having to guess at which topics this particular game cares about is
unappealing.

You shouldn't need to guess, as such; there should be clear threads in the
dialogue as it occurs, prompts, that sort of thing. Much like a puzzle game
which subtly leads you toward its solutions, or Shade which [for the most
part] does its level best to make you head for the next step without
realising you're doing it.

--

As I see it, Emily has done a lot of extremely impressive work, but also
brought herself back around to where Galatea started -- that in the final
analysis, you have to provide lots and lots of different subjects to chat
about, and let the player work to find a way to go. I think viewing the
system as an extended ASK/TELL is a better way of going than viewing it as
an extended conversation menu.

But the other thing here is that we've also just reduced our opinions back
to ASK/TELL vs. Conversation Menu, more or less - topic guessing versus list
progression. Which has been debated before, and generally is considered is
an opinion rather than a Fact either way.

Jon

OKB -- not okblacke

unread,
Nov 18, 2001, 11:53:22 AM11/18/01
to
"Jon Ingold" ji...@cam.ac.uk wrote:
>> Hrm, to me it's just the opposite. Unless every topic I think of is
>> provided for, it breaks mimesis.
>
>Er, really? That's sort of equivalent to saying: Unless every article I want
>to read in the Hitchhiker's Guide isn't there, it breaks mimesis; or unless
>every action I ever think of performing is specifically dealt with.

Well, yeah. I mean, aren't we generally agreed that "you can't see any
such things" and "I don't recognize that verb" are, in general,
mimesis-breakers? As you say, because it is a game, we are able to accept a
certain disparity between the game and the real world.

I do not, when I play a game, expect every possible command to work. When
I type an individual command, however (that command being suggested by
something I've already seen in the game), I expect that command to work.
Similarly, at the outset I don't expect a game to recognize QUANTUM PHYSICS as
a topic, but if it turns out the PC is a quantum physicist, I will come to
expect it.

>[Unless,
>perhaps, the topic is something totally obvious (eg. "HIMSELF", "MYSELF"),
>and there's a continuum of obviousness to consider]

You're right, I was too sweeping in my generalization, and I've revised my
statment above. The fact remains, though, that I have yet to see an ASK/TELL
game which covers even the "obvious" topics.

>> Having to guess at which topics this particular game cares about is
>unappealing.
>
>You shouldn't need to guess, as such; there should be clear threads in the
>dialogue as it occurs, prompts, that sort of thing.

But isn't the "freedom" of ASK/TELL lost if you expect the player to just
follow these prompts? If the player is limited to asking about only what you
goad him into asking about, it's a false freedom.

>But the other thing here is that we've also just reduced our opinions back
>to ASK/TELL vs. Conversation Menu, more or less - topic guessing versus list
>progression. Which has been debated before, and generally is considered is
>an opinion rather than a Fact either way.

I agree.

Tom Smith

unread,
Nov 18, 2001, 4:05:20 PM11/18/01
to
ok, I want to play this game

how do I get hold of a blorb interpreter? I'm too stupid to find one myself...

Tom Smith

Jon Ingold

unread,
Nov 19, 2001, 7:37:37 AM11/19/01
to
> >You shouldn't need to guess, as such; there should be clear threads in
the
> >dialogue as it occurs, prompts, that sort of thing.
> But isn't the "freedom" of ASK/TELL lost if you expect the player to
just
> follow these prompts? If the player is limited to asking about only what
you
> goad him into asking about, it's a false freedom.

But that's exactly the point, isn't it? The game cannot reasonably or
possibly give the player real freedom. As soon as you apply any sort of
story to your Sim-City simulation, you're forcing the plot, the story, and
caging up the player. The trick - in my view - is not to give the player as
much possible freedom as you can but rather to fool the player into
*thinking* that you have. Or rather, into *feeling* that you had. That's
what mimesis is, after all; it's not the game doing everything you want it
to, it's the game feeling to you as though you're in it.

The best moments of mimesis come from when you type something you don't
really expect to work because it's the first thing that pops into your head,
and it *does*. You don't need to achieve these by covering every single
permutation; what you need to do is subtly lead the player but not have him
realise that's what you're doing. [See the beginning of 9:05, whic very
neatly forces you into a certain belief.]

It's not that the player should only be able to ask about what you're
goading him to, it's that you should lead the conversation so that the
natural thing to do is ask exactly what you want him to. And in an
in-your-face list-all-the-topics interface, there is no opportunity for
subtlety.

Jon


Robin Munn

unread,
Nov 19, 2001, 3:29:52 PM11/19/01
to

Instead of replacing menu choices, maybe thinking should add to them?
That way you'd always have the choice to go with your first "gut
response" after all, as if you'd thought about another response but
rejected it. Thus:

Sue: Mary just told me that she's getting married to David
1. Wow, I didn't even know they were dating.
2. Really? He deserves better than her.
> Think about mary's wedding
Slowly, Sue's news about Mary's impending wedding to David sinks in.
While you never liked Mary, you come to realize that she might indeed be the
perfect mate to David - after all, they not only share the same interests,
but he always wanted a woman who could mother him - and after all, that's
exactly what annoys you most about Mary.
1. Wow, I didn't even know they were dating.
2. Really? He deserves better than her.

3. You know what? I think that match just might work.

--
Robin Munn
rm...@pobox.com

OKB -- not okblacke

unread,
Nov 19, 2001, 4:36:22 PM11/19/01
to
"Jon Ingold" ji...@cam.ac.uk wrote:
>It's not that the player should only be able to ask about what you're
>goading him to, it's that you should lead the conversation so that the
>natural thing to do is ask exactly what you want him to. And in an
>in-your-face list-all-the-topics interface, there is no opportunity for
>subtlety.

Okay, the first part of that makes sense, but I don't agree that menus
lock out subtlety. It's just as easy to fill a menu with small-talk as it is
to provide response to ASK ABOUT WEATHER. However, as you've said, we're now
getting into "menus and ASK/TELL under Thunderdome" territory, where opinions
are pretty much all we have left.

Daniel Barkalow

unread,
Nov 19, 2001, 6:07:44 PM11/19/01
to

I'd find that hard to play, especially since I'm not the PC, and don't
have the same information available quickly.

> Anyway, I guess that the objection I see with the 1+ scheme is that it
> adds complexity (in the number of possibilities I'd have to code for;
> in the number of options the player would have to take into account),
> without necessarily adding that much of an advantage. The problem
> here is that players are finding it too easy to play through the game
> on autopilot. So the trick is how to >force< them to make choices
> more often (preferably ones involving thoughtful interaction, not the
> selection of something obvious off the menu), not how to >allow< them
> to choose more.

Well, if the NPC reacted to never adding anything as a normal person would
if you just answered the questions you were asked and didn't say anything
more, the player would have to really choose things more often. I.e., each
turn you'd have to chose whether or not to respond to the NPC, and whether
to say something novel for the NPC to respond to.

I think the only added complexity would be that the NPC would have to
decide which thing to respond to, and you'd have to break the responding
code into one part that deals with the player having given some
information, and one part that comes up with something to say next.

You could also restrict the set of things that can be said as a half-turn:

"1. Yeah..."
"2. No..."
"3. At least he's not a Lovecraftian horror."
"4. Yeah, it's the third time he's skipped school this month"

So the player would have to say something simple and straightforward in
order to change the subject; the other options would count as a full
turn. This would make it easy to decide which thing the PC says should be
responded to on the next turn, essentially just giving the player a way of
changing the subject without ignoring a question or waiting for the NPC to
fail to say anything.

The player would have to balance taking only single turns (which makes the
NPC think you don't actually want to have the conversation, but are just
being polite), taking double turns on the same topic (hard, unless the PC
knows a lot about the topic, doesn't direct the conversation), taking
double turns on different topics (the NPC may try to steer the
conversation back), and just changing the topic without responding to the
NPC (the NPC will probably think you're being evasive). Of course, the PC
can say nothing, as a conversational device, as well.

> [BrenBarn and Jon Ingold seem to have argued eloquently for both sides
> of the ease-for-player vs. challenge-for-player debate. Personally
> I'm in Jon's camp, obviously; and if there aren't a satisfying number
> of topics to which the NPC will respond, then I need to write more
> topics. Simple as that. I passionately dislike the idea of giving
> the player a complete list of topics to which the NPC will respond,
> because that reduces the character, again, to a box whose dimensions
> are known, or a machine with a really large number of buttons that can
> be pressed in any order. And the IF world has plenty of those
> already.

I'd go for a middle ground of a list of all the topics the PC would know
the NPC would respond to; stuff gets added to this list if the NPC shows
some knowledge on the subject, and the player can guess further topics not
directly known (but probably hinted). I think topics should be like
objects: you have a list of ones you've found, there are some made
explicitly available to you, and there are some that have to be dug up.

> A lot of the time I spent on Bo3 was originally intended for writing
> and expanding the dialogue, but wound up hijacked for working on the
> library itself. As a result, there isn't as great a diversity of
> content as I'd hoped, which doubtless contributed to many of
> BrenBarn's objections.]

Yeah, that would make sense. The game seems like it should be much less
redundant, based on the abilities it demonstrates having.

ems...@mindspring.com

unread,
Nov 19, 2001, 11:40:00 PM11/19/01
to
Daniel Barkalow <iabe...@iabervon.org> wrote in message news:<Pine.LNX.4.21.011119...@iabervon.org>...

> On 17 Nov 2001, ems...@mindspring.com wrote:
>
> > Daniel Barkalow <iabe...@iabervon.org> wrote in message news:<Pine.LNX.4.21.011116...@iabervon.org>...

<real-time conversation game>


> I'd find that hard to play, especially since I'm not the PC, and don't
> have the same information available quickly.

Like I said, it's not a *good* idea. (Though if I did implement it it
would be as a raw menu system: there'd be no room for topic-switching,
and no prompt would ever appear on the main game screen. Just a
morphing menu in the menu window. Technically a CYOA, you see, but a
CYOA mediated by the additional factor of time.)



> Well, if the NPC reacted to never adding anything as a normal person would
> if you just answered the questions you were asked and didn't say anything
> more, the player would have to really choose things more often.

Erm. I don't think I'm making it clear enough what I mean. The
player of Bo3 is choosing things all the time-- most frequently,
numbers off the menu. What I want to stimulate is more independent
planning and decision-making and figuring-out. It should feel as
though the player is not being simply led by the nose; and I'm not
sure the insertion of extra decision points would really change that.
As long as the player can cop out and fall back on the game author to
provide the next step of the game, he probably will. And why not?

<stuff about keeping or ceding the conversational initiative>


> The player would have to balance taking only single turns (which makes the
> NPC think you don't actually want to have the conversation, but are just
> being polite), taking double turns on the same topic (hard, unless the PC
> knows a lot about the topic, doesn't direct the conversation), taking
> double turns on different topics (the NPC may try to steer the
> conversation back), and just changing the topic without responding to the
> NPC (the NPC will probably think you're being evasive). Of course, the PC
> can say nothing, as a conversational device, as well.

Hmm. This as proposed seems too mechanical and too predictable to me
(act this way and the NPC will do x...).

Regardless of its virtues, in any case, I think I'll decline to take
this suggestion: the conversation system as it now stands is
approximately as complex as I can deal with, and the inserting
half-turns (with all the extra writing attendant on that) would lead,
not to a more compelling game, but to no game at all, since my brain
would explode first.

You are of course welcome to implement it yourself.



> I'd go for a middle ground of a list of all the topics the PC would know
> the NPC would respond to; stuff gets added to this list if the NPC shows
> some knowledge on the subject, and the player can guess further topics not
> directly known (but probably hinted). I think topics should be like
> objects: you have a list of ones you've found, there are some made
> explicitly available to you, and there are some that have to be dug up.

I still dislike this. Really. I want the conversation to flow from
what the PC wants to ask about; if that's not working, then I need to
a) improve the hints I give about what would be a good topic to follow
up on and b) Make More Topics.

Now, it seems as though a number of other people *do* want it, and I
could certainly implement it. But it will be grudgingly, because I
consider it a total immersion-breaker to thumb through a list of
possible topics: this represents nothing realistic about real life
dialogues, and I consider any use of such a system to be a failure, on
my part, as a designer, to provide alternative and more natural
guidance.



> > A lot of the time I spent on Bo3 was originally intended for writing
> > and expanding the dialogue, but wound up hijacked for working on the
> > library itself. As a result, there isn't as great a diversity of
> > content as I'd hoped, which doubtless contributed to many of
> > BrenBarn's objections.]
>
> Yeah, that would make sense. The game seems like it should be much less
> redundant, based on the abilities it demonstrates having.

There's also another design decision here: the game *could* have been
much shorter on a single playthrough (and thus perhaps less dull [to
those who found it so], and more diverse on replay), except that with
Galatea people complained that they wished they'd been able to see
more of the game. So I deliberately set it up so that certain
baseline topics would always get at least alluded to.

Inasmuch as the conversation game is a genre in itself (and Adam and
others have questioned whether it is, or has a right to be), I think
this will always be a problem: huge amounts of work have to go into
producing something that comes even close to being interesting. As a
ballpark figure, Bo3 implements about 150 topics (I could quote the
actual number, but it would not be useful because some things are
implemented as topics that are actually sort of special cases within
the conversation). There are 422 statements (for the PC) available;
Grant is capable of asking 63 questions and making 174 comments in
addition (ie, that may not have been prompted by any of the
aforementioned 422 statements.) Galatea, I think, has somewhere
around 200 things to say, which means that Grant's conversation
represents over three times as much raw content-- but the topic/menu
system is inherently more difficult to write for than raw ask/tell
simply because you have to produce for every keyword you implement not
one but half-a-dozen or more remarks, some simultaneous, some in
sequence.

The consistent message I'm getting is that this was not enough-- not
enough topics, not enough dialogue in the topics. To fill this game
out to its full potential is probably a project that would take
another several months of work. Shortening the effective length of
the game (by forcing the denouement to occur sooner) might add to the
appearance of diversity by excluding more of the content on a single
play-through: but at the core of the problem is that, given enough
time and a free range, a human will find the edges of any NPC.

Whether it's even worth trying is another issue, of course. This is
not a genre that has much precedent in written material; one might
speak of mystery IF or romance IF or whatever, but Conversation IF is
heir to no literary tradition that I know of. Performative
traditions, maybe: "My Dinner With Andre" kept coming to mind when I
was writing, though I didn't have Wally Shawn's quirkiness at my
disposal or half the bizarre humor of the original. And I still have
this fantasy of using the medium of IF to gratify another long-held
personal desire -- to recast a Platonic dialogue in such a form that
for *once* Socrates' interlocutor doesn't have to *agree* all the
time.

Will the conversation system work in the setting of a "Real Game," as
people keep asking? (The phrasing makes my teeth grit, but ok, I
understand what the question is.) I think so. We'll see, I suppose,
eventually.

At this rate it should only take me another 10 or 15 years to develop
the tools to write the game I wanted to write to start with. And by
then all the characters will have grown old or moved away.

-- Emily

Sean T Barrett

unread,
Nov 20, 2001, 12:57:08 AM11/20/01
to
ems...@mindspring.com <ems...@mindspring.com> wrote:
>I still dislike this. Really. I want the conversation to flow from
>what the PC wants to ask about; if that's not working, then I need to
>a) improve the hints I give about what would be a good topic to follow
>up on and b) Make More Topics.

I think at a certain point you have to decide that a
game design can be a bad game design if it simply has
prohibitive content development costs.

>At this rate it should only take me another 10 or 15 years to develop
>the tools to write the game I wanted to write to start with. And by
>then all the characters will have grown old or moved away.

But I guess you're at least aware of those costs.

SeanB

Eytan Zweig

unread,
Nov 20, 2001, 3:09:43 AM11/20/01
to
>>> What would a true "think about" feature do?
>>>
>>
>>Well, what I'd like to see in a "think about" system is that
thinking about
>>things should (at least sometimes) change the way you feel about
them - for
>>instance, say that you have an NPC conversation game, where the PC
is having
>>a disussion with his friend Sue
>>
>>Sue: Mary just told me that she's getting married to David
>>1. Wow, I didn't even know they were dating.
>>2. Really? He deserves better than her.
>>> Think about mary's wedding
>>Slowly, Sue's news about Mary's impending wedding to David sinks in.
While
>>you never liked Mary, you come to realize that she might indeed be
the
>>perfect mate to David - after all, they not only share the same
interests,
>>but he always wanted a woman who could mother him - and after all,
that's
>>exactly what annoys you most about Mary.
>>1. Wow, I didn't even know they were dating.
>>2. You know what? I think that match just might work.
>>
>
>Instead of replacing menu choices, maybe thinking should add to them?
>That way you'd always have the choice to go with your first "gut
>response" after all, as if you'd thought about another response but
>rejected it. Thus:
>
>Sue: Mary just told me that she's getting married to David
>1. Wow, I didn't even know they were dating.
>2. Really? He deserves better than her.
>> Think about mary's wedding
>Slowly, Sue's news about Mary's impending wedding to David sinks in.
>While you never liked Mary, you come to realize that she might indeed
be the
>perfect mate to David - after all, they not only share the same
interests,
>but he always wanted a woman who could mother him - and after all,
that's
>exactly what annoys you most about Mary.
>1. Wow, I didn't even know they were dating.
>2. Really? He deserves better than her.
>3. You know what? I think that match just might work.
>

That was actually what I wrote initially, but then I decided to use an
example the replaces the menu option. There are two reasons for this -

- One is, that in real life, after we think about something, we might
be less likely to go with our "gut reaction".

- More importantly, from a gameplay consideration, if there are no
penalties for thinking about something everyone will always "think
about" everything, which will just be tedious.

Now, admittedly my example above is less than perfect in demonstrating
the need for this, but it's just an example. In a real game, I'd like
to see situations where thinking about something opens new
possibilities without closing old ones, but also ones where sometimes
you can only say something if you don't take the time to think; and
maybe even places where thinking just closes options and doesn't open
new ones. Then the game could really be different if the PC is played
as a cerebral character who always contemplates everything seriously,
a brash person who never stops to think, or someone in between.

Eytan


>--
>Robin Munn
>rm...@pobox.com

Billy Harris

unread,
Nov 20, 2001, 5:51:17 AM11/20/01
to
In article <a69830de.01111...@posting.google.com>,
<ems...@mindspring.com> wrote:

> Erm. I don't think I'm making it clear enough what I mean. The
> player of Bo3 is choosing things all the time-- most frequently,
> numbers off the menu. What I want to stimulate is more independent
> planning and decision-making and figuring-out. It should feel as
> though the player is not being simply led by the nose; and I'm not
> sure the insertion of extra decision points would really change that.
> As long as the player can cop out and fall back on the game author to
> provide the next step of the game, he probably will. And why not?

I think in lieu of a topic list, better [initial] characterization
would help. I agree with one of the reviews who said that the PC was
generic enough that you were encouraged to identify directly with her,
and then it was jarring to find out she had different opinions than
you. My second time through, I had a much better handle on the PC's
personality and background, and made much more frequent use of the
TOPIC command... I was impressed that 90%+ of the topics I tried were
recognized, and dissapointed that I had very often already exhausted
the topics.

Perhaps as a supplement to THINK ABOUT, the BACKGROUND command could
give a couple of paragraphs of biography about the PC [and NPC?] to
better allow the player to fit into the role.

My other complaint can be fairly translated as "I don't enjoy this
genre"; being goal directed, I wasn't clear what the thrust of the game
was. So I took a surragate role of seducing Grant [since the PC is so
passive, this became almost "Encouraging Grant to seduce me"]. The
first time I played, I did not succeed. The second time I did [at least
to the point of being asked out]. I have no idea what made the
difference between the two games.

--
Billy Harris
wha...@mail.airmail.net

Jon Ingold

unread,
Nov 20, 2001, 11:53:20 AM11/20/01
to
> The consistent message I'm getting is that this was not enough-- not
> enough topics, not enough dialogue in the topics. To fill this game
> out to its full potential is probably a project that would take
> another several months of work. Shortening the effective length of
> the game (by forcing the denouement to occur sooner) might add to the
> appearance of diversity by excluding more of the content on a single
> play-through: but at the core of the problem is that, given enough
> time and a free range, a human will find the edges of any NPC.

I wonder if the sense of unease player's get from reaching a boundary of an
NPC would be diminshed if the game was given a driving-plot as well. To
clarify:- non-linear railroaded story games, when done well, do not feel too
limiting because the player is constantly moving and never gets a feel for
the limits of his location. (See BAP, I think). Here, the player is dragged
along, interested, and so *not looking* for those boundaries. Maybe
something like this could be applied.

So, for example, if the dialogue was not a chat but an interrogation, an
attempt to elicit contradictory facts from a liar with which to trap him, it
may also trap the player into thinking in character, moving in character,
and limit the amount of freedom desired, all in one go.

Of course -- this is a different type of game. But it could use the same
technology (which is powerful, powerful, powerful).

Jon


SteveG

unread,
Nov 20, 2001, 2:13:31 PM11/20/01
to

You'd be needing "Glulxe" ("the Glulx Engine?"), not so much because
you need Glulxe to play blorb games but because this particular blorb
file contains a Glulx game.

Glulxe for various computer systems can be found in the IF Archive's
Glulx interpreters directory:

http://www.ifarchive.org/indexes/if-archiveXprogrammingXglulxXinterpretersXglulxe.html


--
SteveG
(Please remove erroneous word from address if emailing a reply)

OKB -- not okblacke

unread,
Nov 20, 2001, 4:14:07 PM11/20/01
to
"Jon Ingold" ji...@cam.ac.uk wrote:
>So, for example, if the dialogue was not a chat but an interrogation, an
>attempt to elicit contradictory facts from a liar with which to trap him, it
>may also trap the player into thinking in character, moving in character,
>and limit the amount of freedom desired, all in one go.
>
>Of course -- this is a different type of game. But it could use the same
>technology (which is powerful, powerful, powerful).

Yes. I think the reason this hasn't happened yet is that, as Emily said,
we're still working on perfecting the technology. Few people are willing to
spend untold hours writing a conversation model AND an interface to go with
that model AND a whole bunch of dialogue AND a character with a personality,
and then write a full-on story to boot, without checking themselves along the
way by releasing smaller games. (I had a similar situation with Stick it to
the man, where the vast majority of the source code is NPC stuff, although
there it is spread out over many NPCs instead of being concentrated in one.)
It's not hard to imagine that, after writing all that NPC-related code, the
author will want people to pay attention to it, and will thus be less likely to
bury it among other game aspects.

There's been a lot of focus on NPC development lately, though, so with any
luck we'll soon be seeing high-quality NPCs as just a part of high-quality
games. (The costs in file size and slowness are going to be significant,
though.) I certainly look forward to that!

M. D. Krauss

unread,
Nov 20, 2001, 7:28:49 PM11/20/01
to
On Tue, 20 Nov 2001 04:51:17 -0600
Billy Harris <wha...@mail.airmail.net> wrote:

<<<SNIP>>>

> My other complaint can be fairly translated as "I don't enjoy this
> genre"; being goal directed, I wasn't clear what the thrust of the game
> was. So I took a surragate role of seducing Grant [since the PC is so
> passive, this became almost "Encouraging Grant to seduce me"]. The
> first time I played, I did not succeed. The second time I did [at least
> to the point of being asked out]. I have no idea what made the
> difference between the two games.


Woah... I played three times and always got asked out. Didn't know there
could be a different ending! Guess I just make a naturally seductive
woman... Heh.

On a more important point, though, I'd say that not knowing what made the
difference is a sign of success on the authors part. Human interactions
are certainly strange and mysterious at times, and it would be
dissapointing if something like that were obvious.

-Matthew

Daniel Barkalow

unread,
Nov 21, 2001, 3:10:35 PM11/21/01
to
On 19 Nov 2001, ems...@mindspring.com wrote:

> Daniel Barkalow <iabe...@iabervon.org> wrote in message news:<Pine.LNX.4.21.011119...@iabervon.org>...


>
> > Well, if the NPC reacted to never adding anything as a normal person would
> > if you just answered the questions you were asked and didn't say anything
> > more, the player would have to really choose things more often.
>
> Erm. I don't think I'm making it clear enough what I mean. The
> player of Bo3 is choosing things all the time-- most frequently,
> numbers off the menu. What I want to stimulate is more independent
> planning and decision-making and figuring-out. It should feel as
> though the player is not being simply led by the nose; and I'm not
> sure the insertion of extra decision points would really change that.
> As long as the player can cop out and fall back on the game author to
> provide the next step of the game, he probably will. And why not?

Well, the problem I had as a player was that it seemed like the only
polite thing to do would be to respond to Grant when he said anything that
sounded like it was intended to provoke a response; that meant that I
could only use TOPIC when he didn't say anything, just about. This
wouldn't exactly happen if he were less proactive, since the player would
more frequently have to come up with something to say, but it still
wouldn't be at the player's option. Of course, with the actual
implementation, I presume you can simply change the subject when you want
and Grant doesn't get offended, but that didn't seem in character. As I
saw the character, she'd sooner fail to change the topic than leave most
of the things Grant says unanswered, although she'd more frequently say
something shallow to brush off the question, and then change the topic.

> <stuff about keeping or ceding the conversational initiative>
> > The player would have to balance taking only single turns (which makes the
> > NPC think you don't actually want to have the conversation, but are just
> > being polite), taking double turns on the same topic (hard, unless the PC
> > knows a lot about the topic, doesn't direct the conversation), taking
> > double turns on different topics (the NPC may try to steer the
> > conversation back), and just changing the topic without responding to the
> > NPC (the NPC will probably think you're being evasive). Of course, the PC
> > can say nothing, as a conversational device, as well.
>
> Hmm. This as proposed seems too mechanical and too predictable to me
> (act this way and the NPC will do x...).
>
> Regardless of its virtues, in any case, I think I'll decline to take
> this suggestion: the conversation system as it now stands is
> approximately as complex as I can deal with, and the inserting
> half-turns (with all the extra writing attendant on that) would lead,
> not to a more compelling game, but to no game at all, since my brain
> would explode first.

How about just having things you can say that don't invite a response, so
that the player can take control? That shouldn't really increase the
complexity, as it only adds options that generate simple responses, and
makes it less important to tune the proactivity of the NPCs just right.

> You are of course welcome to implement it yourself.

I'd try it, since I obviously don't have any real idea what's
feasible. Unfortunately, I'm terrible at writing dialogue.

> > I'd go for a middle ground of a list of all the topics the PC would know
> > the NPC would respond to; stuff gets added to this list if the NPC shows
> > some knowledge on the subject, and the player can guess further topics not
> > directly known (but probably hinted). I think topics should be like
> > objects: you have a list of ones you've found, there are some made
> > explicitly available to you, and there are some that have to be dug up.
>
> I still dislike this. Really. I want the conversation to flow from
> what the PC wants to ask about; if that's not working, then I need to
> a) improve the hints I give about what would be a good topic to follow
> up on and b) Make More Topics.
>
> Now, it seems as though a number of other people *do* want it, and I
> could certainly implement it. But it will be grudgingly, because I
> consider it a total immersion-breaker to thumb through a list of
> possible topics: this represents nothing realistic about real life
> dialogues, and I consider any use of such a system to be a failure, on
> my part, as a designer, to provide alternative and more natural
> guidance.

I find it necessary as a way of dealing with the separation between the
player and the PC. The PC wouldn't have to take notes as to what has been
discussed recently, but, if the player wants to jump back to something
that was mentioned before, it's pretty difficult. Not for lack of hinting,
but because the hinting was earlier in the conversation as well. I think
the unnatural thing is how TOPIC has to work: the game and player need to
have established a convention by which the player can refer to a topic,
and, with no commnds other than TOPIC, there's no way to get the names or
revisit them. At a minimum, I'd need to know, when I have a menu choice,
what topic the game considers it to be on, so that, if I want to say it
later, I can. Now, a list of topics which have been hinted but not used is
unnecessary, but I find it makes navigation difficult if I can't tell
where I am or get back to somewhere I've been. A list of all of the topics
which have been active but have not been exhausted would make sense both
for playability and as a list the PC could actually be thinking through.

> > > A lot of the time I spent on Bo3 was originally intended for writing
> > > and expanding the dialogue, but wound up hijacked for working on the
> > > library itself. As a result, there isn't as great a diversity of
> > > content as I'd hoped, which doubtless contributed to many of
> > > BrenBarn's objections.]
> >
> > Yeah, that would make sense. The game seems like it should be much less
> > redundant, based on the abilities it demonstrates having.
>
> There's also another design decision here: the game *could* have been
> much shorter on a single playthrough (and thus perhaps less dull [to
> those who found it so], and more diverse on replay), except that with
> Galatea people complained that they wished they'd been able to see
> more of the game. So I deliberately set it up so that certain
> baseline topics would always get at least alluded to.

I think the problem was that the conversation tended to actually go
through all of the topics, not just touch on them enough to make them
logical TOPIC arguments.

I was actually looking for an implementation at a different level; since
the characters are making small-talk, the important thing is not what they
say, exactly, but the way they're interacting. I'd expect the plot to
center more around what happens if you always agree with Grant, what
happens if you keep changing the subject, what happens if keep insulting
his viewpoints, what happens if you keep saying things he finds clever,
etc. It seemed like the extent of the game wasn't logically the set of
topics they would discuss. Of course, this wouldn't be true of a game
where the agenda is to exchange explicit information, which seemed to me
to be true in Galatea, where you actually do care about her story.

> Whether it's even worth trying is another issue, of course. This is
> not a genre that has much precedent in written material; one might
> speak of mystery IF or romance IF or whatever, but Conversation IF is
> heir to no literary tradition that I know of. Performative
> traditions, maybe: "My Dinner With Andre" kept coming to mind when I
> was writing, though I didn't have Wally Shawn's quirkiness at my
> disposal or half the bizarre humor of the original. And I still have
> this fantasy of using the medium of IF to gratify another long-held
> personal desire -- to recast a Platonic dialogue in such a form that
> for *once* Socrates' interlocutor doesn't have to *agree* all the
> time.
>
> Will the conversation system work in the setting of a "Real Game," as
> people keep asking? (The phrasing makes my teeth grit, but ok, I
> understand what the question is.) I think so. We'll see, I suppose,
> eventually.

I suspect it will work better, if anything, because the author would have
the simpler and better understood methods of motivating the
interaction. Not to mention avoiding the thorny issue of small-talk.

Stephen Granade

unread,
Nov 22, 2001, 12:30:30 PM11/22/01
to
buz...@TheWorld.com (Sean T Barrett) writes:

> ems...@mindspring.com <ems...@mindspring.com> wrote:
> >Simple as that. I passionately dislike the idea of giving
> >the player a complete list of topics to which the NPC will respond,
> >because that reduces the character, again, to a box whose dimensions
> >are known, or a machine with a really large number of buttons that can
> >be pressed in any order. And the IF world has plenty of those
> >already.
>
> I think arguably if you nest menus deep enough you can
> regain that level of complexity; providing a list of topics
> to choose from the size of a dictionary doesn't really
> make the game any more trivial. But this requires having
> an awful lot of topics available.

I will chime in with my favorite example of reasonable complexity
through a giant mound of conversation choices: Planescape: Torment. It
uses a menued conversation system, as have countless prior games which
fall into the large cardboard box marked "computer RPG." However, it
gives you *so many choices* over the course of the game that trying to
lawnmower your way through them all is well-nigh impossible. Tied into
that is how some conversational choices close off other avenues of
conversation, and do so in mostly logical ways.

I say this as an example of a system in which, sure, there are a
really large number of buttons which you can press, often in a
locally-arbitrary order. But by the sheer number of said buttons and
judicious application of state-sensitive conversations, the designers
created a conversation system which was astounding.

Stephen

--
Stephen Granade
sgra...@phy.duke.edu
Duke University, Physics Dept

L. Ross Raszewski

unread,
Nov 23, 2001, 3:38:24 AM11/23/01
to
On 22 Nov 2001 12:30:30 -0500, Stephen Granade <sgra...@phy.duke.edu> wrote:
>I will chime in with my favorite example of reasonable complexity
>through a giant mound of conversation choices: Planescape: Torment. It
>uses a menued conversation system, as have countless prior games which
>fall into the large cardboard box marked "computer RPG." However, it
>gives you *so many choices* over the course of the game that trying to
>lawnmower your way through them all is well-nigh impossible. Tied into
>that is how some conversational choices close off other avenues of
>conversation, and do so in mostly logical ways.
>

Possibly my favorite CRPG/Adventureish, 'Protostar', took a similar
tack, presentign you with a conversation menu listing, well,
everything. Of course, other things might be added during the course
of the game, and certain questions could end the conversation (by
boring/annoying/reminding-of-an-important-engagement-elsewhere the
conversee) or trigger a plot-event.

ANd, since I'm in the thread, I'll randomly plug Converse.h, which was
really designed for topic-driven menued conversation (as opposed to
the phototalk-derived libraries, which appear to be oriented more
toward statement-driven menued conversation). (Of course, Converse
will only work under glulx in non-"graphical" mode, but current trends
seem to prefer this style of conversation menu anyway. If there's
demand, I'll see about porting it to GWindows. Though withthe way
GWindows's menuing system is layed out, there might not be any real
ease-of-use benefit in using a library over just writing your own
menu/map code.)

Reply all
Reply to author
Forward
0 new messages