Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

Emergence

10 views
Skip to first unread message

Malcolm Ryan

unread,
Mar 28, 2006, 1:04:40 AM3/28/06
to
I'm currently studying game design, and one of the keep concepts is
that of emgergence. Many of the most enduring games have this property:
the game world is a complex system with emergent properties. Chess, for
example, has many aspects of strategy that are not expressly described
in the rules. Rather, they emerge from the interaction of many simpler
parts. This is even more evident in the game of Go. There are patterns
of pieces which have particular meaning as larger strategic elements in
the game.

On the other hand, interactive fiction games are entirely scripted.
They is no room for emergent effects, every outcome has been
deliberately built by the author. It seems to me that this is because
of the difficulties of language. It is easy to display a complex system
graphically so that the patterns are visible to the player. The
computer doesn't need to 'understand' the patterns in order to
represent them. This is not true of language. If I was to write a
complex physical system and then describe its output verbally, I would
have to be able to recognise the patterns in order to be able to
present them to the reader.

This is a tricky problem and puts IF at a serious disadvantage. Does
anyone have any insight into how we make progress with it?

Malcolm

Markus Zywitza

unread,
Mar 28, 2006, 4:53:48 AM3/28/06
to
So why do you think that this is an disadvantage for IF?

You're mixing up two completely different things here. Emergence is
needed for games that have easy rules but are nonetheless fairly
complex and interesting. You now have two comparison options:

You can either
- compare playing a piece of IF to reading the rules of chess
- regard the usage of simple english sentences as the rules of IF and
the game itself as the emergence of IF.

And even then, comparing chess and IF is far away from making any
sense...

Markus

Mark Thern

unread,
Mar 28, 2006, 4:56:45 AM3/28/06
to
Chris Crawford has been trying to create a system for telling
procedurally generated stories for many years now. These are the most
basic features of his approach:

1. The player's options are limited to what is dramatically appropriate
for a given context.

2. The personality of the other characters are defined by numerical
variables such as Affection, Greed, Honesty, etc.

3. These personality attributes are used in inclination equations to
determine how the characters react to what's currently going on in the
story, i.e. If Brad.Emotional_State == Sad and Marsha.Affection.Brad >
20, Then Marsha.Cheer_Up.Brad.

4. The events in the story are not crafted by hand. Instead, they're
formed from 3 basic elements: a subject, a verb, and a direct object.
Crawford's system places a lot of emphasis on creating an extremely
large number of reusable verbs.

5. Crawford's system supports procedurally generated lying and
gossiping.

6. Crawford system tries to make the player's interactions with the
characters more immersive by giving each character a face that can
display a wide range of expressions.

The first iteration of Crawford's storytelling software, the
Erasmatron, was a spectacular failure. "Shattertown Sky", the game that
he commissioned to showcase his software, was universally panned and
derided. I've never played the game, as it's no longer available for
download anywhere, but its reviews and the transcripts of its gameplay
seem to fully justify its critical reception. Some of this can be
blamed on the writer, as Shattertown Sky has a bizarre premise and an
equally bizarre social environment to explore. However, I think a lot
of it had to do with insufficient pattern recognition, i.e. the story
didn't unfold in a sensible way, and the lack of narrative data, which
meant that the characters had a limited number of responses that the
player would see over and over again. The downright ugly face
technology didn't help either.

Crawford's newest effort, the Storytron, allows for a wider (though
still dramatically limited) number of player options. It also tries to
solve the limited response problem through a more abstract way of
telling the story. If Reginald learns about his wife's infidelities and
decides to swear, the player is not going to be presented with a string
of epithets. Instead he's going to see "Reginald swears." Likewise, if
Great-Aunt Majorie drops the sugar bowl, the output text will be "Great
Aunt Majorie swears" rather than "Oh fiddlesticks!". The face
technology is much, much better as well.

If you'd like to learn more about the Storytron, you can go to
www.storytron.com. If you'd like to see a working example of the
Storytron's abstract storytelling in action, you can go to Crawford's
site, www.erasmatazz.com and download Trust and Betrayal, a social
simulation game that Crawford made around 1987 or so that uses the same
approach.

My take on the Storytron? This isn't going to work either. My
definition of a successful story is one whose characters provoke a
strong emotional reaction in its readers. I've played Trust and
Betrayal, and it never managed to convince me that I was interacting
with anything more than a carefully built set of algorithms. All
storytelling mediums have different strengths and weaknesses in what
they can express, but the Storytron approach is too stripped down to be
effective.

My take on the concept of emergent stories? I would certainly like to
see i-f works with stories that could unfold in wildly divergent ways,
but I don't know if I agree with the idea that a work of i-f should
strive to be infinitely replayable. I think that's a good goal for
games, but not for literature. I would rather read 20 stories that are
told well than a zillion different stories of varying quality.

It would be interesting to see what would happen if an author (or a
team of authors) decided to create an interactive story with so much
branching content --1000+ pages of it, let's say-- that it managed to
approach the ideal of an emergent story. Incidentally, this is what I
think commercially successful i-f would look like.

steve....@gmail.com

unread,
Mar 28, 2006, 8:14:58 AM3/28/06
to
Malcolm Ryan wrote:
> [I]nteractive fiction games are entirely scripted.

> They is no room for emergent effects, every outcome has been
> deliberately built by the author.

I know what you mean, but I think it's useful to challenge the terms of
the argument. Consider this: the game-world simulation produces
emergent behavior by design. You can put stuff in spacial
configurations unanticipated by the writer. It's even possible to solve
puzzles in a way unintended by the writer. (Whether this is another
emergent behavior or a bug is moot.) IF games are not entirely
scripted, indeed most of the stuff is not scripted at all. (Compare the
number of default responses you get, to the number of custom
responses.) What bothers me is that the interesting behavior is almost
always the scripted, or exceptional behavior.

The way to more interesting emergent behavior is stronger AI.

PJ

unread,
Mar 28, 2006, 9:39:52 AM3/28/06
to

steve....@gmail.com wrote:
> The way to more interesting emergent behavior is stronger AI.

Possibly. I tend to see emergent behavior in games as a by-product of
the need to defeat an adversary that is *actively* trying to defeat
you. I can't think of any IF where the game itself is programmed to
try and defeat you. IF games are mostly designed to be static, if
obscure, in their configurations and in the solutions to their puzzles
-- if you solve the puzzles in their basic configurations, you win.

I'm not sure that you actually need real "AI", however you care to
define it, to make the game more adversarial. You could develop
algorithms that compare the number of moves it takes a user to solve a
puzzle or reach a certain point in the game, then introduce something
-- a thief, an earthquake, whatever -- that changes the game state in
some fashion. Depending on whether the player is doing well or poorly,
these changes could be wildly different in effect. Emergent behavior
might be the strategies the user uncovers to either (a) avoid this
state change (b) accelerate it or (c) do something else with the
knowledge that a state change might happen.

It's also true that most people don't play individual IF games
repeatedly over a period of years, the way you play chess over years.
Since every game has similar but slightly different rules/conventions,
it's far less likely that you will see emergent behavior, because once
solved, most people stop playing the game seriously, except for the
occasional nostalgic return.

I think you would also see emergent behavior if there were more
two-player IF games where either cooperation or competion (or both)
were part of the object of the game. But, in that instance, you move
rapidly towards MUDs and RPGs and away from what most consider IF at
this point.

But even without an AI adversary, IF does have some low-level emergent
behaviors that experienced players have adopted and typically use on
any game, including:

-- repeatedly "taking all" just to make sure that you haven't missed
anything
-- carrying even fairly obscure objects around just in case they are
the answer to a puzzle
-- turning light sources off obsessively just in case you might get
"Zorked" by a failing lantern
-- examining every object in every room description
-- using "save/restore" and "undo" to test every option at obvious
turning points in the game;

While we think of these and other similar practices as just what any
experienced player would do, I would argue that these are the core
"emergent" behaviors for IF in its current state of development. AI
might be the holy grail, as you suggest, but there are probably some
interim things that could be done to make the games more adversarial
along the way.

PJ

Adam Thornton

unread,
Mar 28, 2006, 11:59:30 AM3/28/06
to
In article <1143539805.6...@i40g2000cwc.googlegroups.com>,

Mark Thern <mth...@juno.com> wrote:
>My take on the Storytron? This isn't going to work either. My
>definition of a successful story is one whose characters provoke a
>strong emotional reaction in its readers. I've played Trust and
>Betrayal, and it never managed to convince me that I was interacting
>with anything more than a carefully built set of algorithms. All
>storytelling mediums have different strengths and weaknesses in what
>they can express, but the Storytron approach is too stripped down to be
>effective.

I would suggest that you play the Holodeck scene of _Stiffy Makane: The
Undiscovered Country_ for my criticism of the Erasmatron (and the
Storytron, it sounds like). My fundamental objection is that Aristotle
was *wrong*: *most* interesting story does not emerge dialectically from
characters and the conflict that results from the friction between their
natures.

No amount of refining the technology is going to magically produce
interesting stories.

>It would be interesting to see what would happen if an author (or a
>team of authors) decided to create an interactive story with so much
>branching content --1000+ pages of it, let's say-- that it managed to
>approach the ideal of an emergent story. Incidentally, this is what I
>think commercially successful i-f would look like.

How could this be commercially successful?

1000 pages would be nowhere near enough for a wildly branching story of
any length. For 1000 pages, you'd get a sort-of-branching long piece of
IF; to get the degree of freedom you're imagining *I* think you'd need
about ten times that much.

Now: how could this be commercially successful? Most writers can't make
it work full-time when they're writing novels: everyone except the real
A-list authors either lives pretty hand-to-mouth, or has a day job, or
both. Novels already have a well-understood publishing and distribution
model. For this new form, you increase the amount of text required by
between one and two orders of magnitude, add the additional constraint
of "being able to program" to the author's job, and then present it as a
form for which there's no obvious current publication or distribution
model (not a book, and you probably don't want to market it as
software).

Yeah, you probably *could* sell more of them than you can of current IF;
but if it's unable to provide its authors with a living wage, I don't
think you can treat it as "commercially successful."

Adam


signwriter

unread,
Mar 28, 2006, 12:03:21 PM3/28/06
to
Malcolm Ryan wrote:
> Many of the most enduring games have this property:
> the game world is a complex system with emergent properties.

> On the other hand, interactive fiction games are entirely scripted.


> They is no room for emergent effects, every outcome has been
> deliberately built by the author.

There's a perpetually recurring debate about what's important in a work
of interactive fiction. At its crudest, it boils down to "game" vs.
story. No offense, but many of us don't approach a work of IF as a game
qua game, in the sense that chess is; it's an experience, in the same
sense that a work of fiction is - a novel, say, or a film.

> It seems to me that this is because
> of the difficulties of language. It is easy to display a complex system
> graphically so that the patterns are visible to the player. The
> computer doesn't need to 'understand' the patterns in order to
> represent them. This is not true of language. If I was to write a
> complex physical system and then describe its output verbally, I would
> have to be able to recognise the patterns in order to be able to
> present them to the reader.

Undoubtedly true...

> This is a tricky problem and puts IF at a serious disadvantage. Does
> anyone have any insight into how we make progress with it?

...but I have trouble with the conclusions you draw. The tools you're
talking about are good for producing *certain types of works*, and
getting better. I regard those works as consituting a valid medium for
creatives to work in, and for players/readers to enjoy. I see no
"significant disadvantage" - and I think you perceive one because you're
trying to find an entirely different kind of finished product. So yes,
*necessarily*, you need different tools to create them.

I'm sure you don't intend to be dismissive of the existing body of IF,
and I don't intend to be dismissive of the sort of thing you want to
see, but I think the underlying differences between them are more than
superficial. Comparisons may be of limited (no?) utility.

I think, ultimately, your "emergent simulation IF" has to be considered
on its own, not as the evolutionary next step from a cruder technology -
and I think I'd be right in saying that for this newsgroup to accept it,
it will have to take the "fiction" in "interactive fiction" extremely
seriously, not disregard it to focus on games qua games.

Having said all that, I *am* interested; had you come up with any
concrete ideas so far?

Mark Thern

unread,
Mar 28, 2006, 1:25:24 PM3/28/06
to
Branching content is only one aspect of what I think commercially
successful i-f would look like. I also think that it would emphasize
people and relationships over physical simulation, and that it would
achieve this end by limiting the player's options to what's
dramatically appropriate for the current context. The form that I have
in mind lies in the middle between CYOA style fiction and parser-based
i-f. Under this form, creating a widely branching narrative becomes a
lot easier. The CYOA stories that I read as a kid went in all sorts of
odd directions, and their total word count was maybe 60-80 pages worth
of material, if that.

steve....@gmail.com

unread,
Mar 28, 2006, 2:08:32 PM3/28/06
to
PJ wrote, quoting me:

> > The way to more interesting emergent behavior is stronger AI.
>
> Possibly. I tend to see emergent behavior in games as a by-product of
> the need to defeat an adversary that is *actively* trying to defeat
> you.

In terms of Drama Management Theory, the game is in fact actively
trying to defeat the player. A typical chess-program type algorithm is
set up with a slightly different heuristic for "judgePosition()" -- the
idea being that the game "wins" if the narrative is good, and the
player "wins" if the narrative is crappy. It's a lot like rewriting a
chess program so that it maximizes the chance for a tactically elegant
checkmate by either side.

> I can't think of any IF where the game itself is programmed to
> try and defeat you.

You're right: Drama Management has not yet come into its own in IF.

> You could develop
> algorithms that compare the number of moves it takes a user to solve a
> puzzle or reach a certain point in the game, then introduce something
> -- a thief, an earthquake, whatever -- that changes the game state in
> some fashion. Depending on whether the player is doing well or poorly,
> these changes could be wildly different in effect. Emergent behavior
> might be the strategies the user uncovers to either (a) avoid this
> state change (b) accelerate it or (c) do something else with the
> knowledge that a state change might happen.

Yes, this is all part of Drama Management (DM), although the player
does not necessarily have control over the plot points or their
consequences.

> It's also true that most people don't play individual IF games
> repeatedly over a period of years, the way you play chess over years.
> Since every game has similar but slightly different rules/conventions,
> it's far less likely that you will see emergent behavior, because once

> solved, most people stop playing the game seriously[.]

This reminds me very closely of my main beef with AI/DM. Whenever you
have a work that you have to play through several times to appreciate
properly, I get uneasy. I want the appreciation to be in my experience
of the text, rather than some external extrapolation.

Adam Thornton

unread,
Mar 28, 2006, 3:48:33 PM3/28/06
to
In article <1143570323.9...@i39g2000cwa.googlegroups.com>,

Mark Thern <mth...@juno.com> wrote:
>Branching content is only one aspect of what I think commercially
>successful i-f would look like. I also think that it would emphasize
>people and relationships over physical simulation, and that it would
>achieve this end by limiting the player's options to what's
>dramatically appropriate for the current context.

Uh, in that case, wouldn't a hypertext clicky-tree work better than a
parser?

Adam

Andrew Plotkin

unread,
Mar 28, 2006, 2:41:43 PM3/28/06
to
Here, Malcolm Ryan <malcol...@gmail.com> wrote:
>
> On the other hand, interactive fiction games are entirely scripted.
> They is no room for emergent effects, every outcome has been
> deliberately built by the author. It seems to me that this is because
> of the difficulties of language.

I first read this as "IF development language", but you mean the
human-read language that is used for game input and output, right?
I think this point has some validity, but it's also true that authors
have been able to describe some very complex systems in IF prose.

(Some situations will always be better described with pictures, of
course. But you're always going to have to decide what to put into
your game based on what works in the medium.)

I would also say this: an IF player will always *act* as if the world
is fully simulated. If the game is well-done, it will support him in
this. Some elements really are simulated, rather than scripted.
(Picking up and dropping objects, for example, really aren't
hand-written unless the author wants them to be.) Then there is a
(hopefully) rich space of interesting scripted outcomes. It doesn't
matter that there is no underlying super-physics engine, because the
player is (hopefully) going to try sensible things and get sensible
responses anyway.

In other words, whether a system has "emergent properties" is really a
question for the designer. The player isn't in a position to tell the
difference, if the designer is doing his job.

--Z

--
"And Aholibamah bare Jeush, and Jaalam, and Korah: these were the borogoves..."
*
9/11 did change everything. Since 9/12, the biggest threat to American
society has been the American president. I'd call that a change.

Mark Thern

unread,
Mar 28, 2006, 4:07:57 PM3/28/06
to
It might if you limited your design to presenting no more than 10
options at a given decision point. However, in own my work-in-progress,
I'm trying to simulate realistic conversations. To do that, I give the
player a very limited number of appropriate verbs and a list of
keywords. If the player enters "Ask Brad about Marsha", then a menu of
dialogue options about that subject will pop up.

I initially thought that you could do this entirely through point and
click and even built a working prototype in Tads 3. As I tested my
system though, I discovered a problem: I had too many keywords to fit
in a single menu. I tried a number of classification schemes, but the
number of clicks needed to enter a sentence were often as high as five.

After switching to a new authoring tool called Ren'Py and with the
generous assistance of Ren'Py's author, I now have a parser that will
automatically complete your words as you type them. It won't allow you
to make typing errors either. I use it when I'm testing out my
material, and I think it's the most effective solution.

Poster

unread,
Mar 28, 2006, 6:14:32 PM3/28/06
to
signwriter wrote:
> Malcolm Ryan wrote:
>> Many of the most enduring games have this property:
>> the game world is a complex system with emergent properties.
>
>> On the other hand, interactive fiction games are entirely scripted.
>> They is no room for emergent effects, every outcome has been
>> deliberately built by the author.

That's the quality of a good film or a good play, isn't it? You're
captured in the world the author has designed because of its quality --
not because you get to change the direction of the narrative.

> There's a perpetually recurring debate about what's important in a work
> of interactive fiction. At its crudest, it boils down to "game" vs.
> story. No offense, but many of us don't approach a work of IF as a game
> qua game, in the sense that chess is; it's an experience, in the same
> sense that a work of fiction is - a novel, say, or a film.

Well said. Is a poorly-written game very playable? Perhaps, but only if
you don't care about the writing, which begs the point if you're playing
IF.

>> It seems to me that this is because
>> of the difficulties of language. It is easy to display a complex system
>> graphically so that the patterns are visible to the player. The
>> computer doesn't need to 'understand' the patterns in order to
>> represent them. This is not true of language. If I was to write a
>> complex physical system and then describe its output verbally, I would
>> have to be able to recognise the patterns in order to be able to
>> present them to the reader.
>
> Undoubtedly true...
>
>> This is a tricky problem and puts IF at a serious disadvantage. Does
>> anyone have any insight into how we make progress with it?
>

> ....but I have trouble with the conclusions you draw. The tools you're

> talking about are good for producing *certain types of works*, and
> getting better. I regard those works as consituting a valid medium for
> creatives to work in, and for players/readers to enjoy. I see no
> "significant disadvantage" - and I think you perceive one because you're
> trying to find an entirely different kind of finished product. So yes,
> *necessarily*, you need different tools to create them.
>
> I'm sure you don't intend to be dismissive of the existing body of IF,
> and I don't intend to be dismissive of the sort of thing you want to
> see, but I think the underlying differences between them are more than
> superficial. Comparisons may be of limited (no?) utility.
>
> I think, ultimately, your "emergent simulation IF" has to be considered
> on its own, not as the evolutionary next step from a cruder technology -
> and I think I'd be right in saying that for this newsgroup to accept it,
> it will have to take the "fiction" in "interactive fiction" extremely
> seriously, not disregard it to focus on games qua games.

At this point, it sounds like the original author wants to create more
of a sim model than even a game -- I'm seeing something like Sims: Text.
This has been explored in the form of text chat rooms. People can say
(and do, if you count emotes) whatever they want and whole plots,
histories, and so forth develop. Maybe you could adapt that to a game
(say in the blue room the agents could only have set of actions X, but
in the white room, they could do set of actions Y).

I'm just not sure how interesting sim-based IF would be, though.

-- Poster

www.intaligo.com Building, INFORM, doom metal

David Fisher

unread,
Mar 28, 2006, 6:43:33 PM3/28/06
to
"Malcolm Ryan" <malcol...@gmail.com> wrote in message
news:1143525880....@z34g2000cwc.googlegroups.com...

> I'm currently studying game design, and one of the keep concepts is
> that of emgergence. Many of the most enduring games have this property:
> the game world is a complex system with emergent properties
...

> On the other hand, interactive fiction games are entirely scripted.
> They is no room for emergent effects, every outcome has been
> deliberately built by the author.

You may be interested in an article called "IF, AI, and the
confabulating-arranger model of interactive fiction" by Jorn Barger (1994):

http://www.robotwisdom.com/ai/ifdesign.html

... in particular, the section called "Reverse-engineering
emergent-behavior". Here is a quote:

While the artificial-life theorists have been hoping to discover
a few simple laws that will result in a vast payoff of interesting
stories, via the principle of 'emergent behavior', the meta-story
approach might be thought of as an attempt to outdo them by
'cheating'-- first you spell out the stories you want to see
'emerge', and only then do you analyse them and try to 'reverse
engineer' the set of simpler laws that will produce them.

There have also been some past posts about the related topic of simulation &
IF; here is an interesting one:

http://groups.google.com/group/rec.arts.int-fiction/browse_thread/thread/bcac501cdd241b54

For some more threads on this topic (if you feel like reading !), have a
look at:

http://www.ifwiki.org/index.php/Past_raif_topics:_Game_Mechanics:_part_1#Simulation_.2F_world_limits

David Fisher


Malcolm Ryan

unread,
Mar 28, 2006, 6:59:40 PM3/28/06
to
I wrote:
> On the other hand, interactive fiction games are entirely scripted.
> They is no room for emergent effects, every outcome has been
> deliberately built by the author.

Let me be clear. I was not offering a criticism of existing IF. There's
nothing "wrong" with scripted stories. It's just that I want to build
something else. I am an avid MOO builder/programmer and I try to make
my worlds as richly modelled as possible. But every event that occurs
in my world must be an event that I have explicitly scripted. There's
little room for unexpected patterns to arise out of the superposition
of two separate events. The whole is NOT more than the sum of its
parts.

I'm an AI academic and I've studied "interactive narrative" in detail.
I know all about drama management and so forth. Personally, I think
drama management is reaching too far. We don't even have the ability to
recognise and represent drama, let alone to generate it.

For my own research, I'm trying something much simpler (and still
enormously difficult). I have written a simple physical simulation of a
pool table, and I am trying to use it to generate a text-based pool
game. The challenge will be to recognise the important/interesting
patterns on the table, and in the outcomes of a shot, and describe them
to the player. There's no attempt at managing the drama, just
describing it.

In any case, I am only raising the issue in order to explore where it
may lead. I think the development of natural language tools capable of
this kind of description would be useful for building all kinds of
interactive text, be they stories, games, toys or other more 'serious'
endeavours.

Malcolm

ems...@mindspring.com

unread,
Mar 28, 2006, 7:51:50 PM3/28/06
to

Malcolm Ryan wrote:
> I wrote:
> > On the other hand, interactive fiction games are entirely scripted.
> > They is no room for emergent effects, every outcome has been
> > deliberately built by the author.
>
> Let me be clear. I was not offering a criticism of existing IF. There's
> nothing "wrong" with scripted stories. It's just that I want to build
> something else. I am an avid MOO builder/programmer and I try to make
> my worlds as richly modelled as possible. But every event that occurs
> in my world must be an event that I have explicitly scripted. There's
> little room for unexpected patterns to arise out of the superposition
> of two separate events.

I think there are two possible issues here, one of which is descriptive
issue and the other of which is a simulation issue.

Re. description: I'm not sure this makes it all the way to
"unexpected", but one of the things that became clear to me after a
little experimentation with the Reactive Agent Planner NPC library was
that I needed some way of collating and evaluating events before
reporting them. If you have one NPC allowed to do one thing per turn,
then reports such as

Durbin closes the door.

are fine; as soon as you have multiple actions in a turn, like

Durbin opens the door. Durbin goes through the door.

or multiple characters acting, as in

Durbin goes through the door. Louis goes through the door.

the output looks clumsy. So the thing to do is to calculate what
happens in a turn; then do some processing to match particularly
interesting ones and group the rest together mechanically; then run the
actual report phase.

This is what I did with NPC movement in "Mystery House Possessed",
which (despite an assortment of other flaws) does reasonably well at
reporting multiple characters entering or leaving rooms together,
crossing paths, reacting to one another's presence, and so on. I know
some other people have been thinking about related issues -- Nick
Montfort as a general problem in defining description rules separately
from the world model, and Steve Breslin as an outgrowth of NPC action
in general. But I think the main thing that needs to happen, for this
to work, is a separation of the simulation phase (where actions are
worked out) from the description phase (where actions are described to
the player).

And then, yes, you do have the challenge of writing a description
generator; it's at that point that you make decisions about what kinds
of patterns you want to match and describe to the player. Most of the
kinds of work I've done myself so far have been concerned with trying
to make the reports sound like natural English output, so combining
into, e.g.,

Durbin and Louis go through the door.

...but there is no intrinsic reason why these routines shouldn't a)
produce more lyrical output or b) take more complicated factors into
account. Conceivably, a game with some notion of Durbin and Louis as
people with emotions might concoct

Humiliated, Durbin and Louis slink out.

or

Louis walks out angrily. Durbin gives you an exasperated look that
plainly asks why you couldn't hold your tongue; then he follows.

I can see the argument that this much is no more than dressing, laid
over what is still a model comprising lots of uninteresting small
events, and that is perhaps where something more dramatically ambitious
comes into play. "Mystery House Possessed" (which became sort of a
dumping ground for all the ideas I had about this at the time I wrote
it) also tries to find excuses to trigger specific kinds of scene --
I'm not sure it deserves a term as glorified as "drama management", but
it has scripts for kinds of conversation that can occur, which require
participants fitting certain descriptions (e.g., two people who dislike
one another; or three people discovering a corpse; or a couple in love
and a jealous third party; etc.) Whenever it finds the starting
conditions fulfilled, it triggers the scene, using some specific
information about the characters to fill in the details of the script.
But it is still the small-scale simulation of people moving around,
combined with the player's intervention, that determines which scenes
will trigger when.

steve....@gmail.com

unread,
Mar 28, 2006, 8:06:49 PM3/28/06
to

Malcolm Ryan wrote:

> I'm an AI academic and I've studied "interactive narrative" in detail.
> I know all about drama management and so forth. Personally, I think
> drama management is reaching too far. We don't even have the ability to
> recognise and represent drama, let alone to generate it.

Being that you're an AI academic, I'm sure you're aware of a lot of the
groundbreaking work in DM that people are doing in sister genres to IF.
But I'm surprised you know all about DM, but still think it's reaching
too far. You've said a couple things that seem really iffy to me.

You should know that it is a framework for warping the narrative to
guide it towards a satisfying sequence, according to a narrative
heuristic provided by the programmer. The machine itself "recognises"
drama in an even more limited sense than a chess program recognizes a
winning position -- that is, only by evaluating it through and by a
simple criterion provided by the programmer. And even state-of-the-art
programs do nothing to "generate" drama, but only modify what is
already generated by warping the situation in ways also predefined by
the programmer.

> For my own research, I'm trying something much simpler (and still
> enormously difficult). I have written a simple physical simulation of a
> pool table, and I am trying to use it to generate a text-based pool
> game. The challenge will be to recognise the important/interesting
> patterns on the table, and in the outcomes of a shot, and describe them
> to the player. There's no attempt at managing the drama, just
> describing it.

Interesting. You might want to check out the Battle for Walcot Keep,
similar in some ways, availible via:

http://wurb.com/if/game/2688

Andrew Plotkin

unread,
Mar 29, 2006, 1:49:32 PM3/29/06
to
Here, Malcolm Ryan <malcol...@gmail.com> wrote:
>
> For my own research, I'm trying something much simpler (and still
> enormously difficult). I have written a simple physical simulation of a
> pool table, and I am trying to use it to generate a text-based pool
> game. The challenge will be to recognise the important/interesting
> patterns on the table, and in the outcomes of a shot, and describe them
> to the player. There's no attempt at managing the drama, just
> describing it.

That is indeed an interesting problem.

From a game-design standpoint, I think I wouldn't *want* a physical
simulation down underneath. IF works when both the game and the player
(output and input language) can comfortably describe all the important
distinctions of the world state. The "important distinctions" in a
physical simulation of ball movement[*] go down to tiny details of
position and angle. Recognizing all of these without resorting to a
picture is, as you say, enormously difficult. It would be much easier
to build a high-level model and work with it directly.

This is not to say you shouldn't do what you're doing. I'm just saying
that, if I wanted a pool game in text IF, I wouldn't do that. :)

[* no comments from the Thornton gallery, please.]

There's actually an interesting comparison between this and CRPG-style
combat (as in Graham's recent _Reliques of Tolti-Aph_). The kind of
combat in that game is also a numerical simulation. It happens to be
very *easy* to describe in text: you have twelve hit points, the
goblin has four, you hit for three points of damage. Boff! Bam! Pow!

However, my reaction is exactly the same: I'd rather scrap the
simulation and build a high-level model of the scene instead. Map out
a network of interesting ("scripted"!) actions and reactions, rather
than rolling dice.

So perhaps the difficulty of the pool problem is something of a red
herring.

--Z

--
"And Aholibamah bare Jeush, and Jaalam, and Korah: these were the borogoves..."
*

If the Bush administration hasn't thrown you in military prison
without trial, it's for one reason: they don't feel like it. Not
because of the Fifth Amendment.

Mike Roberts

unread,
Mar 29, 2006, 3:23:34 PM3/29/06
to
"Andrew Plotkin" <erky...@eblong.com> wrote:
> Here, Malcolm Ryan <malcol...@gmail.com> wrote:
>> For my own research, I'm trying something much simpler (and still
>> enormously difficult). I have written a simple physical simulation of
>> a pool table, and I am trying to use it to generate a text-based pool
>> game. The challenge will be to recognise the important/interesting
>> patterns on the table, and in the outcomes of a shot, and describe
>> them to the player.
>
> From a game-design standpoint, I think I wouldn't *want* a physical
> simulation down underneath. IF works when both the game and the
> player (output and input language) can comfortably describe all the
> important distinctions of the world state. [...]

> So perhaps the difficulty of the pool problem is something of a red
> herring.

I was getting ready to write pretty much this same thing, so, what he said.

There's one part I want to amplify. You've framed the problem as though it
were an AI problem: to recognize the interesting patterns in a pool game and
describe those to the player. The more I've thought about this, though, the
less convinced I am that this is really the nature of the problem you're
tackling. The fact is, I think a *human* narrator would have a hard time
effectively describing the state of a pool game verbally. Assuming the pool
game is interactive and not just something going on among NPCs, it seems
equally hard for a player to describe their *input* verbally. I mean, you
could phrase input in terms of your intention ("hit the 7 ball into the
northeast corner pocket"), but pool is a game of physics, not intentions.
It seems to me that the real problem here isn't the AI problem of rendering
pool verbally, but in coming up with a verbal pool notation in the first
place.

You raised chess as another similar example earlier. Chess actually does
have a couple of established verbal rendering systems (the "knight to king's
bishop 3" notation, for example, or the algebraic systems), and it's
interesting to note that they operate purely at the component level -
there's no pattern recognition involved, and no AI problem to solve, in
generating descriptions in those systems. I'm at best a recreational chess
player, so for all I know, chess experts might use more advanced notation
systems that operate at a pattern level ("a knight-modified Kurnikov
formation in the third rank" or some such); if such a language exists, it
would indeed be an AI problem to render descriptions with it. But attacking
this problem in an IF context seems almost counterproductive, since most
players wouldn't have any idea what to make of the result; for most players,
you'd be better off just using a straightforward component-level description
of the board.

Or better still, showing it graphically. Even literary (static) fiction
authors occasionally resort to in-line illustrations, when that's the most
efficient way to convey some detail that requires geometrical precision to
be properly understood.

If emergent effects are your real interest, I think you'd be a lot better
off decoupling that problem from what I see as the completely separate
problem of verbally rendering a game of positions and velocities and spins.
Have you looked at the Infocom game "Suspended"? I think it arguably has
some emergent effects, and it works within the kind of physical simulation
framework where the IF rendering is already understood.

--Mike
mjr underscore at hotmail dot com


ems...@mindspring.com

unread,
Mar 29, 2006, 4:14:38 PM3/29/06
to

Mike Roberts wrote:
> There's one part I want to amplify. You've framed the problem as though it
> were an AI problem: to recognize the interesting patterns in a pool game and
> describe those to the player. The more I've thought about this, though, the
> less convinced I am that this is really the nature of the problem you're
> tackling. The fact is, I think a *human* narrator would have a hard time
> effectively describing the state of a pool game verbally.

Yes, I agree with this also.

My favorite simple thought-experiment/test case for the
object-description problem is a card game. The player's hand consists
of specific cards, and it is not hard to model this in IF and to create
instructions to handle the interaction (DISCARD QUEEN OF CLUBS, DRAW
TWO CARDS, PLAY TEN OF DIAMONDS, etc). At the same time, there are
specific patterns one would want to describe specially: flushes and
straights, say, if we're doing poker. The solution is basically to run
down a series of cases, from most specific to least specific, until we
find one that applies, and use that template for the description: check
for the royal flush first, then the straight flush, then... (etc.)

The advantage here is that there's already a well-developed terminology
for describing all the patterns that arise, so one can focus on the
programming, which in that case turns out to be relatively easy. But
it's really not *that* different a problem from a number of others that
arise in IF:

-- describing the liquid created when the player mixes different
amounts of different components, or the color created by multiple
colored light-sources;
-- describing your basic rope in terms of where it's attached or free;
-- describing the machinery that results when several separable
components are plugged into one another or glued together;
-- describing sets of things from which singlets are unusual or missing
(a travel guide set lacking volume Paris-Prague; three men, of whom one
is wearing a hat).

Some of these are harder than others; the liquid-mixture problem
probably gets closest to the complexity of the pool-table problem,
because though we can think of ways to describe most kinds of mixture
(dirt, water --> mud), a system that tracks quantities will allow for a
pinch of dirt in a gallon of water, or a drop in the Sahara, and by
rights those should be named differently. I believe this is actually a
linear algebra problem -- one might define a bunch of vectors that
represent specific named mixtures (1 water by 1 dirt is called "mud"),
then finding the vector which our actual mixture most closely
approximates and using the resulting information about how much it
varies to add nuance (2 water by 1 dirt is closest to being "mud", but
it exceeds in the wateriness dimension, so we call it "watery mud").
This will be most interesting if we have a bunch of named intermediate
vectors as well as the bases ("sludge", "salad dressing",
"Bordeaux-style blend"): these correspond to our more specialized
cases.

Still, in most of these cases I find the challenge lies more in the
subsequent parsing than in generating the description. If I write the
output to distinguish men by whether they're wearing hats, then I would
also like the player to be able to distinguish in the same way: ie, I
want to be able to parse "the man in the hat".

The problem of collating and describing NPC actions is a bit trickier,
because there are so many more things to consider: do we group together
the same action performed by multiple people (Larry and Moe fall down),
or different actions performed by one person (Moe slips and falls), or
different actions performed on the same object (Moe and Larry slip on
the banana)? How do we track relative time within a turn when each of
these things occurs? Differentiate between simultaneous and sequential
effects? Express or suppress incidental information with, say,
adverbial phrases describing the manner in which the action was
undertaken? And so on.

Adam Thornton

unread,
Mar 29, 2006, 5:46:10 PM3/29/06
to
In article <e0ekrs$t38$1...@reader2.panix.com>,

Andrew Plotkin <erky...@eblong.com> wrote:
>The "important distinctions" in a
>physical simulation of ball movement[*]
>[* no comments from the Thornton gallery, please.]

Well, I never!

Well, hardly ever.

Adam

Kevin Forchione

unread,
Mar 29, 2006, 8:35:16 PM3/29/06
to
"Mike Roberts" <mj...@hotmail.com> wrote in message
news:a1CWf.60589$H71....@newssvr13.news.prodigy.com...

> You raised chess as another similar example earlier. Chess actually does
> have a couple of established verbal rendering systems (the "knight to
> king's bishop 3" notation, for example, or the algebraic systems), and
> it's interesting to note that they operate purely at the component level -
> there's no pattern recognition involved, and no AI problem to solve, in
> generating descriptions in those systems. I'm at best a recreational
> chess player, so for all I know, chess experts might use more advanced
> notation systems that operate at a pattern level ("a knight-modified
> Kurnikov formation in the third rank" or some such); if such a language
> exists, it would indeed be an AI problem to render descriptions with it.
> But attacking this problem in an IF context seems almost
> counterproductive, since most players wouldn't have any idea what to make
> of the result; for most players, you'd be better off just using a
> straightforward component-level description of the board.

Grandmasters and woodpushers alike use the same notation in recording a
game. In describing a game one can speak of patterns and themes, and
computer games such as Chessmaster do a fairly good job of translating the
results of a move or a series of moves into text. For instance the game
might say something like, "You move the knight to f3. Black moves his pawn
on h7 to h6. You move your Bishop to g5 pinning black's knight. Black takes
your bishop with the pawn at h6, gaining a piece and a clear material
advantage." All the while you're simply communicating with the game via the
standard notation, or more likely mouse clicks.

--Kevin


Malcolm Ryan

unread,
Mar 29, 2006, 8:46:38 PM3/29/06
to
steve....@gmail.com wrote:
> Malcolm Ryan wrote:
>
> > I'm an AI academic and I've studied "interactive narrative" in detail.
> > I know all about drama management and so forth. Personally, I think
> > drama management is reaching too far. We don't even have the ability to
> > recognise and represent drama, let alone to generate it.
>
> Being that you're an AI academic, I'm sure you're aware of a lot of the
> groundbreaking work in DM that people are doing in sister genres to IF.
> But I'm surprised you know all about DM, but still think it's reaching
> too far. You've said a couple things that seem really iffy to me.
>
> You should know that it is a framework for warping the narrative to
> guide it towards a satisfying sequence, according to a narrative
> heuristic provided by the programmer. The machine itself "recognises"
> drama in an even more limited sense than a chess program recognizes a
> winning position -- that is, only by evaluating it through and by a
> simple criterion provided by the programmer. And even state-of-the-art
> programs do nothing to "generate" drama, but only modify what is
> already generated by warping the situation in ways also predefined by
> the programmer.

Yes, and my argument is that we have very poor tools for expressing
those heuristics so that the results actually line up with what we
think of as real drama. Sure we can build search algorithms which
maximise a heuristic, but calling that 'Drama Management' is begging
the question if you cannot actually write the heurisitc.

Like I said, we need to build tools to recognise and represent drama so
that we can write decent heuristics to guide our search. The search is
the easy bit. I know Michael Young's group at NSCU are working on
trying to build models of "suspense" and "conflict", although I haven't
seen any publcications yet. I can't say I have seen much other work
like this. Do you kniow of any?

Malcolm

Malcolm Ryan

unread,
Mar 29, 2006, 10:01:56 PM3/29/06
to
Mike Roberts wrote:
>
> There's one part I want to amplify. You've framed the problem as though it
> were an AI problem: to recognize the interesting patterns in a pool game and
> describe those to the player. The more I've thought about this, though, the
> less convinced I am that this is really the nature of the problem you're
> tackling. The fact is, I think a *human* narrator would have a hard time
> effectively describing the state of a pool game verbally.

I don't think it would be too hard. If you were playing a live pool
game and were commentating it shot by shot to an blind friend, I think
you could pick out the important features of what happened.
Maybe not quite in exquisite detail, but enough to get a feel for
what's going on. Providing enough exact detail to _play_ the game, is
admittedly more difficult. I envisage it as a series of directed
observations, taking into account intention. Like:

> LOOK AT TABLE

The cue ball is at the top of the table, near the left hand pocket. At
the bottom of the table there is a pair of balls, the eight and the
five, near the right hand pocket, and the seven in the open near the
middle

> CONSIDER FIVE

The five ball is near the bottom right pocket, but the eight ball is
blocking a straight shot. You cannot see any easy way to sink it.

> CONSIDER SEVEN

The seven ball is in the middle of the table near the bottom. You could
probably sink it in the bottom left pocket with a kick shot off the
bottom left rail, or with a straight shot into the right pocket.

> Assuming the pool
> game is interactive and not just something going on among NPCs, it seems
> equally hard for a player to describe their *input* verbally. I mean, you
> could phrase input in terms of your intention ("hit the 7 ball into the
> northeast corner pocket"), but pool is a game of physics, not intentions.
> It seems to me that the real problem here isn't the AI problem of rendering
> pool verbally, but in coming up with a verbal pool notation in the first
> place.

Well, pool is game of both physics and intentions. You plan a shot and
then you execute it. For my game, you will express your intention to
the engine and it will compute the necessary physics. I can already do
this for some quite sophiticated shots. But you're right, coming up
with a parsable language for expressing your desires is difficult.
Describing a shot to sink a ball is not so hard. But there are subtler
plays as well, like positioning a ball near a certain pocket. Or
knocking an opponents ball away.

> SINK SEVEN IN BOTTOM LEFT POCKET

You kick the cue ball off the bottom left rail and it collides solidly
with the brown. The seven rolls straight into the bottom left pocket
with a satisfying clunk. Meanwhile the cue ball stops near the middle
of the table, leaving you a nice straight shot on the five.

> If emergent effects are your real interest, I think you'd be a lot better
> off decoupling that problem from what I see as the completely separate
> problem of verbally rendering a game of positions and velocities and spins.
> Have you looked at the Infocom game "Suspended"? I think it arguably has
> some emergent effects, and it works within the kind of physical simulation
> framework where the IF rendering is already understood.

I'm familiar with the game. Where do you see emergence in it?

Malcolm

Mike Roberts

unread,
Mar 29, 2006, 11:24:08 PM3/29/06
to
"Malcolm Ryan" <malcol...@gmail.com> wrote:
>> The fact is, I think a *human* narrator would have a hard time
>> effectively describing the state of a pool game verbally.
>
> I don't think it would be too hard. If you were playing a live pool
> game and were commentating it shot by shot to an blind friend, I
> think you could pick out the important features of what happened.
> Maybe not quite in exquisite detail, but enough to get a feel for
> what's going on. Providing enough exact detail to _play_ the
> game, is admittedly more difficult.

That's precisely what Andrew was getting at when he said that a deep
simulation might not be the way to go here. If you're not describing the
deep simulation in detail, and you're not allowing the player to make inputs
that relate to the deep simulation, then what's the point of having the deep
simulation? If all you need is to give the player a feel for what's going
on, I expect you could create a convincing enough effect with a lot less
mechanism than a complete pool table simulator.

> For my game, you will express your intention to the engine and
> it will compute the necessary physics.

And presumably this is adjusted by random factors for the player's skill,
and presumably those factors are somewhat opaque to the player, so from the
player's perspective the outcome is somewhat randomized. If that's the
case, it also decreases the value to the player of a deep simulation - once
you have random numbers involved, it doesn't really matter to the player how
complicated the math behind the scenes is, since everything will look more
or less like a probability distribution.

I don't mean to knock the project, by the way. I mean, it sounds like an
interesting enough challenge. I'm just saying that there might be easier
ways of getting more or less the same player-visible effect that fit better
into the IF interaction model.

>> Have you looked at the Infocom game "Suspended"? I think
>> it arguably has some emergent effects, and it works within the kind
>> of physical simulation framework where the IF rendering is already
>> understood.
>
> I'm familiar with the game. Where do you see emergence in it?

It's been a while since I looked at it, but doesn't it have a fair amount of
interaction among the capabilities of the robots? It seems like it was
uncharacteristically simulationist for an Infocom game and had some behavior
that was more rule-based than scripted.

Another one to look at is "Four in One" (http://wurb.com/if/game/476), which
I also think exhibits some rule-based emergent behavior (in this case, in
the interactions among the NPCs).

Kevin Forchione

unread,
Mar 30, 2006, 2:16:39 AM3/30/06
to
"Mike Roberts" <mj...@hotmail.com> wrote in message
news:I3JWf.61598$Jd.3...@newssvr25.news.prodigy.net...

> "Malcolm Ryan" <malcol...@gmail.com> wrote:
>>> The fact is, I think a *human* narrator would have a hard time
>>> effectively describing the state of a pool game verbally.
>>
>> I don't think it would be too hard. If you were playing a live pool
>> game and were commentating it shot by shot to an blind friend, I
>> think you could pick out the important features of what happened.
>> Maybe not quite in exquisite detail, but enough to get a feel for
>> what's going on. Providing enough exact detail to _play_ the
>> game, is admittedly more difficult.
>
> That's precisely what Andrew was getting at when he said that a deep
> simulation might not be the way to go here. If you're not describing the
> deep simulation in detail, and you're not allowing the player to make
> inputs that relate to the deep simulation, then what's the point of having
> the deep simulation? If all you need is to give the player a feel for
> what's going on, I expect you could create a convincing enough effect with
> a lot less mechanism than a complete pool table simulator.

Quite. There should logically be a correspondence between the granularity of
the user interface and the "black box" of the parser/world model.

Btw, a person born blind is certainly not playing with the same model of the
world that a sighted person would. Many persons born blind would, for
example, describe an intersection as a point where 4 roads connect, rather
than where 2 are crossing. Blindness is not the same thing as closing one's
eyes!

--Kevin


Tommy Herbert

unread,
Mar 30, 2006, 6:54:30 AM3/30/06
to
Mike Roberts wrote:
> Another one to look at is "Four in One" (http://wurb.com/if/game/476), which
> I also think exhibits some rule-based emergent behavior (in this case, in
> the interactions among the NPCs).

How about Savoir-Faire? I think Emily Short wrote on her website that
she aimed to make the game-world fairly simulationist, in that puzzle
solutions depended on interactions between properties rather than
specific objects. That means sometimes players come up with answers
that work even though she hasn't anticipated them.

Tommy Herbert

unread,
Mar 30, 2006, 7:08:58 AM3/30/06
to

Malcolm Ryan wrote:
> The cue ball is at the top of the table, near the left hand pocket. At
> the bottom of the table there is a pair of balls, the eight and the
> five, near the right hand pocket, and the seven in the open near the
> middle
...

> The seven ball is in the middle of the table near the bottom. You could
> probably sink it in the bottom left pocket with a kick shot off the
> bottom left rail, or with a straight shot into the right pocket.
...

> > SINK SEVEN IN BOTTOM LEFT POCKET
>
> You kick the cue ball off the bottom left rail and it collides solidly
> with the brown. The seven rolls straight into the bottom left pocket
> with a satisfying clunk.

I can't picture it! As far as I can tell, a cue ball coming from the
bottom-left direction (having bounced off two cushions) hits the seven,
but instead of travelling in the same direction (so towards top-right),
the seven goes into the bottom-left pocket. What's going on?

As others have said, language doesn't lend itself to descriptions of
this kind. I suppose you do get radio commentary of snooker, but I
find that abominably hard to follow. Normally, a person describing a
pool game to someone else would unthinkingly adopt a bimodal way of
communicating - he'd start sketching diagrams with his finger on the
tabletop. I realise this is only an exercise and that your work could
have useful non-pool-related applications, but is there really no
domain that serves as a better test-bed?

Kevin Forchione

unread,
Mar 30, 2006, 11:55:14 AM3/30/06
to
"Tommy Herbert" <tommyh...@gmail.com> wrote in message
news:1143719670.7...@v46g2000cwv.googlegroups.com...

> How about Savoir-Faire? I think Emily Short wrote on her website that
> she aimed to make the game-world fairly simulationist, in that puzzle
> solutions depended on interactions between properties rather than
> specific objects. That means sometimes players come up with answers
> that work even though she hasn't anticipated them.

Interestingly that sort of result (unanticipated) used to be called a
program bug.

No, rather what she's done is abstract the solution to puzzles, and the
interaction of objects up a level. Rather than say that you need object X to
solve in the process of resolving a certain situation, you need property X.
That makes it appear as though magical things are happening, but you still
need to code the interactions, they're simply moved into a more abstract
layer. So she's had to anticipate them, and the properties involved are now
as critically important as the objects once were.

--Kevin


steve....@gmail.com

unread,
Mar 30, 2006, 12:18:54 PM3/30/06
to
Mike Roberts wrote:
> If you're not describing the
> deep simulation in detail, and you're not allowing the player to make inputs
> that relate to the deep simulation, then what's the point of having the deep
> simulation?

This is exactly the direction I was hoping to generalize the
discussion. Putting the pool-table simulation aside, is there anything
this little thought-experiment teaches us about IF modeling in general,
and particularly of interest to me, NPC modeling and drama management?

It's my feeling that NPC-AI systems to this point (from Erasmatron to
Facade, and in my own meager efforts, I myself am probably the *most*
guilty of this), where they fall short, make the mistake of describing
what they model, rather than modeling what they want to describe. So
this would corroborate Mike's point.

But on the other hand, I believe the obvious point that the model is
not the output leads immediately to the less obvious point that it's
impossible to model the right level, and that this kind of problem (the
disjoint between model and output) is inevitable.

Emily writes of a report system she and I have been discussion since
Walcot, which takes the particularities of the simulation (basically
what Mike means by "deep simulation") and translates it into more
presentable language, in part by cutting out some of the detail. Some
such mechanism is necessary no matter what level you model. This does
not mean that the cut-detail was too deep to begin with.

The main distinction is this: of course it's pointless to model what is
absolutely irrelevant to the output, but it is necessary to model what
is translated into output, what is indirectly reflected in output.

Consider a NPC psychological model, where there's a number of axes of
emotion (happy<->sad, friendly<->hostile), which it's convenient to
keep on a graduated scale. (So the NPC might be happy=4, friendly=9 for
instance.) We're never going to see these scalars in the output -- and
maybe the output is not even generated directly from these scalars at
all, but from which of the three or four states the NPC can be in
(determined by some function we've written which takes the
emotion-scalars as parameters, and returns a simple state) -- in other
words, maybe there's more than one level of translation. The model
seems to me perfectly sound, even though the level of detail is deeper
than is reflected in the output.

This is because we're expecting the player to draw conclusions about
the underlying model (that Bob gets angrily gradually or so), to infer
that depth which is *not* represented in the output.

(But this brings me back to my major preoccupation these days, and
which is only tangentially related to this discussion: how far we can
legitimately expect the player to appreciate a dynamic model, without
making the (to me, suspicious and anti-aesthetical) move of saying (to
quote Facade's instructions) "re-play the drama to find out how your
interaction could make things turn out differently the next time."
Doesn't there have to be a more direct way of giving the player a sense
of agency?)

Kevin Forchione

unread,
Mar 30, 2006, 12:19:08 PM3/30/06
to
"Tommy Herbert" <tommyh...@gmail.com> wrote in message
news:1143720538....@u72g2000cwu.googlegroups.com...

> As others have said, language doesn't lend itself to descriptions of
> this kind. I suppose you do get radio commentary of snooker, but I
> find that abominably hard to follow. Normally, a person describing a
> pool game to someone else would unthinkingly adopt a bimodal way of
> communicating - he'd start sketching diagrams with his finger on the
> tabletop.

That would depend, I think, on the person doing the describing. Going back
to the chess analogy... if I describe the game "blow-by-blow" so to speak,
using the algebraic notation, let's say, then your audience would quickly
become lost... and begin scrabbling to set up their chessboards (even I
can't follow a game in my head described like that these days, though I
played a few blindfolded when I was young by having the opponents moves told
to me, and issuing my moves verbally). If someone asked, how did the game
go? I'd hardly respond with "1. e4 e5 2. Nf3 Nc6 3. Bb5...", except for
comic effect. However... you could describe the "bigger picture" by
discussing the well-known patterns involved as the game progressed. For
instance, you might say that the white player opened the game with a Fischer
variation of the Ruy Lopez, etc. In certain circumstances the sequences of
the moves would be unimportant, and you'd simply mention the point at which
they transitioned into the recognized pattern. You'd then go on to
generalize the game: "weakened the pawn structure of black's kingside",
"promoted the a-pawn to a rook." "executed a back row mate in 7".

The drama in retelling the story is thus encapsulated, not in the details of
the game, but in the same kind of information chunking that is conveyed by
the phrase "potted the eight-ball". You see? Unless you're interested in
producing a kind of low-level language that would be akin to the "assembler
of pool". But such a language would be devoid of dramatic power (drowned in
the details, so to speak). Drama, of necessity requires a chunking of
information into larger archetypal patterns that convey meaning to the
broader audience. One could argue that conveying the minutiae in text,
translating the precise details of every step of the game would be a less
realistic representation.

--Kevin


Dan Shiovitz

unread,
Mar 30, 2006, 1:44:00 PM3/30/06
to
In article <Q3UWf.9823$8Y2.7267@fed1read03>,
Kevin Forchione <ke...@lysseus.com> wrote:
[..]

>No, rather what she's done is abstract the solution to puzzles, and the
>interaction of objects up a level. Rather than say that you need object X to
>solve in the process of resolving a certain situation, you need property X.
>That makes it appear as though magical things are happening, but you still
>need to code the interactions, they're simply moved into a more abstract
>layer. So she's had to anticipate them, and the properties involved are now
>as critically important as the objects once were.

And isn't this exactly the definition of emergence? It's a pattern
formed from a series of lower-level interactions. Even though the
*program* might analyze stuff at the property level, the *player* is
interacting at the object level, and any puzzle solutions have to get
created indirectly by the manipulation of those objects.

>--Kevin
--
Dan Shiovitz :: d...@cs.wisc.edu :: http://www.drizzle.com/~dans
"He settled down to dictate a letter to the Consolidated Nailfile and
Eyebrow Tweezer Corporation of Scranton, Pa., which would make them
realize that life is stern and earnest and Nailfile and Eyebrow Tweezer
Corporations are not put in this world for pleasure alone." -PGW

Malcolm Ryan

unread,
Mar 30, 2006, 6:09:24 PM3/30/06
to

Mike Roberts wrote:

> "Malcolm Ryan" <malcol...@gmail.com> wrote:
>
> That's precisely what Andrew was getting at when he said that a deep
> simulation might not be the way to go here. If you're not describing the
> deep simulation in detail, and you're not allowing the player to make inputs
> that relate to the deep simulation, then what's the point of having the deep
> simulation? If all you need is to give the player a feel for what's going
> on, I expect you could create a convincing enough effect with a lot less
> mechanism than a complete pool table simulator.

You may have a point, but I was inspired to do this after playing a
pool game on a MUD which clearly had no "deep simulation" behind it. It
felt unreal. Now, possibly the solution to this is just to build a more
convincing top-down version, but I'm not so sure. In any case, I
thought that this would make an interesting research project. It's a
small example of what I perceive to be a much larger problem. If we
can't recognise the interesting emergent events in a simple physical
system, how do we expect to do the same in a complex social scenario
(like, say, Facade)?

> >> Have you looked at the Infocom game "Suspended"? I think
> >> it arguably has some emergent effects, and it works within the kind
> >> of physical simulation framework where the IF rendering is already
> >> understood.
> >
> > I'm familiar with the game. Where do you see emergence in it?
>
> It's been a while since I looked at it, but doesn't it have a fair amount of
> interaction among the capabilities of the robots? It seems like it was
> uncharacteristically simulationist for an Infocom game and had some behavior
> that was more rule-based than scripted.

I wish to distinguish between interaction and emergence. Interaction
can be scripted. Emergence is when a certain pattern of interaction
arises which wasn't pre-arranged by the author. I don't recall any such
behaviour in suspended .Which is not to disparage the game. It was a
favorite of mine.

> Another one to look at is "Four in One" (http://wurb.com/if/game/476), which
> I also think exhibits some rule-based emergent behavior (in this case, in
> the interactions among the NPCs).

Thanks, I'll check it out.

Malcolm

ems...@mindspring.com

unread,
Mar 30, 2006, 7:00:00 PM3/30/06
to

steve....@gmail.com wrote:
> But on the other hand, I believe the obvious point that the model is
> not the output leads immediately to the less obvious point that it's
> impossible to model the right level, and that this kind of problem (the
> disjoint between model and output) is inevitable.
>
> Emily writes of a report system she and I have been discussion since
> Walcot, which takes the particularities of the simulation (basically
> what Mike means by "deep simulation") and translates it into more
> presentable language, in part by cutting out some of the detail. Some
> such mechanism is necessary no matter what level you model. This does
> not mean that the cut-detail was too deep to begin with.

No, certainly not in the case of NPCs moving around.

There's also maybe a useful distinction to be made between collating
details into a grouped report (Larry, Moe, and Fred all do X) and
actually dropping details out entirely -- in the former case, we've
just taken three separate NPC-report sentences and stuck them together
to make something that reads naturally, but all the same information is
conveyed that could be conveyed in separate sentences. Indeed, one of
the signs of non-ideal reporting (to my mind) is that it clogs the
screen with redundant phrasing or information.

> Consider a NPC psychological model, where there's a number of axes of
> emotion (happy<->sad, friendly<->hostile), which it's convenient to
> keep on a graduated scale. (So the NPC might be happy=4, friendly=9 for
> instance.) We're never going to see these scalars in the output -- and
> maybe the output is not even generated directly from these scalars at
> all, but from which of the three or four states the NPC can be in
> (determined by some function we've written which takes the
> emotion-scalars as parameters, and returns a simple state) -- in other
> words, maybe there's more than one level of translation. The model
> seems to me perfectly sound, even though the level of detail is deeper
> than is reflected in the output.

Yes. This is really yet another manifestation of the liquid-recipe or
light-color problem: you have an assortment of states that are actually
somewhat hazily defined (the state of being a martini, the state of
being maroon, the state of being well-disposed towards the person
you're talking to), and it might look peculiar to the player to shift
too rapidly from one state to another, because there's a grey area. For
a long time I was dissatisfied with modeling this with scalar
components even so, because I wasn't sure how to create a description
that wouldn't still *look* like rapid switching: Bob with an anger of
29 and an amusement of 30 and an attraction of 30 is all flirty, but
then nudge it just a little to anger 30, amusement 30, attraction 29,
and he turns into a snarky bastard. From the player's perspective this
is still too abrupt, I think, though you've added at least the nuance
that these feelings can build over time. It's possible to build
ludicrously-detailed switch statements to try to improve the
descriptions here, but the results can be pretty spaghetti-like.

I now think the solution involves, as with the liquids, defining each
named state (flirty, sarcastic, whatever) as a vector (composed of all
your scalars), and then describing the actual state in terms of the
nearest named vector, with a modifying adjective ("dry martini,
reddish-maroon, bitterly amused") reflecting the difference between the
named state and the actual state.

Probably this is all stuff that has been covered, better and in more
detail, by people who deal with AI academically (and they may have a
different, superior solution, as well), but this is more or less novel
territory to me; certainly when I wrote the code to describe mixtures
in Savoir-Faire, it was much more based on special cases and much less
on an intelligently worked-out system.

> (But this brings me back to my major preoccupation these days, and
> which is only tangentially related to this discussion: how far we can
> legitimately expect the player to appreciate a dynamic model, without
> making the (to me, suspicious and anti-aesthetical) move of saying (to
> quote Facade's instructions) "re-play the drama to find out how your
> interaction could make things turn out differently the next time."
> Doesn't there have to be a more direct way of giving the player a sense
> of agency?)

I think there is, but you may find it also anti-aesthetical: telegraph
important choices and modes of interaction to the player. Make it clear
that there is a point of decision coming up, though preferably frame it
in the most general terms your model will handle -- ie, suggest to the
player that it is important how he treats Bob, rather than hinting that
it would be good if he smiled a lot.

Or, if that won't work, make it very clear later how the results stem
from player action.

There's a silly easter egg in City of Secrets: there are security
cameras everywhere, and I added a response for >WAVE AT CAMERA; it
doesn't do anything important at the time, but if you wave at a camera
early in the game before one critical scene, that scene includes
replayed security footage of yourself waving. I stuck this in partly
because it seemed amusing, but also partly because I wanted to remind
the player that even apparently innocent actions might be tracked and
have some effect on the story. This apparently worked better than I
anticipated, because I heard from several people that they found this
very effective.

If one wanted to do something more useful with the idea, I could
imagine our hypothetical NPC Bob, when he finally gets really angry,
mentioning all the assorted small things the player did to tip him over
the edge. This would require some additional storage of information,
but the effect could be kind of interesting; and it doesn't
particularly violate the plausibility of the story and characters,
since arguments often seem to involve Person A presenting Person B with
an annotated list of grievances.

Approaching this from the other side, I'm trying to think of places
where I felt that I had an obvious effect on the course of the game
even though I didn't replay. The example I can think of is only
semi-satisfying, because the choice in question was explicitly
presented at the beginning of the game: I played Bolivia By Night only
once, but I was impressed with the fact that it obviously adapted the
story to fit the fact that I'd made my PC female. Someday I may go back
and replay, but since I was given the choice up front and then the text
reflected the results, I feel confident that players who chose a male
PC got different text in several places.

It was also clear to me in Slouching after one play that I'd affected
the action and how, though this did make me want to go back and
investigate the other possibilities; but I think Slouching is designed
to work as a composite of all the possible stories you get by playing
through all the endings, so I didn't feel this was a waste of time. I
didn't replay Bolivia because I felt that I had seen all the major
content, and the modifications, while noticeable, were not going to
make playing a second time a very different experience.

ems...@mindspring.com

unread,
Mar 30, 2006, 7:22:16 PM3/30/06
to

Malcolm Ryan wrote:
> I wish to distinguish between interaction and emergence. Interaction
> can be scripted. Emergence is when a certain pattern of interaction
> arises which wasn't pre-arranged by the author.

Okay -- though what do you mean by "pattern of interaction"? This
thread seems to be covering in different places all of the following:

1) the world-model can get into a very large number of possible states,
not all of which are easily described in language. How do we form
phrases to talk about things like position, spin, shape, etc., when
these are modeled at a level that we usually represent in drawings or
mathematical formulae? Your pool-table example seems to fall into this
category, and I agree that it may be quite hard to write descriptions
about it because it is simply hard to describe in language. I'm not
sure that constitutes a reason to despair of emergence in IF, though --
assuming I understand the term.

2) the world-model can get into a very large number of possible states;
we have natural linguistic ways to talk about these states, but it
would still be a burden for the author to write every one of these
descriptions by hand. How do we condense this information into an
auto-generated description, highlighting any exceptionally interesting
combinations? Into this category fall, I would say, chess notation,
hands of playing cards, the colors of mixtures of paint, complex NPC
moods, and a host of similar things. I would include also all the
variant descriptions for "NPC 1 and NPC 2 walk into the room from the
south", given a large number of NPCs and possible directions of entry
and exit, though this is mostly a question of cleaning up and making
presentable a set of reports that do already convey the necessary
information, rather than trying to condense abstract data into verbal
terms in the first place.

3) the world-model can get into a very large number of possible states
but behaves according to carefully-written rules, with the result that
the author does not have to specifically anticipate everything that the
player might do in order to provide sensible responses, and even to
allow the player to accomplish his goals within the world. This is
where the multiple-solution puzzle stuff comes in; it's also (I think)
the only one of the three categories that I would think of as involving
"interaction patterns" instead of "description patterns" (or something
along those lines).

Obviously the way the model is described to the player does have a
powerful effect on how well the interaction works, but (1) and (2)
remain largely output-processing; at best we might rig them so that the
parser will in appropriate circumstances understand "the red-brown
paint" or whatever, but mostly we are not giving the player carte
blanche to talk to the parser in new phrases, or to undertake any new
kinds of action himself.

Mike Roberts

unread,
Mar 30, 2006, 9:50:29 PM3/30/06
to
"Malcolm Ryan" <malcol...@gmail.com> wrote:
> I wish to distinguish between interaction and emergence.
> Interaction can be scripted. Emergence is when a certain
> pattern of interaction arises which wasn't pre-arranged by
> the author.

That's roughly the same meaning I use, although I'd generalize it somewhat
to say that emergent effects are those that aren't trivially predictable
from the rules governing a system, but which nonetheless follow necessarily
from those rules. The use of the term in gaming is a specialization of its
use in describing physical systems. Fluid dynamics, for example, can be
seen as the emergent effects of underlying molecular dynamics.

A characteristic of emergent effects in physical systems is that the
emergent effects are complex enough and internally consistent enough that
they seem to be governed by their own separate physical laws. Those
higher-level laws are essentially just mathematical approximations, in that
the true laws governing the emergent behavior are the more fundamental
underlying laws. But the approximations are often so accurate over so wide
a range, and the fundamental laws so computationally intractable when
applied to complex systems, that the higher-level law is the only practical
way to approach problems in the space.

Which could be seen as applying to IF in the ways I was talking about
earlier.

> I don't recall any such behaviour in suspended.

Perhaps not; my recollection was that it did have some, but as I said, I
haven't looked at it in a while.

Mike Roberts

unread,
Mar 30, 2006, 11:21:14 PM3/30/06
to
"Kevin Forchione" <ke...@lysseus.com> wrote:
> Going back to the chess analogy... if I describe the game "blow-by-blow"
> so to speak, using the algebraic notation, let's say, then your audience
> would quickly become lost... If someone asked, how did the game go? I'd
> hardly respond with "1. e4 e5 2. Nf3 Nc6 3. Bb5...", except for comic
> effect. However... you could describe the "bigger picture" by discussing
> the well-known patterns involved as the game progressed.
> The drama in retelling the story is thus encapsulated, not in the details
> of the game, but in the same kind of information chunking that is conveyed
> by the phrase "potted the eight-ball". Drama, of necessity requires a
> chunking of information into larger archetypal patterns that convey
> meaning to the broader audience.

I very much agree. The complication when you make it interactive, of
course, is that the player *could* try to poke her nose into the details,
even if they're not (in the author's judgment) important dramatically. For
example, you convey the story about how the PC got off to a good start in
the chess game, but seemed to lose focus after a dozen moves, and the NPC
opponent ended up winning. But then the player wants to EXAMINE BOARD to
see the layout at the end, to see if it's really a checkmate; and
furthermore compares the ending configuration with earlier descriptions to
see if you can really get here from there. If you actually let them see the
detailed configuration at the end or during the game, you open up a can of
worms, because then you have to maintain internal consistency throughout.
So you can either keep it vague, limiting the descriptions to the same level
of detail where the drama is taking place, or you can do the full
simulation.

José Manuel García-Patos

unread,
Mar 31, 2006, 4:45:42 AM3/31/06
to

> -- describing [...] the color created by multiple colored light-sources;

This was already done several years ago. The results were satisfactory
game-wise, story-wise and even simulationist-wise. The game was called "A
veces". It can be downloaded from here:

http://www.geocities.com/jene_if/descargas/aveces_4.zip

All The Best.
José Manuel García-Patos
Madrid

ems...@mindspring.com

unread,
Mar 31, 2006, 1:53:36 PM3/31/06
to

José Manuel García-Patos wrote:
> > -- describing [...] the color created by multiple colored light-sources;
>
> This was already done several years ago. The results were satisfactory
> game-wise, story-wise and even simulationist-wise.

Cool. Do you know how it was implemented?

Several of the things on this list have been done before, but don't
have a sort of standard implementation that makes them easy for people
to include.

> The game was called "A
> veces". It can be downloaded from here:
>
> http://www.geocities.com/jene_if/descargas/aveces_4.zip

Hm -- this raises a tangential question I've been meaning to ask for a
while. Is there somewhere a list of the standard IF commands in
Spanish? I'd be interested in the French, Italian, and German ones as
well, for that matter.

My reading comprehension is a lot better than my active vocabulary in
these languages, so trying to play a non-English language game usually
involves a lot of verb-guessing and eventually giving up in
frustration, even though I can usually make sense of the descriptions
and so on. Obviously I should work on my language skills, but maybe
someone besides me would also find useful, say, a side-by-side chart of
the standard IF commands in each of several languages (which might then
be of use to speakers of other languages who wanted to try English
commands, too).

Anyway, thanks for pointing the game out.

ems...@mindspring.com

unread,
Mar 31, 2006, 1:56:38 PM3/31/06
to

ems...@mindspring.com wrote:
> Obviously I should work on my language skills, but maybe
> someone besides me would also find useful, say, a side-by-side chart of
> the standard IF commands in each of several languages (which might then
> be of use to speakers of other languages who wanted to try English
> commands, too).

Sorry, on re-reading this sounds demanding, which is not what I
intended. I'm curious to know whether such a thing as a comparative
IF-vocabulary list exists already; if not, I'd be happy to collaborate
in producing one, if anyone but me thinks it would be useful.

There.

And thanks again for the recommendation.

steve....@gmail.com

unread,
Mar 31, 2006, 5:17:53 PM3/31/06
to
Malcolm Ryan writes, quoting me:

> > You should know that it is a framework for warping the narrative to
> > guide it towards a satisfying sequence, according to a narrative
> > heuristic provided by the programmer. The machine itself "recognises"
> > drama in an even more limited sense than a chess program recognizes a
> > winning position -- that is, only by evaluating it through and by a
> > simple criterion provided by the programmer. And even state-of-the-art
> > programs do nothing to "generate" drama, but only modify what is
> > already generated by warping the situation in ways also predefined by
> > the programmer.
>
> Yes, and my argument is that we have very poor tools for expressing
> those heuristics so that the results actually line up with what we
> think of as real drama. Sure we can build search algorithms which
> maximise a heuristic, but calling that 'Drama Management' is begging
> the question if you cannot actually write the heurisitc.

This is the correct level on which to attack DM. It doesn't claim to
recognize drama, so don't refute that; it doesn't claim to generate
drama, so don't refute that. It *does* claim that it's possible to
write a narrative-heuristic. If you want to refute DM, I think you have
to tackle it there.

> Like I said, we need to build tools to recognise and represent drama so
> that we can write decent heuristics to guide our search. The search is
> the easy bit. I know Michael Young's group at NSCU are working on
> trying to build models of "suspense" and "conflict", although I haven't
> seen any publcications yet. I can't say I have seen much other work
> like this. Do you kniow of any?

Nothing you haven't, just the well-rehearsed stuff. Drop me an email,
though.

José Manuel García-Patos

unread,
Apr 2, 2006, 3:46:15 PM4/2/06
to

>> This was already done several years ago. The results were satisfactory
>> game-wise, story-wise and even simulationist-wise.
>
> Cool. Do you know how it was implemented?

Nope. But the source code is freely available from here:

http://ifarchive.heanet.ie/if-archive/games/mini-comps/spanish/nanocomp3/n3-aveces.zip.

Just keep in mind that it was written in InformATE, which is the same as
Inform, but with the keywords translated. The email address of the author
must be somewhere in that zip file just in case you have any questions.

> Several of the things on this list have been done before, but don't
> have a sort of standard implementation that makes them easy for people
> to include.

There can't be a standard implementation of any of the things that you
suggested. I'm writing an article about that. Hopefully, I will finish it
in time for next SPAG.

> Hm -- this raises a tangential question I've been meaning to ask for a
> while. Is there somewhere a list of the standard IF commands in
> Spanish? I'd be interested in the French, Italian, and German ones as
> well, for that matter.

I don't know what you mean by IF commands. LOOK, EXAMINE, TALK TO, ASK,
and all that? I never thought of that as standards, but just make a list
of what you consider as such in English and I guess I can give you the
Spanish equivalents. I don't see the usefulness of that, but it wouldn't
take a lot of my time to translate a bunch of words.

ems...@mindspring.com

unread,
Apr 4, 2006, 7:41:44 PM4/4/06
to

José Manuel García-Patos wrote:
> >> This was already done several years ago. The results were satisfactory
> >> game-wise, story-wise and even simulationist-wise.
> >
> > Cool. Do you know how it was implemented?
>
> Nope. But the source code is freely available from here:
>
> http://ifarchive.heanet.ie/if-archive/games/mini-comps/spanish/nanocomp3/n3-aveces.zip.

Okay, thanks.


> > Several of the things on this list have been done before, but don't
> > have a sort of standard implementation that makes them easy for people
> > to include.
>
> There can't be a standard implementation of any of the things that you
> suggested. I'm writing an article about that. Hopefully, I will finish it
> in time for next SPAG.

No, probably not: I phrased this very sloppily. I do not expect there
to be a way to cover these things uniformly in a library, as I've said
before elsewhere. (Other people may disagree with me, but as far as I
understand it, you and I *do* agree.)

On the other hand, I think that some of these may be expressions of a
single problem -- how do you determine descriptions on the basis of a
combination of scalar states within the world model? -- and for that
there may be some algorithmic approaches which might form the
underpinnings of future libraries (plural).

> > Hm -- this raises a tangential question I've been meaning to ask for a
> > while. Is there somewhere a list of the standard IF commands in
> > Spanish? I'd be interested in the French, Italian, and German ones as
> > well, for that matter.
>
> I don't know what you mean by IF commands. LOOK, EXAMINE, TALK TO, ASK,
> and all that? I never thought of that as standards,

In the sense that there are built-in commands in the libraries, it
might be that a basic list could be formed. But it also sounds as
though I'm the only person who wants this, so perhaps the thing for me
to do would be to poke through Italian, Spanish, etc., versions of
Inform and other languages, and extract such a list myself; no need to
make other people help if it's just for my own benefit. But thanks for
the offer.

José Manuel García-Patos

unread,
Apr 5, 2006, 7:11:03 AM4/5/06
to

> No, probably not: I phrased this very sloppily. I do not expect there
> to be a way to cover these things uniformly in a library, as I've said
> before elsewhere.

Where exactly did you say that?

> In the sense that there are built-in commands in the libraries, it might
> be that a basic list could be formed. But it also sounds as though I'm
> the only person who wants this, so perhaps the thing for me to do would
> be to poke through Italian, Spanish, etc., versions of Inform and other
> languages, and extract such a list myself; no need to make other people
> help if it's just for my own benefit. But thanks for the offer.

You're welcome. And good luck with the poking. If they ask me, the only
place where I see such a list could be helpful would be an appendix of the
IF Theory book. I don't know, you might be right on this one, but if you
really think it's useful, you should at least put an example of its use.
Playing games in other languages? What would be the use of the list if you
don't understand the descriptions? And if you do, what do you need the
list for? I just want to understand this.

Sophie Fruehling

unread,
Apr 5, 2006, 7:54:04 AM4/5/06
to
José Manuel García-Patos <jo...@cervantes.org> writes:

>> In the sense that there are built-in commands in the libraries, it might
>> be that a basic list could be formed. But it also sounds as though I'm
>> the only person who wants this, so perhaps the thing for me to do would
>> be to poke through Italian, Spanish, etc., versions of Inform and other
>> languages, and extract such a list myself; no need to make other people
>> help if it's just for my own benefit. But thanks for the offer.
>
>You're welcome. And good luck with the poking. If they ask me, the only
>place where I see such a list could be helpful would be an appendix of the
>IF Theory book. I don't know, you might be right on this one, but if you
>really think it's useful, you should at least put an example of its use.
>Playing games in other languages? What would be the use of the list if you
>don't understand the descriptions? And if you do, what do you need the
>list for?

As someone who has been playing IF in a number of languages, I can
only agree with Emily that such a list would be very helpful. Even if
you understand the description more or less, that doesn't tell you
which verbs exactly you can use. Hell, I've had problems with thinking
of verbs to use in German, which is my native language.

--
Sophie Frühling

The cube tastes like sugar. You are suddenly surrounded by a herd
of moose. They start talking to you about a moose-load of things.

ems...@mindspring.com

unread,
Apr 5, 2006, 12:52:30 PM4/5/06
to

José Manuel García-Patos wrote:
> > No, probably not: I phrased this very sloppily. I do not expect there
> > to be a way to cover these things uniformly in a library, as I've said
> > before elsewhere.
>
> Where exactly did you say that?

Here and follow-up posts:

http://groups.google.com/group/rec.arts.int-fiction/browse_frm/thread/871fb9289259771b/6fdf66a04f3027b7?lnk=st&q=analysis+of+72+newbie+transcripts&rnum=1#6fdf66a04f3027b7

> I don't know, you might be right on this one, but if you
> really think it's useful, you should at least put an example of its use.
> Playing games in other languages? What would be the use of the list if you
> don't understand the descriptions? And if you do, what do you need the
> list for?

Mm, difference between active and passive vocabulary -- most people
understand more words than they know how to use themselves.
Particularly with romance languages, I can guess the meaning of a lot
of words when reading that I would not be able to come up with on my
own.

José Manuel García-Patos

unread,
Apr 5, 2006, 6:54:53 PM4/5/06
to

Could you please be more specific? That was one of the longest threads
that I can remember, and the link above doesn't point to any of your
posts. Just say: numbers such and such, please. Thank you.

> Mm, difference between active and passive vocabulary -- most people
> understand more words than they know how to use themselves.
> Particularly with romance languages, I can guess the meaning of a lot
> of words when reading that I would not be able to come up with on my
> own.

Ok, so you have your list and now you can EXAMINE things in German. Now
what? Could you solve puzzles too?

Don't get me wrong. It's probably a good idea, and sure it's going to be
helpful for a lot of people, but at some point you'll have to get the
dictionary if you want to complete the game (or if you just want to enjoy
it). Also, input is not only verbs, but also objects. And sometimes
objects are not in the descriptions (which is bad practice, in my
opinion). How does your list help you with that? How does it help you with
number, gender, declinations, cases, etc.?

Now, if you had told me: It could be useful for translators (as the
translations would keep a unified and universally accepted set of
commands), that would be a completely different thing.

José Manuel García-Patos

unread,
Apr 5, 2006, 7:07:38 PM4/5/06
to

> As someone who has been playing IF in a number of languages, I can
> only agree with Emily that such a list would be very helpful. Even if
> you understand the description more or less, that doesn't tell you
> which verbs exactly you can use. Hell, I've had problems with thinking
> of verbs to use in German, which is my native language.

Did you have problems with those commands Emily wants to put in the list
or with any other verbs? Was it one of those cases of "exact word
syndrome"? If so, what does it have to do with the issue we were
discussing? I sometimes run out of ideas too in Spanish games,
but a list of "frequently used verbs" is hardly ever of any help. If I run
out of ideas is because I need some weird verb not in the supposed list
of most common ones.

Again, don't let my skepticism prevent you of doing anything you find
useful.

ems...@mindspring.com

unread,
Apr 5, 2006, 7:21:48 PM4/5/06
to

José Manuel García-Patos wrote:
> > Here and follow-up posts:
> >
> > http://groups.google.com/group/rec.arts.int-fiction/browse_frm/thread/871fb9289259771b/6fdf66a04f3027b7?lnk=st&q=analysis+of+72+newbie+transcripts&rnum=1#6fdf66a04f3027b7
>
> Could you please be more specific? That was one of the longest threads
> that I can remember, and the link above doesn't point to any of your
> posts. Just say: numbers such and such, please. Thank you.

72, somewhat, though it's long; then the followup exchange with Steve
Breslin at 106, 107 (and some more comments at 110, 111, 115, etc.)

> Don't get me wrong. It's probably a good idea, and sure it's going to be
> helpful for a lot of people, but at some point you'll have to get the
> dictionary if you want to complete the game (or if you just want to enjoy
> it).

Sure; it's mostly a question of doing a little prep work to make the
basic stuff easier on myself. At least in English language IF, the need
for verbs not in the main verb list is usually not too great. One would
need the dictionary eventually in any case.

> Now, if you had told me: It could be useful for translators (as the
> translations would keep a unified and universally accepted set of
> commands), that would be a completely different thing.

Maybe it could do that too, though unfortunately I do not know of very
many translation projects going on. I'm not opposed to the idea, though.

José Manuel García-Patos

unread,
Apr 5, 2006, 8:03:42 PM4/5/06
to

> 72, somewhat, though it's long; then the followup exchange with Steve
> Breslin at 106, 107 (and some more comments at 110, 111, 115, etc.)

Thanks again.



>> Now, if you had told me: It could be useful for translators (as the
>> translations would keep a unified and universally accepted set of
>> commands), that would be a completely different thing.
>
> Maybe it could do that too, though unfortunately I do not know of very
> many translation projects going on. I'm not opposed to the idea, though.

Nice phraseology. I'm not opposed to the idea. So we have your
blessing? I humbly thank you, well, well, well. By the way, you may not
know of many translation projects going on, but you know of at least one,
and that should suffice.

"And, yes, maybe it could be helpful in translations, but the main goal of
the list is that I can play Italian games I can barely understand! I want
that to be perfectly clear." Ok, whatever Your Majesty says.

ems...@mindspring.com

unread,
Apr 5, 2006, 8:57:28 PM4/5/06
to

José Manuel García-Patos wrote:
> >> Now, if you had told me: It could be useful for translators (as the
> >> translations would keep a unified and universally accepted set of
> >> commands), that would be a completely different thing.
> >
> > Maybe it could do that too, though unfortunately I do not know of very
> > many translation projects going on. I'm not opposed to the idea, though.
>
> Nice phraseology. I'm not opposed to the idea. So we have your
> blessing?

No, you don't have my blessing: I don't give blessings, not being under
the delusion that they're at all meaningful. I occasionally express
support for a project, by way of attempting to cheer on someone else's
hard work which I think worthwhile. But in this case I wasn't doing
either -- I just meant, oh, gosh, something like "that sounds like
another fine use for this hypothetical list, and one which I had not
even thought of." "I'm not opposed..." meant merely that if I make such
a list, I would be happy to share it with anyone who wanted to use that
list for translation purposes; I was not issuing any blanket statement
about translation projects in general.

I did, as you say, know about your project, but creating a style guide
for a "unified and universally accepted set of commands" seems most
meaningful when there are several projects happening at once. I do not
recall hearing about any others under way at the moment, but I'm happy
to be wrong about this.

Anyway, I am sorry I offended you. I'll try to watch my words more
carefully in the future, and, for the record, I do not imagine that
anyone here needs my approval for any reason, unless collaborating with
me on something directly.

José Manuel García-Patos

unread,
Apr 6, 2006, 6:54:53 AM4/6/06
to

> [..] creating a style guide for a "unified and universally accepted set

> of commands" seems most meaningful when there are several projects
> happening at once. I do not recall hearing about any others under way at
> the moment, but I'm happy to be wrong about this.

Well, you realize of course that if there are so few projects is due to
the difficulties of translating not only the text of the game, but also
the program and the parser. And you should realize that anything useful is
still useful even when it isn't used. So, if nobody needs the list today,
who cares? They will need it tomorrow. It is useful even if it is not used.

Nanotechnology, for example, isn't widely used today. But would you say it
isn't useful? Just wait and see.

It works like this: If some people want to do something, but find out it
is too difficult, most of them will not do it. But if this same thing is
made easier, then more people will be able to make it. So, if you help
translators, the number of translation projects will raise. Today there is
no help at all. Well, there you have it: No translations. But that doesn't
mean there are no people wanting to translate games. Ergo, any help is
welcome.

> Anyway, I am sorry I offended you. I'll try to watch my words more

> carefully in the future...

You should always keep in mind that a particular need (your playing games
written in foreign languages) is almost never more important than a
general need (translation of those same games). That's what I criticized.
That you seemed to imply that your own need was more important than a
general, objective need. (And now you'll say: "No, I didn't mean that. I
phrased it sloppily.") It's ok. I wasn't offended. Just think of other
people when you write. The fact that you don't care about translations
doesn't mean everyone agrees with you.

solar penguin

unread,
Apr 6, 2006, 7:26:17 AM4/6/06
to
José Manuel García-Patos <jo...@cervantes.org> wrote:

>
> You should always keep in mind that a particular need (your playing
> games written in foreign languages) is almost never more important
> than a general need (translation of those same games).

Surely the list would be useful for _everyone_ who wants to play games
written in foreign languages, not just for Emily. Wouldn't that make it
a general need?

So, when are you going to post your list here? Or are you really going
to withhold from all of us, just because you've had an argument with
her?

--
___ _ ___ _
/ __| ___ | | __ _ _ _ | _ \ ___ _ _ __ _ _ _ (_) _ _
\__ \/ _ \| |/ _` || '_| | _// -_)| ' \ / _` || || || || ' \
|___/\___/|_|\__,_||_| |_| \___||_||_|\__, | \_,_||_||_||_|
|___/
http://www.freewebs.com/solar_penguin/

** This is base one. Keep half your group out of a musical instrument.

** Elements will come across more than a computer program.


José Manuel García-Patos

unread,
Apr 6, 2006, 7:54:53 AM4/6/06
to

> Surely the list would be useful for _everyone_ who wants to play games
> written in foreign languages, not just for Emily. Wouldn't that make it
> a general need?

She said she was the only one interested. Sophie expressed interest too,
but at this moment it is unclear to me if they were talking about the same
thing. That's why I asked Sophie about it.



> So, when are you going to post your list here? Or are you really going
> to withhold from all of us, just because you've had an argument with her?

I don't have a list. And I didn't have an argument with her. Also, Emily
didn't want me to help her in the making of the list. So, why should I?
Remember what I said? If it had any use at all it would be in
translations. I was never interested in it. I just suggested an objective
reason why it should be done, which I think is better, more useful,
helpful and general than the one she gave. That's all.

Do you have a list? Maybe she'll take your help. Do you want to play games
in foreign languages? Do you want to translate games? None of the above?

solar penguin

unread,
Apr 6, 2006, 8:57:53 AM4/6/06
to
José Manuel García-Patos <jo...@cervantes.org> wrote:

>
> She said she was the only one interested.

I thought she only meant she didn't know if anyone else was interested.
I know I'd be interested in it.

>
> And I didn't have an argument with her.

"That's not an argument, it's just contradiction."

> Do you want to play games in foreign languages?

Yes. It might help me to _finally_ learn a foreign language. I'm
getting so desperate that I'm willing to try almost anything, no matter
how unorthodox. Whenever I've tried to learn one in the past, I've
always given up when I couldn't understand those fiddly little words
that don't really mean anything.

For example, the English word "put". The phrase "She put the lamp on,"
refers to her lighting the lamp unless it's part of a sentence like "She
put the lamp on the table." Foreign languages use totally different
verbs for these variations of the concept of putting-on. That's what
makes them so confusing. (It gets even worse when she starts putting on
an accent or some clothes!)

So even if I do remember that a foregin verb has something to do with
putting-on, I then have to remember which type of putting-on goes with
that verb. And that means having to remember how the foreginers divide
up the concept of putting-on in the first place. So, it's not
surprising that I'll probably guess the wrong one and say something that
literally means "She lit the lamp onto the table," or "She placed the
lamp alight," or something equally meaningless.

There are similar problems with all the other fiddly little words too.
I can never remember them all. At least making mistakes with a game
would be less embarrasing than making them with a person, and would give
me a chance to practice until those words finally start sinking in.


> Do you want to translate games?

See above.

--
___ _ ___ _
/ __| ___ | | __ _ _ _ | _ \ ___ _ _ __ _ _ _ (_) _ _
\__ \/ _ \| |/ _` || '_| | _// -_)| ' \ / _` || || || || ' \
|___/\___/|_|\__,_||_| |_| \___||_||_|\__, | \_,_||_||_||_|
|___/
http://www.freewebs.com/solar_penguin/

** I'll miss the sea... But a person needs new experiences... They
jar something deep inside, allowing him to make me a report when we're
finished.


Sophie Fruehling

unread,
Apr 6, 2006, 10:24:03 AM4/6/06
to
José Manuel García-Patos <jo...@cervantes.org> writes:

>
>> As someone who has been playing IF in a number of languages, I can
>> only agree with Emily that such a list would be very helpful. Even if
>> you understand the description more or less, that doesn't tell you
>> which verbs exactly you can use. Hell, I've had problems with thinking
>> of verbs to use in German, which is my native language.
>
>Did you have problems with those commands Emily wants to put in the list
>or with any other verbs?

I'm talking about standard IF verbs, and I think Emily has been
talking about these, too. OTOH, I can't see why such a list would be
useful for translating, especially if you tranlate a game from, say,
informate to some other version of inform, since these have the
same standard verbs.

One example: it took me quite a while to figure out how to put
X ME in Spanish. Okay, this is more than the verb, but such a thing
should be put in the list, too, I guess.

Sophie Fruehling

unread,
Apr 6, 2006, 10:28:12 AM4/6/06
to
Sophie Fruehling <sfrue...@LOVELY-SPAM.aon.at.invalid> writes:

>OTOH, I can't see why such a list would be useful for translating,

[...]

On the other other hand, I think I can think of a use after all. :)

Rikard Peterson

unread,
Apr 6, 2006, 1:45:19 PM4/6/06
to

"Sophie Fruehling" wrote in
news:olb732hcakhamscro...@4ax.com:

> I've had problems with thinking of verbs to use in German,
> which is my native language.

I can say the same about Swedish.

Rikard

José Manuel García-Patos

unread,
Apr 6, 2006, 8:14:24 PM4/6/06
to

> One example: it took me quite a while to figure out how to put X ME in
> Spanish. Okay, this is more than the verb, but such a thing should be
> put in the list, too, I guess.

Any game with a help command says how to do such things. You speak
Spanish, Sophie, so you know the difference between EXAMINAME and EXAMINA
OBJETO. It's the same verb, the same action, but not the same word. Don't
you think it's confusing for a foreign player? Does a list solve these
cases? How long should that list be to cover all of the exceptions?
Consider this: RUFEN SIE AN. The verb here is ANRUFEN. Do you think that
someone who knows nothing about German would know that from a list? I, for
one, don't think so. But again, I'm not here to stop any of you from doing
what you feel like. If you think this is useful for playing, go ahead.
Never mind me. My opinion -- and hopefully my last word on the subject --
is that if you do that list, you won't be wasting your time, but you'll be
doing it for the wrong reason. It won't help you play games unless you
already have a pretty good knowledge of the language. And I know, because
I used to do that in the eighties. I didn't know enough English and I had
to play with a dictionary and a very limited "active" and "passive"
vocabulary. I never even finished a game.

This answers solar penguin too, so I think I'll take a rest from
discussions now.

Benjamin Caplan

unread,
Apr 26, 2006, 8:34:25 PM4/26/06
to
>UNTOPIC

I think that to form more analog world models, it is valuable to look
at mainstream gaming. (I find this ironic because I always thought of
IF as the world leaders in game design -- while everyone else was
worrying about sound fidelity and frame rates, Graham Nelson was
writing _The Craft of Adventure_.) RPG battles are completely
rule-based, sometimes to very good effect. FPSs allow arbitrarily
sophisticated tactics because the player can place the protagonist in
any of thousands of subtly different positions on the battlefield.
But I think that RTSs and empire sims are the closest to what we're
looking for: essentially object-oriented, exhaustively rule-based, and
arbitrarily complex. In addition, they aren't as closely tied to battle
as the other genres; Starcraft might make a fairly good model for a
chemistry puzzle, for instance.
What do y'all think?

John Bruce

unread,
Apr 27, 2006, 6:02:36 AM4/27/06
to

Interesting suggestion. Though I'm afraid it doesn't really tie back
to IF, I do know a group that's doing something not unlike your final
suggestion. They're attempting to create a game focused on the various
roles of the internal organelles of a human cell, and the model is one
that shares more than a passing similarity with Starcraft, at least in
the sense you're discussing.

--
With All Due Respect,
John Bruce

Benjamin Caplan

unread,
Apr 27, 2006, 10:03:51 AM4/27/06
to
I don't see why it shouldn't tie back. _Metamorphoses_ showed that
rule-based systems can work for IF, and _One Week_ did
resource-management problems. Neither of these strikes me as being
terribly difficult to reproduce, once you know it's there.

David Fisher

unread,
Apr 27, 2006, 6:37:06 PM4/27/06
to
"John Bruce" <aras...@gmail.com> wrote in message
news:1146132156....@t31g2000cwb.googlegroups.com...

> Though I'm afraid it doesn't really tie back
> to IF, I do know a group that's

> attempting to create a game focused on the various
> roles of the internal organelles of a human cell

OrganelleComp !

Transcript exerpt:

A worried looking vesicle approaches you. "The mitochondria are
restless, sir ... word from the nucleus is that an uprising is imminent."

Through the cytoplasm you can make out the forms of several rogue
mitochondria. Distant drum beats sound ominously from the far side
of the cell. The endoplasmic reticulum waves gently in time with the
beats, like seaweed at the bottom of the ocean floor.

> x reticulum

Which reticulum do you mean, the smooth endoplasmic reticulum or
the rough endoplasmic reticulum ?

David Fisher


0 new messages