Rick Smith's Reactions

Once upon time, some people went to Phrontisterion 4...

Held on June 26 to 27 there were 19 people in attendance. The group was approximately evenly split between the techies, and the artists / designers / writers.

Three people in attendance had written programs to allow Interactive Fiction (I.F. hereafter). They were Brian Magerko, Jeff Rawlings & Chris Crawford, so the talk was considerably less theoretical than in previous years.

The talk was a round table discussion with everyone asking questions or making comments as they chose.

Definition of I.F.: Interactive Fiction (I.F.) is not computer games & is not Interactive Storytelling (that term has been used by the people making text based puzzle games). It is used here to mean writing programs to tell stories based on the actions of one or more users of the program. They may be player characters (PC’s) or a more third person fate like being. It has proved to be very difficult to do this and the field lacks strong examples of what this entertainment form will eventually become.

Chris has finished his new book on I.F. & had sent out a galley version of it to the attendees. Most of us had read it before the conference began. The talk was structured around the chapters of his book, until mid Sunday, when we started discussing other books on the subject. This write up will hit the high lights (in my humble opinion) of the two days of talk.

We began by talking about stories & the characters in them. The point was made early, that the people in games were little more than objects, and the ways that they interacted with each other are typically both simple and violent. Interesting drama can not be made when the only verbs allowing characters to interact are ’shoot’. Generally stories are about people and games are about goals.

A point was made that I found very interesting: if the animation is very cartoony people accept it for what it is. Animation that is very close to photo realistic causes people to notice the tiny flaws in the animation. With artificial personalities the same thing may occur: if a artificial personality is close to how we expect people to react & is _almost_ right then the character feels monstrous. This emphasizes the difficulty of making software that can create and control interesting characters. The importance of emotions to drive behavior in software characters was emphasized. The game Ico (where you must carry a person with you who is unable to walk on their own) was given as an example of a game with stronger than normal characters. Gordon pointed out that the thing that was really important was the mental state of the player of our story worlds. If the user of our I.F. is enjoying themselves and thinks that a cool story is going on, then the job is done. A good point, (which partly explains the popularity of the Sims I believe) but something has to be there first to get the user’s imagination going. The discussion of how important graphics and sound was to I.F. It was pointed out that MUDS had tens of thousands of players, but graphical MUDS like Everquest and Origin Online have millions of players. (Are they really up to millions of regular players?) The first Interactive Fictions may have text based interfaces while we are figuring out the art form, but they must get graphics before they will become a medium of mass communication. Chris brought up his thesis that space is of limited use in drama; all needed movement can be done on various stages with out any relationship between them in Cartesian coordinates. A few counter examples were offered but generally the group agreed with his point.

We then discussed Machinima. I had never heard of this but apparently it is a type of story telling using interactive level editors (such as Unreal Editor or Macromedia Director) to create characters and stages. People move on these small levels and dialog is added to make stories. Thus game tools are being used to make small movie like entertainment. This is not interactive and we moved on.

Chris pointed out that we needed LOTS of verbs for interesting stories. (In his experiments he found that more than 2,000 verbs were needed to allow interesting drama to happen.) This is one of the major difficulties with the art form.

Consider various user interfaces in computers:

Joystick with several buttons: ~Max 24 verbs. Command Line interface: ~Max 50 verbs. * GUI interfaces: ~Max 300 to 400 verbs. * Language: Many thousand verbs.

* A command line interface or a GUI can have a theoretical limit of nearly an infinite number of verbs, but in practice user memory places a practical limit on how many commands are allowed. Power users who are highly expert on this can push this higher of course.

Chris thus feels that a simplified language is needed. There was much discussion on this. It turns out that evidence suggests that a class of languages know as creoles are extraordinarily easy for humans to learn, and they may be the based on the fundamental language circuitry of the brain. Very interesting stuff, see Chris’ writing for more details. Chris suggests I.F. should use an artificial language based on creoles.

This caused a lot of argument: people would not want to learn a language to play a game. Chris offered the counter example of Siboot. In that game, people did learn the small language very quickly and easily. Gordon pointed out that if people think that they have to learn a _language_ to play a game, they just won’t buy it. True, but if it is perceived to be a subset of English, I think that people won’t obsess about it.

People pointed out that we might be able to get away with fewer verbs and lots of adverbs. This is true but they still must be learned by the users and still must be programmed by the programmers so I was a bit unsure how much this would gain us.

Next we talked about I.F. itself. First, would it be popular? Laura pointed out that the mediums which have stories (books, TV and movies) are multi-billion dollar industries. That suggests that people are willing to pay serious money for stories. Interactive entertainment (computer games) are also a multi-billion dollar industry even tho it reaches a very limited market of hobbyists. (Game industry propaganda aside, most games are still played by young males. There are very few women are much interested in the violence in games.) The Sims is a non-violent game that significant number of women bought and it is now the most successful computer game in history (in terms of sales). However the industry has not been able to repeat the success of The Sims in other games.

Anyway, Laura suggests that I.F. WILL appeal to a mass market including women and there is no reason it could not someday become bigger than movies (as games have already done) or traditional computer games. I strongly believe in the power of interactivity and agree with her. Laura has made a graph showing the relationships between various entertainment media which graphically suggests the size of the market for I.F. I believe that this graph will be posted on the BLOG that will be formed for this group.

Also interactivity is the competitive advantage for I.F. If you want a non-interactive entertainment (multi-media for example) well, TV or movies will do it better. Only with interactivity does the computer really shine. Story telling on computers must use interactivity or it will be done better and cheaper else where.

Using I.F. where all characters are Player Characters (PC’s) was brought up. Using the computer as a communication device to bring people together in an I.F. (everyone is an actor) is a possibility. I have my doubts, most interesting stories that people buy are created by a single artist, not by a collection of people. (However, if it is an interesting enough virtual reality the users may be willing to pay for it.)

The point was made that trying to tack a story onto a game sucks, you do not get software that tells stories, at best you get some context for the violence in the game. As long as games have verb sets of 40 or fewer verbs, you CAN NOT have drama. If you are to have good stories in games two things must happen: first the writer and game designer must be working together from the very start as equal partners. Second, the game designer can NOT say: the only verbs you can have are jumping, moving, shooting and picking up ammo. If that sort of constraint is put on the writer, he or she has been in fact demoted.

It was pointed out that those interactive entertainments that have poorish content that have succeeded are those that allow people to SHARE the content. People put their home pictures on websites, they post blogs and tells stories of their Sims people. The ability for people to share or talk about their experiences can leverage the success of Interactive Fiction (I.F.). Thus we may wish to consider how to make it easy for people to share stories about what the I.F. did. Perhaps a replay function allows people to recreate an especially interesting play thru of a story world and let their friends see how they did would be useful.

A discussion was made of the ’two cultures’. The way techies often communicate alienates and puts off non-techies. Many artists make little or no effort to understand math or science. I.F. will require story world builders who are both techies who have an artistic bent. About 3% of the people producing salable artistic content have this ability to span both worlds. Some discussion was made on if people can train to be able to think both ways. I believe that this can be done, but it is difficult. Programmers, read widely: all SORTS of different books. Do try writing for ordinary people, on non-technical subjects. (Fiction or non-fiction.) Artists, learn some math and programming.

There was a short discussion on forms of thinking. Noun based verses verb based thinking. (For example the English language is a noun powerful language where as Latin or Hopi put more emphasis on their verbs.) Building story worlds will require a verb-centric way of thinking which comes hard to some people. (Programming requires verb based reasoning which is an additional reason for artists in any form of interactive entertainment to practice programming.)

Also intuitive and pattern based reasoning was compared to sequential based reasoning. Both will be needed for I.F. (with a bit of fuzzy logic IMHO).

Current styles of building story lines were discussed. A branching tree is unworkable. (For example, a short story I analyzed had about 100 decision points with often several decisions possible at each point.) Assuming a minimum choice of a yes / no decision, that is 2 to the 100th power of story elements to be created. (Trillions of trillions of elements.) So in practice story trees are tightly constrained. One way of trimming this impossibly huge tree is to ’kill them if they stray’. If they deviate off the main story, they die. This is basically a linear story along with the mechanic of going back to the last save point. This awkward in my opinion; it breaks the suspension of disbelief. Fold back gives the person a choice but very soon both choices come back to the same node. This is irritating to people as they soon find that their choices are not meaningful. A linear story with state variables can allow you to pick a cluster of different endings at the end of your linear story. This is the strategy used by the Japanese Comic Book games. I’ve played a couple of these, and it is irritating having so little control of what is going on. Most of your choices do not affect anything meaningful and having to make all of these choices that don’t accomplish anything grated on me.

(Of course ALL of these strategies can be used a LITTLE bit when appropriate, with out damaging the story. But using them enough to trim the impossible sized story tree spoils the interactivity except for the simplest stories.)

Chris suggests a story web as a solution to the above problem. It is discussed in his book on interactivity.

A discussion of what forms of interactive entertainment are there now. There are role playing games, improv, heckling in comedy clubs, storytelling to kids and the play "Tony and Tina’s Wedding". A number of problems were discussed in the play Tony and Tina’s Wedding. First it is VERY expensive, and the audience is relegated to low quality interactivity. They are not at the center of the drama, the Non-Player Characters (NPC’s) Tony and Tina and their family are. There was some discussion of Neverwinter nights which allows a GM to make a graphical rpg. Good idea but it requires so much work it has limited the adoption of the game so far.

Emergent stories were discussed. Create an interesting environment, & maybe good stories could be told AFTER the fact. The human mind is good at making stories out of pretty flimsy material. This has been called the ’man in the moon’ effect. (The brain finds order even if there is not any.) This can work but we are looking for something which gives the story world writer a bit more ability to discuss important themes. Good stories communicate something. Also, we have lots of stuff now that people can do and tell stories about it afterwards. (Stories of battles in Total Annihilation for example.) But even tho websites with such fiction exist, does not make Total Annihilation an example of an Interactive Fiction software project.

Personality Modeling is used to help drive and distinguish the behavior of the Non-Player Characters (NPC’s). Standard personality models created by psychologist and the way D&D models personality were discussed. The feeling was that render unto D&D that which is D&D’s. Basically the models that have been built by others are optimized for the use that the designers of these models need. The Lawful - Chaotic / Good - Evil model from D&D is designed to help the players of that game know how they should run their characters The models designed by psychologists are designed to help people or understand people. However we want a personality model that will help our NPC’s make decisions. In other words, custom design a model that will drive the behaviors that you wish to model in your I.F.

Some time was spent on the discussion of values for personality models. Should the personality attributes be orthogonal or more ’realistic’ where you might have several values that say similar things? I feel that the personality attributes should have as little overlap as possible. Basically we are already taking on a very difficult task, by making the different values independent, we simplify our job some.

Character growth in stories is very important and we would like to capture this with personality models. I don’t see any major problems, the data structures just need to be set up with some care.

A drama manager is a software system that watches what is going on in the story & ’helps’ it along if the story seems to becoming boring. Kinda like a game master who throws a wandering monster at his PC’s anytime they talk for more than 5 minutes.

To do a drama manager well is a very difficult task, but a few of the people working on I.F. have been attempting it. Jeff’s system (if I understand what he said) will watch the player character and try to deduce what the motivations or goals of that PC is. When the drama manager knows this, it can place dramatic delays and problems in the player’s way that will correspond to the ebb & flow of the story.

One idea I thought was very clever was using a drama manager to rate the players experience on an ’applause meter’. This idea came under sharp criticism as it turned interactive fiction into something more like a game & it constrained the person’s choices with values the drama manager used to score the drama.

I agree with the above, but...

The I.F. art form is so new I feel uneasy about closing off avenues of approach. Also it is a dream for many people to be an actor or actress who gets a thunderous round of applause after a performance. If our software wants to give people that experience we pretty much need the drama meter to gage the actions of the PC’s.

Finally if the users of the system know that they will get ’high marks’ for choosing interesting or dramatic behavior (rather than the safe or cautious behavior) then they are more likely to be dramatic. If everyone does this, they are more likely to have fun, which is the whole purpose of all this after all.

At this stage we began looking into the implementation details of Chris’ Erasmatron engine. Pick up his book for details, as he says it much better than I can cover in this short summary.

Tho it is very difficult, having the NPC’s and PC’s trade information is very important. Lies also drive a huge amount of drama. So we would like our software actors to be able spy on each other, gossip, blackmail each other, trade secrets, etc.

In Erasmatron version one, the PC spent so much time dealing with minor, boring gossip it hurt the fun of the story world. In the current version (version 4) that Chris is working onthis is automated, but the player loses control of the what to reveal. It is a tough user interface problem as we would like them to have the choice of what secrets to reveal (dramatically important) but don’t want to have the player bogged down with boring discussions (dramatically important to avoid).

I suggested making custom verbs to ’socialize’ (where boring gossip is automatically shared) and a ’reveal secrets’ verb which allow the player to reveal important information. (Since Chris already has code to rate the importance of information this should be doable.)

Anticipation is also dramatically important. For example, Lancalot beds Gueniver. He likes to gossip with his close friends, he likes to share information that is important to them & he likes to talk about information that is important to him. So he rushes at once to Arthur to tell him...

There are two ways to handle this. First write recursive routines that will look at how Arthur will react before telling him, the second is write some sort of dramatic inference engine which will weight the dramatic values of situations. Both are tough problems but Chris feels that the recursive strategy would be simpler to implement.

We discussed ’roles’ to place more context around social situations. Depending on how you approach the problem of implementing I.F. roles can be fairly straight forward or be a really complicated, driving force in the design. Kyle talked about a system he was thinking about where he used roles as a central means of driving drama.

We discussed sequencing where one event drives another event and so on. Chris uses a diary of events to keep track of this information.

Chris spent some time emphasizing the importance of development environments for building story worlds in I.F. He says that at least 2 to 3 times more effort must be spent on making the tools for artists easy for them to use over (the big job) of making the story engine itself.

We then did a review of other work that other people have done Most was of only tangential interest to I.F.

We discussed Facade’ where you are a guest at a dinner party given by a 30’s old married couple. You discover that their marriage is on the rocks, and in fact it may disintegrate while you are at the party. Depending on what you do you may hasten the break up or buy the couple some time.

Facade’ is very tightly constrained and is pretty much hardwired for this one situation, but it is still a huge step forward. Anyone interested in interactive fiction should take a look at it.

Grand Text Auto is a BLOG that talks about stuff of interest to people in the I.F. field.

Brian talked about his technology. It tries to figure out the goals of PC’s and uses this to build the story. He puts down plot points that must occur, but the player character is given wide latitude between these plot points. The major limitation is that he does not use language, communication is done via physical interaction (a stylized body language?) and you can not interact with the NPC’s. You are a ghost who observes the interactions of the PC’s and can affect things a bit by revealing your self at various times.

Brian said that from what he learned in the conference that perhaps language was not as tough as he thought, and may put in a simple system in his next version!

Jeff’s story world engine has a number of interconnected stories in it. It will watch the user and try to pick which story line is the most appropriate at this moment. Thus a fair bit of what is going on is moving form one story line to another. Story lines can be suspended and then later picked up again.

If I understand it correctly, the major work is in the drama engine that tries to deduce the user’s goals and pick an appropriate story for the PC. I would like to see more on this, maybe next year he would have a demo to show us?

I am probably making some mistakes in describing Brian’s & Jeff’s technology, the two did not talk about their work for as long as I would like. I confess that I am a bit vague on the details. (Chris has an advantage over them, as Chris’ ideas are written down in a book that we had some weeks to study ahead of time.)

The remaining time in the con was spent in general discussion of what ever subjects interested people. Andrew Glassner wrote a book saying that I.F. can not be done. (However if you do not use a narrow definition of plot but instead use a meta-plot where a plot arises as people explore a story space his argument is largely undermined.

Understanding Comics by Scott McCloud was spoken highly of to anyone who does creative work. I.F. will likely borrow some ideas from comics so this will be especially valuable to us.

A discussion of marketing I.F. was hotly debated. This feels premature to me (but that didn’t stop me from putting in my two cents worth!)

Syberia and Syberia 2 were discussed briefly. They are games where nothing blows up. They are produced by Similtronics.

We decided to make a BLOG that would allow us to discuss the subjects further. Watch Chris’ website for a link to it.

Chris is making Erasmatron 4 object oriented so that the artist is given a working story world (tho boring) at the start, and progressively modifies it. I think that this is a huge step forward, which will make the technology much easier to for people to explore.

And that was it, I had a lot of driving to do so was one of the earlier people to leave the con. I enjoyed Phrontisterion 4 more than any of the others I attended. It was more practical and I learned a fair bit.

...and everyone lived happily ever after.