Archive for August, 2005

Notes on Façade

Thursday, August 4th, 2005

After having played Façade a dozen times, I feel that I have enough experience with the product to put together a piece of text with my impressions and thoughts.

First of all, I really encourage everyone interested in game design and narrative to try the game before reading on. It is a free download (of hefty 800Mb) for the PC…

When starting up the product, I was appalled that I had to wait almost a minute for my 2.8 Ghz, 1Gb Xeon machine to fire up the game. Apparently the startup does an enormous amount of data-crunching and I am very curious what that crunching consists of. Anyways, the game begins with a short audio sequence, then I am asked for my name and off I go.

The first time I played through the drama, I was truly spooked at several occasions. I really felt deeply for these characters and wanted to help them. It ended with me being thrown out of the apartment. I felt almost devastated and immediately started over again. The following few playthroughs were not that pleasant. Suddenly, it was apparent that I was experiencing scripted “blocks” or sequences glued together by some underlying machinery. My hopes for the product fell, but I continued to play and gradually started to “abuse” the game which was great fun and I found many interesting details and parts of the game that I haven’t experienced during the first few playthroughs.

So, what differs Façade from a branching-tree story game? I believe that question can be answered from several points of view. First of all, from a player’s perspective I am sad to say that there is not very much difference. Sure, Trip and Grace has facial expressions and body language to accompany their utterances. There is also a sense that the characters have some level of mindstate, something that is not necessarily a hard thing to implement technically in a branching-tree game (although very hard to author). So, given the time the authors has spent on Façade, I believe that the same result could be accomplished, perhaps even in less time, with pre-scripted and branching-tree solutions.

The reason I as a player don’t see the cool new AI in action is that the game has serious problems with interaction. There are clearly predefined “moments of interaction” in the script, where I am free to influence the drama. These moments run real-time so they are easily missed. Also, there are problems with feedback so I had a hard time judging if my input actually had been noticed. Many times, I felt like shouting at walls (for example, if you get Grace to storm out to the kitchen and follows her there, she is suddenly immune to kisses. No reaction, nil, void, zip.) Sometimes this might be a bug, but it happens so frequently I believe it is a fundamental design choice.

However, from a developer’s standpoint the difference between the advanced AI approach and the branching tree might be enormous. Under the hood of Façade is a custom behaviour language (ABL), a drama manager, a NLP system and other stuff that hopefully aids the author in creating interactive drama. I have not looked into the technology side of Façade, nor authored for it, so I cannot judge whether the tools and engine actually serves its purpose. Future products based on the Façade engine and toolset have a lot to prove.

There are other serious problems with Façade. Some things I have already mentioned, but sadly there is more. For example, the problem of “addressing input” is painfully apparent. I tend to comment on something Grace says while Trip rambles on, then he interprets my comment as directed to him. Also, it does not help that the game continuously misinterprets statements, often to the opposite of what was intended. Chris Crawford has been advocating symbolic interaction-languages and I definitely think he has a point. Façade makes a bold go at allowing any type of input, but it leaves me with many questions. What is really understood? The engine misinterprets things frequently and there is no telling what parts of what I tried to say were understood. Did the game stumble on the word “rhino” or was it the use of negation that messed things up? Unfortunately, this boils down to a serious feedback-problem. Trip and Grace never once told me “What in God’s name are you blabbering about?” or anything along those lines. They just silently looked at me, adding to my frustration. Going as far as using a specially designed symbolic language as Crawford suggests is perhaps to go too far, but input limitations and a clear focus on verbs might solve some of these problems. I believe the game would benefit if it was able to answer these questions: what am I allowed to do? What can I talk about and in what way?

When making Riddick, we talked a lot about AI and what features of AI that is important. Our discussions boiled down to the idea that the state of an AI is meaningless if it does not communicate what it is doing, feeling or is up to very clearly to the player. For example, a guard who is on alert must show that to the player (this is why the AIs of Metal Gear emits the famous exclamation marks). If you design an AI that could be sad and cry and run away scared and all those other nifty things, you’d better make sure all those moods and states are communicated to the player. Otherwise you are just spending time on things that will only confuse the player and make the experience frustrating.

So, how does Façade manage in this respect? Frankly, I do not know. I know that I can see if Trip is angry or if Grace likes it when I flirt with her and that is very cool. However, the game is so extremely resource hungry that I suspect there is a lot more going on in those AIs that I cannot ever experience. For example, Trip can start shouting “Uhh, you’re driving me insane!” without being provoked (at least I did not understand what suddenly drove him mad). So, again we have some problems with feedback.

Now I have gone on and on about the flawed aspects of Façade. I know that there are people who believe that the game is a milestone in gaming history. Yes, it is an interesting, bold and enjoyable experiment but what the engine and the ideas behind it is capable of seems yet to be fully revealed.