Skip to main content

Dialogue systems in KOTOR and beyond

Some basic design questions

Designing a game means solving a problem with many, many variables. It invariably involves making tradeoffs. The hard part is making those tradeoffs well. To do this, you need to have your priorities straight.

I've been playing BioWare's Knights of the Old Republic (aka KOTOR) the last few days, and while thinking about what I didn't like about it, I came up with a couple of questions (far from a complete set) which are pertinent to ask while designing a game.

  1. What kind of choices do you want the player to be able to make? What kind of freedom can you, as the designer, handle? Where do you offer true interactivity (to use Mark Barrett's definition: choice that affects, as opposed to merely revealing, outcome), and where do you use magicianship, to use a charming term coined by Lee Sheldon?

  2. How do you convey to the player which choices will affect gameplay and which ones won't? Or rather, how do you manage the player's expectations? If selecting skin color of the player character will affect gameplay, the player needs to know this. Similarly, if insulting a non-player character has no real impact on the game, but the player has to play 20 hours to find out, he or she may be disappointed, to the point of quitting the game.

  3. What is the information the player needs to make the choices you want him or her to make? This is, basically, user interface design.

  4. How do you present that information so it does not destroy suspension of disbelief? This is a basic struggle unique to interactive entertainment: the conflict between making a usable application and a compelling experience. It is quite easy to design a user interface that allows for very easy control, but at the same time robs the game of all its charm.

The answer to these questions depends entirely on the kind of game you're making, on the goals you're trying to achieve with that game. KOTOR's designers have answered these questions differently for different parts of the game.

Dialogue and non-dialogue in Knights of the Old Republic

In pretty much all of it's gameplay systems except dialogue, it freely mentions the underlying game mechanics, even referring to "rolls". So a description of the stealth device will mention that enemies will have to make an awareness roll to determine if they can see you or not. I don't like this: I think it destroys suspension of disbelief, it is not really relevant information (not to this level of detail anyway), and in a sci-fi setting it is so easy to come up with a better description that fits the game's fiction (see Deus Ex for a good example).

But while most of the gameplay systems in KOTOR are very transparent, in the sense of exposing the underlying mechanics, the dialogue system is almost completely opaque. Dialogue options are presented as phrases. This is very classical and non-controversial, and it reinforces suspension of disbelief. But I think it has its downsides. These systems tend be non-systemic. In other words, the available choices for any particular situation are largely determined by a game designer, as opposed to being governed by the rules of a simulation. It is usually hard to determine exactly what effect a given phrase will have. The game designer has to clearly convey to the player what actions and intentions his or her character will express with a given phrase. But this requires heavily on interpretation of text, without the help of intonation or other additional information. You can try to solve this with very precise writing, and probably some amount of exaggeration. But precision and exaggeration will affect the tone of the phrase. How can you offer the player the choice to subtly allude to something if all phrases on offer have to state much too clearly what they mean? How can you guarantee the quality and consistency of tone over many thousands of these lines?

In KOTOR and in other games that use dialogue systems of this kind, I constantly find myself translating the phrases into actions, and then choosing based on the action. But because of the problem mentioned before, it is not easy to determine which action will be executed by a given phrase. And once my character has uttered the selected phrase, silently in the case of KOTOR, it is not easy to determine the effect of my action from the reaction of my dialogue partner (although the intonation and animation make it a bit easier).

What I would like to see, and have tried to work on in the past, is a dialogue system which is action-based instead of phrase-based, and which is much more systemic and clear. It will not have the layer of illusion that the phrases provide, which will put a stress on suspension of disbelief. But a coherent system in itself is a big enabler of suspension of disbelief. (Finding out how to simulate in an entertaining way, how to design a system that allows players to make interesting choices while maintaining suspension of disbelief, is one of the central challenges of game design.)

Captain Blood had such a dialogue system. So had Chris Crawford's Siboot: Trust and Betrayal, probably one of the more ambitious, if obscure, attempts along these lines. And so did The Sims. The Sims is a very good example of the pros and cons of a "dialogue" system that is very systemic and very transparent. The conversations in The Sims are coherent and rich in interesting choices, but do not deal with sophisticated subjects and are not presented in a sophisticated way. Yet many, many people believed in and enjoyed these conversations.

Of course, in a game that tries to tell an interesting story involving slightly more sophisticated creatures than Sims, this dialogue system would stick out like a sore thumb. That is why most games carefully constrain the possible situations and choices, and try to keep the player from being bothered by these constraints. This is what Lee Sheldon refers to as magicianship. It's a good term, seeing how it invokes distraction, and subtly persuading people to play along, because it will entertain them.

KOTOR's dialogues have some systemic aspects. There is a neat persuasion system which allows you to try and persuade people, the success of which depends on the active characters persuasion skill. Persuasion phrases are preceded by the word "[persuasion]", indicating something about the action behind the phrase. Some phrases serve to retrieve information and one can detect a system behind them: the "I'd like to ask you a question" phrase which opens the topic selection submenu, then the "What can you tell me about blah" phrases. (Something which we did in Albion, and a bit better in my humble opinion.) But despite these few elements, KOTOR makes me just as frustrated as other games that emply phrase-based dialogue systems.

I may appear to be contradicting myself by criticizing KOTOR for being too transparent in its non-dialogue gameplay systems, and too opaque in its dialogue system. However, it is important to make a distinction between exposing game mechanics - making information available to the player - and dressing them up - presenting them in a way which does not clash too much with the game's fiction. KOTOR exposes its rules and values outside of dialogues, but doesn't dress them up - in fact, it goes out of its way to make references to elements which are neither part of the game's fiction nor even of the underlying code i.e. die rolls. (I know, they may well actually simulate die rolls inside the code, and this probably doesn't annoy the hard-core AD&D players who are BioWare's core audience as much as it does me. But still.) In the dialogue system, KOTOR does not give me the information I want as a player to make the choices I want to make. This is more forgiveable because nobody, including me with my big mouth, have done much better. But what disappoints me is that nobody has done better, or has even tried, for the last, what? ten years? If not more...

New directions in dialogue systems

What bothers me about current games is that interpersonal interaction has not significantly evolved for well over a decade. Like water running towards the lowest place, we have found a local optimum and we've stuck by it.

I don't have the answer, but the question remains fascinating to me, and I know I'm not alone. Apart from the action-based interface described above, I think there are some other promising directions:

  • The diplomacy systems that can be found in strategy games. Diplomacy is a way to model interpersonal interaction. Strategy games tend to do more simulation and offer more true interactivity than action-adventure games, but make fewer attempts to present in-game events dramatically.

    What is interesting here is that strategy games rarely try to hide all the values they simulate, yet they remain compelling because of the richer interactivity. What if dialogues in more story-driven games had more complex interfaces, but combined with richer interactivity?

  • Entity AI in action-adventures. Combat is also a form of interpersonal interaction. Add the simulation of stealth/awareness and morale, and things start to get interesting.

  • Interaction with NPCs in GTA3 and GTA:Vice City. The player characters and NPCs in these games react to in-game events in pretty subtle ways. There is the obvious "running away when a madman points a flamethrower at you", but GTA3 was also the first game I know of that actually simulates jostling people on the sidewalk, which is a great way to solve the problem of people blocking your way, and which adds a lot of character to the game. The player character makes rude gestures at honking cars, and in GTA:Vice City he will make remarks that are pertinent to the situation and his mood. This makes the game much more lively and compelling.

What I find interesting about the last two approaches is that they add richness without significantly extending the interface. You can still do the same old things: run, shoot, etc., and apart from visibility or wanted level gauges, nothing has been added to what is commonly referred to as the user interface. But the underlying simulation has become richer, and the depiction of the state of that simulation has followed, through the use of animation and speech.

Conclusion

There is much more to be said about dialogue systems. The issues that come up when designing a dialogue system affect the whole game: you cannot design it in isolation. The phrase-based model, which has been in use for over a decade, has severe limitations. Using a different model could lead to satisfying new forms of gameplay, but it will take some time to find out how to do this, and to get players and other stakeholders used to this.

Finally, I may appear to be unfairly picking on KOTOR. I was just very excited to be able to play a console role-playing by the one of the best Western RPG developers, only then to be disappointed that so little has progressed since I last shipped an RPG in 1995.