Dialog - choices vs keyword guessing

Storytron. www.storytron.com

In my teens Ultima 7 came out. I loved that game (both parts) because it gave a sense of being able to go anyway in a world and do anything.

I was aware that this was somewhat illusionary - only skin deep. The biggest example of this was the dialogue that consisted of the several word options - just keep clicking till you've read everything - style of interface (like Morrowind). I thought this was a let down because it did not make me have to think about what questions I needed to ask. Ultima 6 was better in this respect because it was free text.

I thought that the future would be bright - that all the elements in Ultima 7 would be developed further in future games. With physics engines a reality now, worlds could have more of the go anywhere, do anything feel. The week link is still interacting with NPCs. Of course, Doom came out and adventure games largely died.

The problem with free text is that its hard for the software to understand. I once saw the suggestion of an interface where the player constructs their sentances using a limited vocabulary set and grammar structure i.e. the game forces them to pick from a word list (which possibly expands as new things are discovered) in a set manner so that the software easily understands meaning. This would allow a great deal of freedom in dialogue, and makes the player have to think about the questions/responses, whilst letting the programme understand exactly what the person is saying.

Couple this with advanced AI like information sharing between NPCs and I think you'd get a truly immersive world.

Joe

Hello Kest,

I've made in the past similar thoughts with what you are talking about.

Since I was a kid I particularly liked open and non-linear worlds and always the choice driven dialogs were ruining this.

Recently, I evaluated the use of chatbots (like http://www.alicebot.org/) for use in rpg/adventure games.

There are some benefits like:

- Human like behavior
- Extensive dialogs

But in the contrary I see the following difficulties in a possible implemantion:

- They respond correctly only in small and well formed sentences
- It takes a long time to prepare the dialog material for the bot
- The user must still guess in a way what he needs to ask the bot (although this not always a disadvantage)
- Hard to have voice recorded dialogs (obvious, huh?). Text to speech technology nowadays is quite good although it still sounds like a robot (oh, but it is a robot! :)).
- Hard to make the dialog flow towards where the plot of the game requires.

Ok, propably my assumptions are quite subjective. I think the subject of choice vs keyword guessing dialogs is fundamental for the creation of deep and realistic games.

P.S. The link that I gave above (http://www.alicebot.org/) features a markup language (AIML) that is a derivative of XML especially designed for the creation of the material needed for a bot to interact.

And SILVIA is an evolution of Alicebot.

Joe60, what you've described is very similar to storytron so it was probably that you saw.

It works like this:
players perform actions, by describing what they want to do, using an unambigious vocabulary and unambigious grammar provided by the interface.
The engine interprets this precisely as its unambiguous.

for instance:
pick up - rock.
make deal with - sally (sally - takes - bike, I - take - $100).

then all characters and props involved react accordingly.
characters react to what you do and to what other characters do, thereby creating a living community. If there is no reacting to be done, nothing happens until the player does something or some scheduled event occurs.

The main principle is that any action that you can perform in the game is available at all times, to all characters, privdided it is dramatically interesting.

The upside is that it achieves what a lot of us are looking for in terms of giving the player total control of dialogue, and equality of dialogue with the other characters in the game.

The downsides are first that the team have very specific goals in mind for its use and it'll be some time before they consider lending it out. So you either use their tools to make interactive stories from it, and have it played with their front end or think of something else to power the characters in your game.
Dimitri project by any chance?
Second that characters don't have long term goals, and there is little flexibility in deciding what influences the way characters react outside of the parameters they have chosen. There is no ability to take into account a character's intelligence for instance, and no way to include that. I'm pretty sure this kind of customization is planned for the future though!

Another interesting thing about it is that its such a big project because it is the development of the editor and the player, they want others to write games for it. Without the editor it would not be such a big task.

Quote:

Original post by Kest Normally, I would say a good, solid, multiple-choice dialog system is as good as text-based dialog can get. [...] I wonder if there isn't another way to go about achieving this. Some way to give the player the potential to make their own decisions, without handing them the options, but not requiring them to play guessing games. Any thoughts or ideas?

Yes... why are you equating dialog to decisions? Surely dialog is just one type of decision, and that it's possible for a game to offer many free-form and creative approaches to solving a puzzle without even requiring dialog at all.

I think keyword guessing is always going to be a bad idea, and Storytron's method just seems inherently awkward and mechanical. Kylotan makes a good point, though. Don't give your player menu options like "kill him" and "free him." Give your player the ability to use a deadly weapon and the ability to unlock doors.

One of the problems that has been observed with selection-based dialog systems actually comes from the implementation of the system. If you only provide selections in a dialog box that are relevant, the player knows that they need to simply click on all of them eventually and they will get all the information they need. Really it's the same thing as being able to hit a key and immediately know what objects you need to search because they are highlighted. It is impossible to miss picking up the cool objects that are in the desks, chests, etc. because you know where everything is.

If you were to provide numerous red herrings (either in places to look or in the dialog selections) it would generally take some effort to get the loot/information that you need. It would be less like leading you by the nose through the stuff the designer wants you to find/know. For example, what if every NPC that you can talk to has a battery of questions such as "how are you?" "What is your job?" etc.? That would be a lot of relatively useless stuff to wade through in many cases - but may be relevant in a few cases. It would be up to you to decide to ask if this particular person needs to be asked that particular question. You have now created a hybrid system - lots of potential choices, but no guessing. That doesn't completely obscure the esoteric information that you need to ask a particular NPC, however. You would simply need to scan the list for things that stand out. Of course, another major drawback here is that there needs to be a LOT of content creation going on to cover all the new contingencies.

It's a rough issue to solve.

Hm...

Why not choose by your actions? Do away with the player's ability to talk back. (I think) in Half-Life, the protagonist never actually said anything. The characters still talked to you, and interacted.

I like the idea of choosing with actions, like Logodae said. It's easier to make a free-form action system than a free-form dialogue system that emulates human behaviors accurately.

Quote:

Original post by Kylotan Surely dialog is just one type of decision, and that it's possible for a game to offer many free-form and creative approaches to solving a puzzle without even requiring dialog at all.

Communication with NPCs is where the majority of important decision making takes place. It is just one type of decision [making], but it's the hardest to offer free-form and creative approaches with. Or at least it seems to be.

Quote:

Original post by Logodae Don't give your player menu options like "kill him" and "free him." Give your player the ability to use a deadly weapon and the ability to unlock doors.

That's an interesting idea, and it works with that scenario, given that there's a door or a way to exit the scene. But there are many choices that would be impractical or impossible to represent through gameplay.

Quote:

Original post by InnocuousFox For example, what if every NPC that you can talk to has a battery of questions such as "how are you?" "What is your job?" etc.? That would be a lot of relatively useless stuff to wade through in many cases - but may be relevant in a few cases. It would be up to you to decide to ask if this particular person needs to be asked that particular question. You have now created a hybrid system - lots of potential choices, but no guessing. That doesn't completely obscure the esoteric information that you need to ask a particular NPC, however. You would simply need to scan the list for things that stand out. Of course, another major drawback here is that there needs to be a LOT of content creation going on to cover all the new contingencies.

I agree that this type system, or something derived from it, has a lot of potential. The main problem with it is that these dialog options that are always present could never relate to extremely specific situations.

Quote:

Original post by Kest Quote: Original post by Logodae Don't give your player menu options like "kill him" and "free him." Give your player the ability to use a deadly weapon and the ability to unlock doors. That's an interesting idea, and it works with that scenario, given that there's a door or a way to exit the scene. But there are many choices that would be impractical or impossible to represent through gameplay.

Well, you'll need to pick choices and gameplay elements that work well together. Either figure out what sorts of scenarios your gameplay can support, or approach it from the other side -- decide how you want your "big choices" to play out, and then design the rest of the game to use the elements necessary to those choices.

And for the record, I think that multiple-choice dialog is a good choice mechanism for a lot of scenarios, and I wouldn't suggest that you eliminate it.

I think the Storytron idea will eventually become workable, but it still has a long way to go before it moves from "frustratingly annoying" to "fun". There are many ways to simulate dialogue without needing to actually create an AI capable of interpreting dialogue.

One method that I thought was particularly interesting was Indigo Prophecy. In it, instead of being given a few pre-written sentences and choosing between them, you would be given three or four keywords and a fast timer, and you needed to quickly choose the keyword that matched closest to what you wanted to say, otherwise the game would just pick a default one for you. The reason this was an interesting system was because it actually did simulate some of what you're talking about in your first post. Some people would never think to release the prisoner unless they had a big, obvious "You may go free" choice sitting there in front of them. But some people would have that as their first reaction to the situation and would be wanting to make that decision before even knowing it would be an option in a dialogue tree.

So using your example, there would be a cutscene for a bit where the prisoner is talking to you and you're talking to him about how your god demanded that he be killed. Suddenly, four options pop up with a timer bar quickly counting down: "Regret", "Inevitable", "Coward", "Run". Someone who hasn't even considered the idea of letting him go would look at the four options and not fully understand what they mean, and would probably pick "regret" or "inevitable" for something like "I'm sorry I have to do this, but I have no choice", or "coward" if they sided with the god and agreed that he deserved death. They might see the word "Run" and not understand why it's there, and perhaps think it means that they themselves might run away, which is no good, or that it's an admonishment to the prisoner for running away from battle. But someone who was sitting through the first part of the conversation thinking "There must be some way to not have to kill this guy", upon seeing the four options, would know that telling the prisoner to "Run" sounds like the best of the four choices.

It's a little hard to explain if you haven't seen it in action, so I highly recommend checking out Indigo Prophecy if you get a chance. I thought it was very refreshing; it was a little stressful having to pick an option really quickly without knowing exactly what my character was going to say, but it led to me really feeling like I was inside the character's head, rather than the usual feeling of an uninvolved participant in a "Which response does the developer think is Lawful Good" quiz.

Quote:

Original post by makeshiftwings Suddenly, four options pop up with a timer bar quickly counting down: "Regret", "Inevitable", "Coward", "Run". Someone who hasn't even considered the idea of letting him go would look at the four options and not fully understand what they mean

This is another really interesting concept. So the basic idea would be to represent choices with something that would only be recognized by those already desiring them. I don't like the idea of a timer, but I understand its purpose. I don't think the timer would be necessary if there were a more reliable way to associate the choices with their hidden meaning.

In a way, the design without a timer would be similar to what InnocuousFox brought up. Something like "concept" buttons. "Mercy" or "Compassion" in this case. Reminds me of the old adventure games, like Maniac Mansion, with "push" and "pull" actions.

Quote:

Original post by Kest Quote: Original post by makeshiftwings Suddenly, four options pop up with a timer bar quickly counting down: "Regret", "Inevitable", "Coward", "Run". Someone who hasn't even considered the idea of letting him go would look at the four options and not fully understand what they mean This is another really interesting concept. So the basic idea would be to represent choices with something that would only be recognized by those already desiring them. I don't like the idea of a timer, but I understand its purpose. I don't think the timer would be necessary if there were a more reliable way to associate the choices with their hidden meaning. In a way, the design without a timer would be similar to what InnocuousFox brought up. Something like "concept" buttons. "Mercy" or "Compassion" in this case. Reminds me of the old adventure games, like Maniac Mansion, with "push" and "pull" actions.

I think you could do it without a timer, but for me, the timer was the most important part in making it seem engaging and real. Thinking back to Indigo Prophecy, there were times when one of the words in the choice made no sense to me, so I didn't pick it, and then only much later did I come to a realization of what it was. Situations like "Oh! Window! He was the guy that I saw outside the window! Of course!" If I had as much time as I wanted, I could have sat there for a while and thought about what "Window" meant, and in that case, it would have been almost as obvious as them just putting "Hey, didn't I see you outside the window earlier?" as a dialogue choice. The timer made it so that the game rewarded you for actually remembering things on your own, or figuring out a unique approach. I think it also made conversations seem a little more real, since you usually are "on the spot" in a real conversation; a paused dialogue tree is much more like responding to an email, where you have lots of time to decide what you want to say.

Quote:

Original post by makeshiftwings If I had as much time as I wanted, I could have sat there for a while and thought about what "Window" meant, and in that case, it would have been almost as obvious as them just putting "Hey, didn't I see you outside the window earlier?" as a dialogue choice.

If the two designs were mixed to create some type of "concept commands", then it would not be possible to have extremely specific concepts like "window". But perhaps a concept could be "recognition". It would still allow you to make that claim, if it seemed relevant to you as the player.

Several (possibly many) commands would always be present in dialog situations:

- Compassion
- Recognition
- Forgiveness
- Subterfuge
- etc

These would be listed in a sub-menu or somewhere near the typical multiple-choice dialog lines. Certain concept actions (possibly several) could be duds in situations where they make no sense. But when they do relate, they would override the dialog line options and execute a non-reversible command or statement. It would be impossible to "explore" the system. And it could often be dangerous to randomly experiment with them unless a command makes sense.

The biggest problem would probably be from the designer or story writer's perspective, having to deal with the constant presence of all of the commands. If the player always has this multitude of options, it would be impossible to limit their choices as many games often seem to do. I suppose the commands could sometimes fail, even when they relate to the situation. As long as it doesn't punish the player for trying.

Quote:

Original post by Kest Quote: Original post by Kylotan Surely dialog is just one type of decision, and that it's possible for a game to offer many free-form and creative approaches to solving a puzzle without even requiring dialog at all. Communication with NPCs is where the majority of important decision making takes place.

Says who? It sounds like you're making massive assumptions here that you're not sharing with us. Your assertion doesn't hold for most first person shooter games, or most strategy games (real time or turn based), puzzle games, sports games, war games, fighting games, flight sims, economic sim games, or music/rhythm games. Even most RPGs just use dialog as a way of advancing the plot and revealing information rather than giving you distinct decisions to make.

So just which games are you discussing here, given that it's not a problem in the vast majority of genres?

Quote:

Original post by Kylotan Quote: Original post by Kest Quote: Original post by Kylotan Surely dialog is just one type of decision, and that it's possible for a game to offer many free-form and creative approaches to solving a puzzle without even requiring dialog at all. Communication with NPCs is where the majority of important decision making takes place. Says who? It sounds like you're making massive assumptions here that you're not sharing with us. Your assertion doesn't hold for most first person shooter games, or most strategy games (real time or turn based), puzzle games, sports games, war games, fighting games, flight sims, economic sim games, or music/rhythm games.

I was talking about my own project. But what difference does it make whether the majority of it takes place there in your own project or not? Are you suggesting that it's pointless to discuss?

Quote:

So just which games are you discussing here, given that it's not a problem in the vast majority of genres?

I'm not discussing any specific games. I'm looking for ideas to improve decision making options within dialog situations.

Good dialogue is about being able to express what you want to, and for other characters to understand and react appropriately. For that, the minimum that characters and the player need is to be enabled with the following:

Goals, memory, ability to construct sentences from individual words, ability to digest sentences.

Any disagreement with that?

I've always thought that a good dialog system. Should take in to account conversation style, keywords, and npc responses.

By style I mean being able to choose how you want the character to approach the current interaction. The player could choose a style from a list of possible styles like:

Flirty,
Calm,
Aggressive,
Abusive,
Submissively,
Jokingly

The choices available to the player should all be relevant to the current situation, or lead off of an npc response.

The problem with most keyword conversation systems in games is that they don’t take into account the flow of conversation. After all doesn’t make sense that your follow question to “Did you kill your husband?” would be “How are you today?” It should be a question relevant to how that person responded to first question.