Hello! I'm Tom. I designed a game called Gunpoint, about rewiring things and punching people, and now I'm working on a new one called Heat Signature, about sneaking aboard randomly generated spaceships. Here's some more info on all the games I've worked on, here's the podcast I do, here are the videos I make on YouTube, here are some of the articles I wrote for PC Gamer, and here are two short stories I wrote for the Machine of Death collections.
Andrew: Just linked the book club to you, boosting your...
Luv as Smith: hello I am extremely excited about your game and...
God: Jesus didn’t mean what he said, I had a talk to him...
I started thinking out loud about it on Twitter, but didn’t want to swamp your timeline with it, so I’ll paste what I said so far and continue it here:
Made me realise how many judgements and assumptions we make in saying “Violence is easy to simulate.”
We mean that when you simplify violence into something that is easy to simulate, the fidelity you lose isn’t stuff we care about.
Whereas when we simplify social relationships between people to the same extent, the things we lose seem like a big deal, and the simplification feels crude.
One simplification we’ve learned to accept is a finite list of response options to dialogue. That makes conversations ‘easy’ to simulate, and allows for some authorial intent in defining the character you play, at the cost of some freedom for you as the player. Eg. “I want to tell the Illusive Man to fuck off!” vs “Commander Shepard wouldn’t do that in this situation.”
How much freedom you lose, and how much authorial intent is imposed, is largely up to the developer. Most err on the extreme side of authorial control – myself included. In Gunpoint, only a few dialogue choices in the whole game take you down a meaningfully different path. Most get you a specific response, so that the conversation reads well, then immediately return you to the same state you would be in regardless of what you said. How do I make that not seem odd? I just never give you the option to say something that *would* dramatically alter the course of the conversation.
How far can you go in the other direction? How ‘free’ can a dialogue tree get? There’s always going to be a limit on how much ‘stuff’ you can produce, so a limit to how many options you can account for and offer to the player. Branching gets expensive fast: giving the player four options for their first line only requires writing eight lines – four options and four responses. Giving them four options for the line that leads on from there means writing another 32. The next one takes 128. The one after that takes another 512. And after you’ve written those 680 lines of dialogue, you have an 8-line conversation.
So, that’s why people don’t do that. That’s especially why people making voice-acted games don’t do that. And that’s why sometimes when you say something in a BioWare game, the character’s response sounds a lot like it was written primarily as a reply to one of the things you could have said but didn’t.
So there are a bunch of tricks to make people’s choices loop back to the same point or sound like they made a difference when they didn’t, including defining enough about who they are in this world that you get to restrict them to a certain course of action.
But giving people lots of options is only the most naive, brute-force solution to allowing them more freedom. I said that simplifying the real-world concept of a conversation down to a dialogue tree ‘costs’ us some freedom we care about. In a way that restricting us to effectively only shooting someone in the ‘head’ or the ‘body’ doesn’t. So the challenge in making dialogue trees better isn’t necessarily just ‘reduce the loss of freedom’, it’s ‘reduce the loss of freedoms we care about’.
One freedom we almost always lose in dialogue trees is the option to just bark something nonsensical or offensive and refuse to participate, but that’s not one many people really, strongly desire in game after game. It might be nice to do it once, but ultimately it’s the freedom to not play the game that has been made for you, and you already have that.
So I’m wondering if there’s a way to focus your option-creating efforts on the options players will most want, so that the freedoms they’re losing are the ones they care least about exploring. Obviously ‘good writing’ gets you some of the way there, but that’s not a slider we can vary freely. ‘Testing’ helps too, and that at least is something we can just decide to do more of. But there’s a limit: it’s tough to observe players on a large scale, particularly with meagre resources.
So wouldn’t it be cool if, at the bottom of every dialogue tree in a game, you had a box called something like “What I wish I could say here:” As a player, any time you’re not happy with your options, you write something in there and it’s sent to the developer. It doesn’t help you personally, right now, but the dev can see “Boy, a lot of people really wish they could tell the Illusive Man to get fucked when he essentially recruits you into his terrorist organisation purely by having a spaceship that looks like your favourite spaceship.” Or “Wow, seems like people think they should be able to contact this Alliance Navy they’re an officer of, and who would sure like to know they’re alive now, and could totally solve all the current problems pressuring you into joining a terrorist faction.” Or “Jesus, seems like Tom is really bitter about this one particular non-choice in Mass Effect 2 even though we did an otherwise peerless job in blending an authored character with meaningful player choice over the course of a vast trilogy.”
If I do story in Heat Signature, I won’t do this – it’s better for a game that’s really ‘about’ the story or choices. What I am planning, though, is to write several short, self-contained stories rather than one over-arching one. As you can see from the maths, branching stories get exponentially more expensive the longer they are, so short ones are super efficient and don’t need to compromise so much on player freedom. And you can still have them feed into each other in a few binary, easily accountable ways – though I don’t know how far down that route I’ll go.
I Usually Use Random Generated Names: I feel like there doesn't need to be as many branching choices or a cynical dialogue if the game itself is entertaining. It just seems like something you would add only if there isn't enough content to keep the player interested and entertained. When I see excessive dialogue like that I feel like it's an attempt to shoehorn in more dialogue where it's not needed and it takes me out of the game.
Stories are prescripted are they not? I understand the concept of using interaction in the game to immerse the player but we have to realize that it's a story and not a conversation simulator. Going out of your way to give the player so many options can detract from what you want the plot to be. Planning what emotions the player feels and in what order is necessary to have something sensical. Giving the player what they want is not necessarily good for the story. If the player is meant to feel restricted or uncomfortable maybe that is intentional.
Just thinking out loud, no guarantee my thoughts are logical.
@mtrc: Interesting! Years ago my first videogame was almost based on the insane idea you mention that you just infinite branch. I knew this amazing writer who wrote the funniest dialogue, and the idea was that we would genuinely just keep branching as deep as we could. Ultimately it got too hard to plan, here's a screenshot from the planning doc (I hope Jon doesn't mind):
Nowadays there's some cool research on text adventures that do interesting things with branching. A lot of them are coming out of Mark Riedl's lab (@mark_riedl on Twitter). One student is working on a project that crowdsources new branches by asking people playing another game (or on Amazon's Mechanical Turk) to tell stories similar to the game's scenario. The system then tries to understand what they're saying and put them in as new branches of the story.
Branches feel so chunky and easy to play with that you do wonder why people don't do more with them. One thing that might be tricky in your example is: how do you know what was influencing someone's desire to say this right now? Is it because they have a particular personality, or because they made a certain choice? It's a cool set of problems to solve.
Oracizan: The Walking Dead is one of the biggest culprits in the "I'm going to give you choices lol j/k" arena, with many of their branching paths eventually or immediately converging.
However, I feel like they do something brilliant to make those paths feel different. After offering a binary choice like "save X" or "let X die", they then allow the player to clarify via dialogue with NPCs why they did what they did. This works especially well with TWD games, because a large theme in those is deciding who you are going to be in the face of the zombie apocalypse. Thus, "I did X because Reason 1" feels like a completely different option from "I did X because Reason 2", and these reasons are an inexpensive way to make a small tree feel like an expansive oak.
This is almost the inverse of the strategy of explaining the lack of player choice by defining who the character is (not that TWD doesn't employ that as well). This strategy offers the same limited game-affecting choices, but allows the player the choice of deciding Why in addition to What.
Jason L: Ref, as ever, Leisure Suit Larry's development. As a text-parser game they could just log everything testers wanted to do and rewrite their responses to account for most of it.
MJG: "So wouldn’t it be cool if, at the bottom of every dialogue tree in a game, you had a box called something like 'What I wish I could say here:' As a player, any time you’re not happy with your options, you write something in there and it’s sent to the developer."
I'm reminded strongly of http://xkcd.com/1350/
Awpteamoose: Something I've not seen in too many games is just dropping the written speech in dialogues altogether. IRL whenever we engage in dialogue is either to get the desired outcome (a favour, information, waste some time, etc). This outcome can usually be summarised in a short description, even if the interaction was very long. Sims games do this reasonably well as we do not hear the actual words spoken (well, we sort of do), but the pictures provide the descriptive information that can be used to understand what is going on. Since we understand that the same picture appearing in different dialogues does not mean that the same thing has been said, it doesn't feel like reused dialogue lines and feels like emergent storytelling.
My point is that a game can be done with more player freedom in dialogues and various other social interactions if we just substitute descriptions in place of speech.
Tim W: Interesting discussion! I recently started playing Fallen London and was intrigued to see how that handled what seems to be (to a beginner at least) an almost infinite well of choices and things to do. This is similar to what Awpteamoose described: instead of choosing what dialogue to say like in Mass Effect, you are instead choosing from a wide range of things to do, each one quirkily described by an anonymous narrator. They also recount whether you succeeded or failed, which then affects your stats, which in turn determines what you can and can't do in future. It's a level of abstraction away from specific dialogues that seems to be an efficient way to manage choice. And like Awpteamoose said about The Sims, when repetitions do occur they are less noticeable than repeated dialogue - in this case too, the narration is so fun and engaging that I rarely mind a repeat.
Allowing player-submitted ideas for dialogue options is an interesting concept. I noticed in one or more of the Mass Effects that you could turn on some kind of live monitoring that I assume told BioWare (anonymously, hopefully) how I was playing the game. So I could imagine that going one step further as Tom says, albeit on PC versions only - unless I can record using my Xbox headset what I would like to say instead ;)
Jason L: Oh yeah. I don't know which came first but see last week's Crate and Crowbar for Chris Thursten's remarks on Sunless Sea and Tom's contrasting it with FTL.
Jason L: Brain dropped the dependent clause somehow. 'Last week's' C&C is #79.
Tim W: Thanks Jason, I'll check that out!
blubb: Your thoughts definitely seem to be relevant but in my opinion only to a very limited choice of games. As one of the comments already mentioned quality of the story is much more important than player choice. One of my all-time favorites is the Metal Gear Solid brand which is actually very linear. But it feels like playing a film. A film in which the characters have depth, the story is at least of average quality and the presentation as well as the dialogues are interesting, presented and written very well. The gameplay is also quite simple and straight-forward ensuring high quality. Story still needs content and for the amount of content to seem natural you need a lot of people to provide ideas. In my opinion for small studios the idea for gameplay is much more important.
Abraham: Are dialogue trees really the best way to simulate a conversation in a game? They're the best we've been able to approximate communication, but they're basically a non-game mechanic. At best, they let you make and justify a decision, but conversations are never just a series of decisions. Usually, dialogue trees are used to break up the monotony of reading a block of text, but we already know that huge text blocks are a bad way to provide information to the player. This is why things like audio diaries are a much preferred technique to accomplish the same goal.
I think the real problem with designing an engaging "game-conversation" is that the conversation is hard to turn into a game. And that's because almost all video games are built around the concept of spatial reasoning. What we usually consider "action games" are really just spatial games. Errant Signal has a good video on this subject: https://www.youtube.com/watch?v=wSBn77_h_6Q Shooters, fighters, racing, sports, platformers. These are all games whose challenge comes from mastering spatial reasoning. Almost all games outside of this category are made up of rpgs and puzzle games.
And what's been tried so far is to turn conversations into puzzle games (think Oblivion's Persuasion Wheel), usually with poor results. I think to tackle the obstacle of "social interaction in games" we should do what computer games have always done: look back to more primitive games that handle the concept well. There are a lot of popular party games that don't involve violence or spatial reasoning in the least. A game like "Apples to Apples" might make an excellent foundation for computerizing social interaction.
Brad V: Hmm. Yeah, it does feel like the first step towards a better interaction system is to move away from "What do you want to say?" to "What do you want to do in the conversation?"
I've been through the process of sitting down, writing out the dialog tree choices and diagramming out the options from both sides. On the writing side, you pretty much end up having to write silly lines that overstate the intent behind that choice; and on the player side you end up with the task of reading each dialog choice and deciphering what it's supposed to accomplish.
So if your choices were things like:
and other actions to manipulate the social variables, then actions to use the social variables like
- Dig for information about a specific topic (which brings up a list of things to ask about)
and, for when you really don't care specifically what happens next
- Just make conversation or small talk
You'd probably get a more fun system out of it, and not have the hassles of building a dialog tree.
No, seriously.: I love you Tom Francis. And I don't regret this one bit.
Chris R: Fantastic read from you all and not just Tom. Thanks guys!
Joseph: Ok, perhaps someone here might be able to point me in the right direction, and it totally relevant to this post - There was a blog post recently that had some stuff about the possiblity of Turing-Test passing (or nearly) NPCs allowing for functionally human conversation, and I read it, and I forgot to bookmark it, and I cannot find it for the life of me because I think I found it from twitter and we all know that twitter is basically shit for archiving anything. Has anyone seen this?
Josh Grams: I know I'm months late to the party, but it seems to me that we should be using choices to broaden/narrow/flavor the possibility space, rather than branching the entire story. Choices could affect individual aspects of a story, rather than the story as a whole. Nicky Case suggests this (http://blog.ncase.me... ...e-stories/) and much of Emily Short's work seems to be along these lines...I dunno, maybe it's more work than it's worth, but it seems like there are situations where you could do it fairly cheaply...