Why I hate ASK ABOUT and TALK ABOUT

For what it’s worth, I don’t want to focus the player on “‘what’s the magic password for me to say to this character’” either (though I realize that that is the experience some players have had with my work).

What I want them to think about is “what in the range of conversation here is important and interesting? what am I interested in digging into more?” This implies a kind of conversational space that isn’t purely physical, and invites the player to explore it in a similar way.

Ideally, actions that will significantly affect the NPC’s state are described in a way that clues the player in about that.

What I was arguing wasn’t “you can’t do these things without suggesting specific conversation topics”; what I was arguing was “you can’t do these things purely through environmental cluing.” I actually do think there’s a lot of interesting stuff one can do by specifying a speech act (INSULT DOUG, BUY A COW, etc.), or by breaking down conversation in some completely other way (choose when to interrupt someone, what emotional tone to set, which of several debating tactics to use next).

The reason I think your environmental method is limited is that it represents a second-order kind of agency. It requires the game to draw conclusions about what I mean when I set the teddy bear on fire or rip a photograph in half or lay the polished rock on top of the gravestone… or what I mean by glancing at a poster or reading a book or dropping a box of chocolates. That’s doable enough when the story is very well defined. It’s a lot harder when the story can branch or the protagonist is making significant choices or is in an unknown knowledge state. (Mysteries are hard exactly because the game is trying to assess what the player knows and give ways to translate that back into in-game actions… without providing overly obvious hints about what to do next. There are games that belong to the literary genre of mystery that don’t do this – Death Off the Cuff is an obvious example – but that’s because the player is not being asked to deduce anything himself.)

So the more things a protagonist could reasonably be doing or thinking in this situation, the more explicit the connection needs to be between the player’s action and the protagonist’s intent. Otherwise, the player is thrashing around with minimal control.

Edited to add, especially since it drives me nuts when other people say categorically that something is impossible in interactive stories: when I say “I think you can’t do X,” I’d be happy to be proven wrong by example. But I mean that I don’t see a good way of doing it. If I were confronted with the design challenge of writing a game that was all about communication by environmental interaction, I would constrain it pretty sharply in the ways I described in order to make it work.

Yeah, I’m all for that. Good luck!

See, that’s the part that I personally would identify as IF’s unhelpful holdover from an earlier era: the relentless trivial physicality, the obsession with open and closed boxes, with doors and buttons and lights. We have gradually moved away from some of that, and light source puzzles (for instance) are much less common than they used to be.

But I do think it would help if our conventions included more allowance for abstract or impressionistic treatments of setting (yes, this room is an office; no, I didn’t implement the desk, because it’s not at all important and the player’s attention is meant to be elsewhere); and more willingness to move the action to a partly-abstracted sphere. The room only gets you so far.

This is a nice discussion. I’d like to see as many different approaches as possible tried, but in general I like conversational approaches that give me agency – so long as it’s done well, which is of course the trick. I’ve shared Paul’s frustration with “What topic do I need to mention to advance the game state?”, but I think my problem there is not with the topic-based conversation system, but with the idea that the purpose of the conversation is to advance the game state.

That’s my experience with a couple of different conversation systems and kinds of game. What felt like a lack of agency wasn’t being able to select only from a menu or from the topics that had been introduced, it was when I couldn’t use the options to control the PC the way I wanted. And that was heavily correlated with having a particular thing I wanted to do in the conversation, rather than using the conversation to roleplay.

So In Galatea (topic-based and Best of Three (multiple-choice), I feel like conversation usually flows smoothly (sometimes I dead-end in Galatea, but there’s usually a way to restart it). But in Alabaster (multiple-choice) and the comp version of Death off the Cuff (basically topic-based), I felt like I lost control over the PC; sometimes I was being pulled along behind the PC and sometimes I was trying to drag him with me. Which is to say that sometimes the PC would make leaps that I hadn’t seen yet, and sometimes I would have figured something out and couldn’t figure out how to get the PC onto that topic. And I think that the difference is that the latter two are both mysteries, in their way, where it’s natural to try and push the conversation to a particular outcome. You’re trying to find out what’s going on and master the conversation.

Whereas in the other two games the interaction with the NPC is the point. It was OK in Best of Three when the conversation menu revealed something I didn’t know about the PC, because I didn’t feel like those were things I was supposed to figure out or that I’d missed out on something by doing it; and in Galatea I was usually serving as a sounding board for her, so the PC tended not to blurt out something I didn’t want. (Well, often he was an ass when I didn’t want him to be. But that was part of the point.) In fact the bumpiest experiences I’ve had with those games have been when I was trying to push them to a particular outcome.

Now I’d really like to see more done with taking different tones, though that seems like something that wouldn’t be easy without a multiple-choice inference. And I really like the idea of choosing an outcome for the conversation, and then having it spool out turn by turn; maybe you could have different choices for tones and strategies within it. Though that could turn into a game of “pick the right choice to bring about your preferred outcome.” Having conversation develop from your actions organically and automatically (in katz’s sense, where it happens when you do a certain thing) seems like a nice idea, but I don’t know if it’d work where NPC interaction is the main point; it seems like it works best where the NPC is commenting on what you’re task, as in A Day For Fresh Sushi. It’s interesting that Pacian brought up Walker and Silhouette as an example of this, because I would’ve thought of it as a keyword/topic-driven system; actually it’s got all three kinds of conversation. The first scene has keywords (some of which trigger actions and accompanying conversation, some of which are conversation topics), the interrogation lets you pick the tone from a multiple-choice menu, and other scenes, especially the big climax, have conversation triggered by other actions.

And while I was typing this I started Jon Ingold’s new multiple-choice story, whose first set of options is “Yes/No/Tell the Truth/Lie.”

Yes, yes, yes. Unfortunately a lot of this is pretty thoroughly baked into the world model of I7 at least. One promise of multiple-choice/keyword systems is that they let you do things at a more abstract level, though there are tradeoffs.

@Paul: I would say that “in IF the environment is the story,” is precisely one of those “leftover assumptions from the birth of the form.” While we may look askance at the CYOA and hypertext communities’ lack of mechanical complexity, they look askance at our obsession with setting. So while I empathize with your “simplify, simplify” impulse, I’m suspect of its direction: even if your WIP proves its concept swimmingly, it may be regressive, not progressive.

I think about the dialogue-by-inches thing a lot. (Thanks for the term.) My WIP has a fancy-pants parser, but still, the player enters input text, not prose. The command prompt is changed so the whole construction looks exactly like a sentence. But still, it’s not prose. So I’m concerned about unintended incidental characterization as well as the quality of the whole WIP, being funneled through parser-mouth that way.

(My direct opinion on ASK/TELL: guess-the-noun is worse than for Examine since there’s no room description to mine for possibilities, but, there’s nothing wrong with the default ASK/TELL response be a (possibly in-world) TOPICS command.)

(Noticing that link points to i-f rather than I-F, there’s this awesome Inform 7 extension called Command Prompt On Cue, and I really must petition Merk for a horn-tooting smiley.)

While I can’t say whether Ask About / Tell About is the best interface, I think it works well in some situations. Notably, I’ve enjoyed Murder Mystery stories that use this interface; perhaps that’s owing to the premise that has the player operating as a detective where there are certain expected lines of questioning. Infocom’s The Witness comes to mind, which was quite limited given the state-of-the-art from 1983. More recently, Chris Huang’s “An Act of Murder” does a good job with this interface.

Not surprisingly, I’m using Ask About / Tell About for my first IF story which is of the same genre. I’m totally open to suggestions on how to best use this style of dialog to make it work well within the constraints etc.

–Zack
www.z-machine-matter.com

Just to be clear, I don’t think an interest in setting is necessarily holding IF back – I’d like it not to be the only thing we do, but setting is a strength for interactive games of all types, and I think we do well to keep exploring the possibilities there. But I do think it’s cramping if we always do setting at the lowest possible level, with interest predominantly in small acts of physical manipulation.

(Disclaimer: I have a WIP that is very much about setting, but I’ve tried to focus the interaction on actions that are a little unusual for IF, and streamline the opening doors sorts of activity so that they’re as automatic as possible.)

Thanks and I have tried Galatea and enjoyed it but felt it was an exception because the ASK ABOUT is essentialy, in that case, the only interface. In fact I think you could probably recode it such that every single thing typed is interpreted as a conversation topic (no ASK or ABOUT required and no other types of commands understood), and it wouldn’t really disturb the functioning of any of the actual pathways through the game. So, in that case, the game still only has one mode, really — in that case what I’d prefer is that the remaining trappings of the verb/object system be entirely discarded, and let the parser just be what it is — a sequence-sensitive keyword search. But in agreement with you, it was satisfying it was one of the most satisfying IF conversations I have seen but not really as satisfying as following a good, intriguing conversation into which I didn’t inject any words. There’s just something about that word injection and the conceptual and stylistic ranging limitations that puts on one half of the conversation, that diminishes the act of conversation.

For example, how do you represent utterances that are meant to affect the other person but without any grammatical content at all? How can you express sarcasm? How can just say, ‘duh’? How is the game ever going to interpret the huge range of conversational effects I really need to hold up my end of an entertaining conversation? Thus, conversation are just better when half the decisions aren’t being directly driven by the parser. I want an indirect connection because I want to read whole conversations ripped out of the game designer’s mind, stylistically intact and with no compromises in dynamism and pace.

Haven’t really got into Blue Lacuna yet — I just started playing it again the other day. I’m not sure what you mean by overlaying but I guess I will find out.

LOL, you and tggdan3 both mentioned solutions that use the name comma phrase thing. Funny because I also dislike that interface as a way of ordering people around – I prefer the method where you actually assume their POV for the duration, that I first saw in Suspended and Hitchhiker’s Guide. (I think they used the comma method to switch POVs but thereafter you could just type as them.)

Yeah I agree with you that we shouldn’t get fancy with the ASK/TELL system of course, because I already think the system itself is too fancy. Sure, anything is awful when poorly executed, but my POV is coming out of having experienced all too often an ASK/TELL interface that kind of stank up and slowed to a grind an otherwise fun and well-written game. So, maybe they can be perfectly executed but it seems like it’s a lot harder to achieve that, it’s like the triple-axle of IF, so it should be a whole lot rarer than what it is, no? At least, I wish it were. And I definitely wish that it weren’t this automatic thing that people seem to resort to now to reveal backstory. I’d much rather open a mailbox and read a leaflet than have a pretend conversation that is actually just a bad conversation which happens to also have the property that it reveals backstory.

I guess you could put a different spin on it and take away from discontent like my own, the lesson that maybe ASK/TELL is difficult enough to design well around, that it’s not a great stock way of revealing background information and might be better reserved for situation where the outcome of the conversation really matters. If there were just way fewer of these things and they didn’t turn out to be inconsequential to the story almost every time, my patience might not have such a short fuse for them.

Here’s a thought experiment I like to do sometimes: if I have already solved your game, can I play it through completely from beginning to end without talking to anyone? If I can, then I’m usually not too impressed with the wisdom of implementing all those elaborate conversational interludes, plus it seems my ‘agency’ wasn’t really required. This describes a surprising number of games that use ASK/TELL. (Not all of them though — I think the very first, was it Deadline? Could not be solved this way, which is perhaps telling as to how these commands were originally envisioned to affect outcomes… quite strongly, or else not really be used at all?)

Paul.

Are they numbered? I hope they aren’t numbered… that bugs me, too. (Man, am I ever picky.)

That’s a sweet idea. What I’m using to replace the ask/tell default responses at present is just ‘[So-and-so] can speak for himself.’ (This works because my game is in third person instead of second person.) Rather flip, I know. Maybe I should include an anti-tutorial, for veterans. 87

Not bad, so basically you are restricting to one topic choice per character. This is a nice way I think of restricting the field without necessarily coming off as robotic (which as you pointed out earlier, is usually a lost battle anyway – it be seem robotic even with 20 topics). I’d still to prefer to avoid even the one topic myself, but I’d love it if more games went your way in #1.

Or even a sort of thumbs-up/thumbs-down system. Could work. My own interest still lies, primarily, in directing player agency somewhere else entirely and then having those choices bounce back into character, smoke-and-mirrors style. But as you say, it’s a matter of taste and that goes for choices of solutions even more so than identifying problems.

P.

P.S. I’ll be back later to pick up responding where I left off, but I have read ahead and some great points are being made…

It’s interesting you put it this way, because I think of it as a pace issue as well – but interaction pace. I don’t expect players to want to read more than a paragraph or two of text in response to a single turn, unless something really amazing and drastic has happened (which should be infrequent). If I’m routinely printing more than about a quarter screen of text in response to a move, I’m probably doing something wrong.

This isn’t exactly true – “x something” is sometimes kind of natural and important, “a something” and “t something” occasionally yield different results, there are some implemented "galatea, do something"s, and a few other actions too. OTOH “a something” is probably a pretty enormous percentage of the commands typed into the game.

Well the purpose of my game is not solely to demo this, but this ideas are definitely in there.

I misinterpreted your reference to conversations automatically happening earlier, although the definition you do give I still find doesn’t really cover everything. Where does indirectly giving a command that allows it to happen fit in? It doesn’t seem to be squarely automatic or non-automatic according to your definition and so that’s what concerns me — I don’t want people to think that I am advocating even more robotic, clockwork-like diallogue that depends only on factors extrinsic to the player’s choices (such factors even if randomised usually feel like ‘clockwork’ to me).

Wow, thanks for scaring up those links for me — very interesting, especially that first one. The second one is also of interest. The techniques represented in your third link don’t interest me so much — too on the nose, and I guess ‘it’s just so on-the-nose’ is a pretty apt description of what it feels like to me to step through most dialogue trees and to type topics at a character, telling them what to talk about. There is a school of thought apparently that ASK ABOUT etc. are not even direct enough, so we need to specifically feed buzzwords into an NPC’s brain sort of ‘behind the scenes’ of the conversation. This solution feels even worse to me — perhaps the ultimate extension of it would be a command like, ‘> STORY, BE ABOUT LOVE.’ I don’t want to decide that, I want to discover it, and that’s I guess how I feel about topics and why I don’t want responsibility for them.

Isn’t that just what it’s like when you have a puzzle to solve. This second-order agency that you accurately define, and the uncertainty it creates; wouldn’t the resolution of this uncertainty be some great subject matter for a puzzle? Why is it any poorer than the uncertainty you get when you don’t know (I won’t call it a magic word anymore) what to ask about?

And if the environment is well-designed in sympathy with the story’s point, figuring out how it can echo into character and what that means, is the whole ball game, IMO. I definitely don’t see it as a negative, I see it as good artistic space because the author is defining fresh relationships between objects that you have to explore. Whereas I percieve stocking an ASK ABOUT system as more like filling in a dictionary. (Not that it can’t be artistic but you have to fight against something to make it so — as opposed to coming with some schema for what objects mean to characters, which is more like just building a real world.)

I’m trying to avoid too many categorical statements myself, but I still don’t see all the constraints; I see a fertile green territory there yonder. 8)

Paul.

I think that you and Ron who have stated/implied that focus on setting is ‘regressive’ (to use Ron’s term) in IF are conflating two things that seem very similar but which actually don’t logically need to be conflated: the containment tree, which of course drives all parser-based IF, and the neanderthal puzzle (‘How do I bring LIGHT to this place? Grunt.’)

I don’t know, I suppose I can’t really push this point much farther without drawing all my crazy ideas together and showing how they interact. I don’t use the containment tree in a standard way — I let emotions bleed into it. I believe it’s necessary, because I think the object is to IF as the shot (or the cut, to an editor) is to cinema: it’s the fundamental unit of meaning for the form, and to think of it as regressive or old school is I think to throw out the baby with some bathwater there. I understand this opinion because I understand what it’s rejecting and I reject that, too. Light puzzles: mostly lame. Inventory management puzzles: mostly lame. There are all these conventions that are now kind of lame, I agree, but the answer isn’t trying to supersede the default containment model, just like the answer to the pokey shots of early cinema wasn’t to denigrate and move away from the idea of the basic unit of meaning being ‘the shot’. The real fertile territory in cinema was opened up when people came to see the shot in a different way: when people opened up the ‘shot space’ to encompass far more than it ever had. The same basic unit of meaning: stretched artistically into extreme flexibility; this won out over more stagey concepts of how to evoke emotion, even though as Eisenstein showed us with his montage experiments (he showed the same shot of an old man’s face interecut with different things, making it seem like he had different emotions), it’s all smoke and mirrors. And yet creating smoke & mirrors from the simplest, most non-emotional seeming toolset, actually ended up to be the most flexible way of creating emotion on screen.

I think the indirect nature of the relationship between the shifting of a camera’s POV and the audience’s feelings, have a lot to do with that. It allows an artist to construct their own language for what things mean and to communicate those meaning using the simplest language possible: juxtaposition. Applying the same sorts of artistic ideas and principles to IF to me suggests rehabilitating the way we use the object tree: rather than building outside of it.

If you had shown early filmmakers many of today’s films with their dream sequences and flash cuts and flashbacks and flashforwards, and POV shots and extreme-close-ups, and animated sequences, and etc. etc., and these are all SHOTS, but they would likely say, ‘That’s wrong. That’s not what a shot is for.’ This is what I hear when I read that IF is moving beyond the object tree. There is nothing beyond the object tree: to me exploring what you can express with an object tree is the whole point of the exercise and we do not seem to have come close to plumbing this territory whereas conversational trees seem to me played out. All about perspective I suppose.

That’s just it: the conventions do allow for this. There is nothing in the Inform (for example) language that forces you to make a room an actual physical room, and an object an actual physical object, or to describe them in mundane, unemotional terms. This is a view that we are actually imposing on the object tree with our literallism, and it isn’t necessary: you can break this rule as you please. Anyway, once again — this is the sort of thing where probably very few are going to get what I mean without a demonstration — it’s in my head, though! And soon it’ll be in a game. 8)

So, conversations are only good when they’re irrelevant? Heh heh it sounds absurd but I kind of agree that the attempt to capture player agency for some important goal is part of the frustration factor. (The other part is just wading through wads of mostly irrelevant backstory that’s been shoved into a character’s mouth with no narrative aplomb, Babylon 5-style – that’s what it is, most IF characters interact with me as if they all live in the overexplanatory, mealy-mouthed world of Babylon 5. 87)

So it’s kind of damned if you do and damned if you don’t. If it’s irrelevant then it’s irrelevant. To the extent that it’s relevant, it tends to be overwritten to the point that frustration sets in.

My formula for good writing is pretty simple: hook them hard with something intriguing, and then string them along with something fun, and use whatever respite from the winds of impatience you’ve earned to express some of your philosophy. The ASK/TELL system is very hard to work around that suspense-focused principle. If it opens with a strong hook, then the interface feels like it gets in your way and the stringing along is no longer fun, and my interest in the hook fades. If there is no strong hook, then it’s definitely not a wise moment to be sending the player into a complex conversational cul-de-sac, at any rate. (Like even if I were writing a linear story, the moment when my strong hooks are played out is the worst possible moment to launch into a long conversational digression, because there’s nothing holding people.)

‘Is the conversation fun’ would be a huge plus that would mitigate the problem, but it’s very hard for an ASK/TELL structured conversation to come out as ‘fun’, I have found. Which is sad, because conversations are usually the most fun in literature.

Wow, interesting. I wonder what drove the different choices for different situations, I will have to play this game, for sure.

Paul.

I disagree with this pretty strongly (and I referenced why above in a reponse to Emily). Anyway, there is a reason I prefer to write IF over CYOA and it’s exactly this: the object tree is a strength, not a weakness. Otherwise, aren’t we all in the wrong medium, and wouldn’t just a very sophisticated CYOA be better? Are all the ASK/TELL afficionadoes eager to upconvert to CYOA but just don’t have a sophisticated enough CYOA tool, yet? If the object tree is so regressive, then maybe it should be abandoned in favour of CYOA, by those who feel that way?

I like CYOA and have experimented with it particularly for projects that weren’t susceptible to a parser, but I’m more into IF precisely because I am inspired by the possibilities of what I can explore through and express with a simulated text object tree. I’m inspired by all the things that don’t seem to have been done with it, all the things that could be. Not regressive at all in my book, and I personally feel that’s a bit of a misunderstanding of the underpinnings of the form to think so, the way it is when people say that ‘comic books are about action because they have pictures’. So maybe pictures are regressive and speech bubbles are the future? If that sounds weird, that’s how this sounds to me. It’s like tarring the heart of a neutral form with the brush of the way a few practitioners have used it. Most designers have used the object tree in mundane ways thus the object tree is mundane and true human abstraction requires a ‘higher’ system — I disagree with this. The object tree IS already maximally abstracted. Nobody says a room has to be an actual room — it could be a state of mind. It would be more fun for me to explore weird ways of using the room/object interface than to explore abstraction using a much more unmoored and nebulous topical interface.

This sounds really interesting as an experiment — looking forward to it!

Paul.

Alright I’ve probably mounted the best defence I’m capable of on this, and would probably only write in circles going forward, so I’m going to consider my case rested and will not waste people’s time henceforth restating things further at length, but to mop up some neglected replies…

Yes, I do agree that in terms of classical IF these are the types of stories where this interface worked best and was even almost fun — when you were playacting as a detective. This interface was added very, very early (Deadline) — so why didn’t Infocom then export it to every other game to handle conversations? They kept it very restricted to only the mystery-style games. But that’s not really what I see today. Not that Infocom is the be-all and end-all, far from it, but they were pretty smart people, and they must have had a reason for these choices… I feel that they at least in some ways felt the way I do about this interface. It’s a necessary evil, perhaps, to be avoided if possible. In a straight-up detective game, where you play the detective, perhaps it is not so possible to avoid — I’m willing to accept this as true as far as it goes.

I agree with this. I don’t want to read a whole long dialogue all at once. But there are many ways to break up long texts and some have a bigger impact on ‘pace’ (which is hard to quantify in IF — how long does it take to think up a topic heading?) than others. Also pace in narrative is completely contingent on your proximity to the most recent strong narrative question: the further away you are from it, the ‘longer’ every word becomes. This becomes really important in the case of ASK/TELL, where overly deep conversational trees tend to lead you pretty far from the point of it all and tend to return you feeling you have wasted your time and are now in the same position with less interest. Again, as you have argued, with the right technique these concerns can be allayed, but it’s pretty difficult.

Point taken. I was close, but no cigar. 8)

Paul.

Not irrelevant, but an end in themselves rather than a means to another end. Which solves the pacing problem you mentioned, I think; you’re at no remove at all from the narrative question.

Does that mean the conversation has to be the whole game, as in the two I cited as ones I liked (and, come to think of it, the two I cited as ones I didn’t like)? I don’t think so – you could definitely have a game where you explored and talked to people, and you were just talking to them in order to talk rather than to unlock some new part of the game. And you could have a game with puzzles and also with conversations (or other NPC interactions) that didn’t serve the puzzles but were interesting in their own right. “She’s Got a Thing for Spring” does that, I guess, to the extent of disabling the timer on the puzzles while you’re talking to the NPC. (I didn’t enjoy the puzzles, though.)

This I definitely agree with. One of the things I’ve been working on for a long time is the use of NPC conversation direction to keep the conversation moving forward. So for instance, as an author, I’d define a scene with a list of things the NPC definitely wants to cover. The player would be able to ask questions, but as he reached the end of a productive line of investigation, the NPC would redirect the subject onto the next important element – so you’re constantly moving forward, not just exhausting a static tree. (This is what Alabaster does.)

So you would be talking to them just to talk rather than unlock a puzzle, but it’d still be part of some proximate narrative question whose answer I assume would be revealed in the course of the conversation, do I have that right? If so, it sounds like great advice. 8)

Sounds like an improvement — with my level of impatience though I’m not sure I would notice the difference on my first encounter with it. I’m rather guilty I’m afraid of cancelling out of anything that even smells like a dialogue tree after trying only one or two topics and only returning to it when I’ve exhausted other avenues. (At which point, sometimes it can end up impressing me — like talking to Eruera in Aotearoa was not bad, and seemed absolutely necessary, but I still felt he was overtopicked, and yet I felt obliged to plumb everything available in the end, for clues.) Haven’t really played Alabaster yet — I definitely will. I keep noticing how many authors’ names there are on it and get curious about how that collaboration happened.

Paul.