nicked on 25/10/2017 at 05:54
I think a good conversation system would be emergent and dynamic, but that pretty much requires teaching the computer to talk.
Chade on 25/10/2017 at 23:39
Did anyone ever try (
http://www.interactivestory.net/) facade back in the day? I tried a few times but never got anywhere with the AI.
I enjoyed reading about Chris Crawford's attempts to do interactive storytelling back in the day. He had stories of, for instance, Lancelot and Guinevere cheating on Arthur, which for Guinevere would be a very prominent event, so she would rush off to tell Arthur all about it ... problems like that. I can't find that particular story on (
http://www.erasmatazz.com/) his site anymore though.
I get the feeling that the best work on intelligent agents is now happening throughout the rest of the tech industry, and I'm hoping that matures and trickles down into the gaming industry at some stage.
Starker on 26/10/2017 at 05:56
I did and I still think NLP is a dead end. It might work for something short and nonsensical like Little Pink Best Buds, but I think there are inherent problems with it, starting with the expectation that the player can type anything and be understood.
On the other hand, has anyone played Emily Short's Galatea? One of the best dialogues in gaming. Also, Progue from Blue Lacuna was a similarly complex reactive character.
I think the way forward might just be better dialogue and better NPC's whose personality is molded by the way the player interacts with them.
Chade on 26/10/2017 at 14:17
Downloading Blue Lacuna now ... (downloading does not mean dedicating the time to play tbh, but I hope I'll get to it soon)
I'm not current with the latest games, but as of a few years ago it seemed that conversation systems were one of three varieties:
1. Info dump with a bit of window dressing, in the form of (hopefully well written and interesting) characters. Can certainly add a lot to the game but is still ultimately just an info dump.
2. Puzzles with fairly obvious win/lose condition, immediate in-game effects, and generally little in the way of longer-term consequences (or if there were longer term consequences they were of the form I complain about below). Possibly a good opportunity for role playing. Still feels very gamey due to the obvious win/lose conditions and immediate rewards.
3. Pushing a handful of values on a slider (e.g. good/evil). The slider is usually very thinly disguised and stereotypical, and generally the player will have a particular "direction" they want to go in which will remain unchanged for the whole game. The sliders are either player attributes, or if they are about other characters, it's usually literally "but enough about me, let's talk about you: what do you think of me".
I think a few things are needed to have really interesting conversations with game characters: first they need a broader range of actions at their disposal with more dramatic potential, otherwise what are the conversations for? Secondly the progression of the story should be reflected in the states and desired states of the characters around the player: the way I interact with them and the things I try to get from them should change as the story develops. Thirdly, well ... I think there just needs to be a certain amount of complexity (or at least subtlety) so that I can't mentally model the world as a handful of items in a spreadsheet.
Thirith on 26/10/2017 at 14:30
Quote Posted by Chade
1. Info dump with a bit of window dressing, in the form of (hopefully well written and interesting) characters. Can certainly add a lot to the game but is still ultimately just an info dump.
That's something I've really come to dislike in (especially fantasy) RPGs. I don't mind background texts that are long and expositional, if they're well written - but put them in the mouth of a character, and in most cases that character ceases to be believable. People don't talk in long info dumps, in particular in regular conversation. I understand that you're trying to create a world, but in most cases it's infinitely better to have characters talk like they already know that stuff and they're pretty certain you do too; long reams of exposition-by-conversation are almost always there for the benefit of the player rather than the player character, and it's to the detriment of the character doing the talking.
It's one of the reasons why I didn't particularly enjoy
Pillars of Eternity. The quality of the writing was okay, but I rarely felt I was adventuring alongside characters; my party members were basically various flavours of exposition deliverers.
scumble on 26/10/2017 at 15:10
I think the writing in DOS2 is a bit better, or possibly I'm thinking of the way character interaction is pushed on you when you encounter characters that are relevant to your companions quests. Just that makes the party seem a little more like something you're not 100% in control of.
Pyrian on 26/10/2017 at 17:47
Info delivery isn't optional. Just getting people to understand things is challenging, and the results of failure are truly unacceptable in a way that nitpicks about believability just aren't. The issue is in no way limited to games. Most movies are brimming over with exposition thinly disguised as conversation (e.g. Wallace's monologue in Blade Runner 2049). And the worst part is, failure to understand doesn't typically lead people to merely not understand; most people will jump to the wrong conclusions and judge your work based on their own imagination. So, yeah, people are always trying to deliver information in ways that are in-character and entertaining, but don't kid yourself that it's easy, or that you can just "take out" the exposition.
Take Chade's suggestion that sliders over time shouldn't be representable on a spreadsheet. That's easy. What happens, though, is that things happen and the player doesn't understand why. Happens to me a lot in Headliner, honestly, which has a bunch of hidden interacting sliders with results that aren't always obvious, especially in the non-extremes. Imagine KotOR without the big light/dark slider. You'd be overrun with people wondering how they ended up as a Sith "despite" consistently siding with the "good guys".
In Glade Raid I've taken to just outright displaying the dialog-result statistics. I want the players to understand the connection between their decisions and their long-run consequences as much as possible.
Chade on 27/10/2017 at 00:25
Quote Posted by Pyrian
Imagine KotOR without the big light/dark slider. You'd be overrun with people wondering how they ended up as a Sith "despite" consistently siding with the "good guys".
Well, in KoTOR
2, perhaps ... KoTOR is just too straightforward IIRC.
Quote Posted by Pyrian
Take Chade's suggestion that sliders over time shouldn't be representable on a spreadsheet. That's
easy. What happens, though, is that things happen and the player doesn't understand why. Happens to me a lot in
Headliner, honestly, which has a bunch of hidden interacting sliders with results that aren't always obvious, especially in the non-extremes ... In Glade Raid I've taken to just outright displaying the dialog-result statistics. I want the players to understand the connection between their decisions and their long-run consequences as much as possible.
That sounds hard, not easy ;). Most things are easy if failure counts ... anyway, I get that it's really hard to make this work well, and I assume that's why we haven't done it yet. But if the goal is to develop believable conversation systems then you can't get around the fact that this implies my mental model of the character has to be a character, not a couple of variables.
That said, maybe if the characters had more dramatic actions at their disposal and my interactions changes with them as the story developed, maybe it wouldn't matter so much if my mental model of the characters was pretty simple. At least I'd be thinking hard about the consequences of my interactions, even if the actual course of the conversation was fairly predictable ...
demagogue on 28/10/2017 at 07:50
Firewatch had a good conversation system. It wasn't so much gameplay as much as the player was trying to direct the conversation in interesting directions.
Having multiple routes is better for storytelling purposes than gameplay purposes perhaps.
