Panel discussing AI directors and player experience management.
Jurie Horneman, Stephane Bura and Gwaredd Mountain.
Somewhat interesting discussion on the use of experience management, although mainly interesting things were from Gwaredd talking more about Silent Hill from his other talk.
Question (Gwaredd) If you could profile different player types could you improve the game around that?
Stephane: With achievements for instance – the metagame has this. But it would be cool to get it in elsewhere.
Question: Jurie, how does it involve the online space?
Jurie: Metrics in game design with social games, funnelling players around, needs constant analysis of player data and testing.
Question: How difficult is the problem? can we solve it?
Stephane: We’ve been doing that since the beginning of videogames. The game understands when you press the jump button you want your character to jump. The game understands walking slowly means silently. Already done all the time; very easy. Only complicated thing is the amount of time we want to spend doing it. We just have to do it.
Gwaredd: Tomb Raider only measured 4 metrics, expected many more. Profiling physical things are a lot easier – harder to analysis psychological things, and there is a lot of things we can do and some are doing this stuff which is cool.
Jurie: Saw the Half-Life postmortem about how the data of health tracking over time means sometimes there is not enough healthpacks (nb: Was implemented a bit in Half Life 2), should be done more often.
Stephane: People also feel if they are being watched. People like it, like the Metal Gear Solid achievement tracking showing save times/amounts/health used.
Question: What about the online social space tracking like Facebook?
Gwaredd: Social space collects more long term non-realtime data then change things to see if the metrics change.
Stephane: Problem is when it is done well you don’t feel it when doing real time changes.
Question: What about natural language processing to add more interaction?
Gwaredd: It is like Natal and so forth, adding more human computer interaction is a good idea. Natural language processing is where games needs to go, but we’re currently not very sophisticated so can’t understand it – if we can only recognise key words rather then inflexion it is going to break the forth wall.
Stephane: I agree, but to go past this problem we just ask him what he means rather then wait for input. Asking for feedback during or after things the player does instead of inferring it.
Phill’s comment: Psychological/physiological profiles are definitely noisy, hard to get proper arousal. Best suggestion is looking at a recording when they’re playing. Posture and attention they are paying to the game is so much clearer.
Question: In terms of the Big Five, you talked about some events that might change these; any concrete examples?
Gwaredd: There are big key events, the psychotherapy sessions mainly. Smaller indications in real time games. Could have used introversion/extroversion – not typical ones, but how they are energised rather then how social they are. Could have done more of that but too much content to create.
Question: Did any of your team think to take some people other places then fear? Taking them from greed to something else for instance? Lifting people up from bad fears?
Gwaredd: More like a serious game then entertainment; Mess peoples heads in a good way not really considered; would be hours of talk.
Question: How do you use these techniques in other mainstream or future games?
Jurie: As a game design you’re looking for what players want from your game. Lots of varied people so as a consequence need this kind of bang for the buck profiling if it is implemented.
Stephane: What it is includes more things then profiling; it’s everything the game knows about the player. This includes tutorial things like learning the game and checking they understand it.