The most recent post, “Friends Like These”, on Duncan Fyfe’s Hit Self-Destruct concerns itself with the believability and limitations of NPCs. He uses a game that’s fresh in all our minds, Fallout 3, as a means of example, first considering the general NPCs scattered throughout the D.C. Wasteland, and then specifically the various companion NPCs, both human and non, that can accompany the player.
The crux of Fyfe’s argument is such:
Any NPC is believable to a point, and as soon as their scripted routines are disrupted all the flaws become quickly apparent. Bethesda largely prevents that from ever occurring, except, inexplicably, in this case. It’s a technical issue. The companions can’t be programmed to exhibit a convincing array of responses to all the emergent possibilities conceivably generated in an open-world playground. Games aren’t able to simulate human behaviour at the level which Fallout 3 requires to be consistently credible.
This is a point that I’ve personally been struggling with for some time. The question that remains is one I posed to Fyfe: Is this truly a technical restriction or a limitation inherently present in the medium itself?
Let’s imagine, for arguments sake, that we can create a game with perfectly photorealistic graphics. Let’s also imagine that any and all NPCs in said game have somehow been programmed to react realistically and with a full range of responses to anything the player says or does to them. Of course it’s debatable whether or not we’ll ever be able to accomplish either, but this whole scenario is purely hypothetical for a reason.
The characters we interact with in this imaginary futuristic game, will we somehow treat them differently than how we’re inclined to treat NPCs currently? They may be capable of exhibiting a full range of human emotion, but can we actually empathize with them? Or love them? Or hate them?
My guess is no, and that it’s an inherent limitation in gaming. We will continue to recognize that these NPCs, no matter how believable within the confines of the game, are still non-sentient. They may be programmed to have a rich and extensive backstory or to interact in a meaningful and enduring way with the game world or the player’s own story, but we will never consider them even remotely human. It’s because they exist in the world only during the time when we’re actually interacting with them; we don’t really care what they’re doing during the moments we’re not talking to them or during the moments we’re not trotting off together into combat.
I realize it’s probably in poor form to conclude by contradicting myself, but it bears mentioning that we have begun to reach a level of believability, even sympathy, with our NPC interactions, as Fyfe himself points out. How many of us refused to harvest the Little Sisters? Or felt remorse over killing Colossi? Or sought to find the most diplomatic solutions to Fallout 3’s moral quandaries?
Are we all being conditioned in some sort of modernized Milgram experiment? (Not so much Milgram’s hypothesis with regards to the acceptance of authority, but more of the experimental method itself: could a human inflict pain on a completely believable AI?) When will a game developer force us to torture or kill an NPC we’ve grown to “love” over the last fifteen hours of play? We have games in which the player is penalized for acting evilly. When will we have games that punish the player for making a decision that’s morally right?
I accept your hypothesis to an extent. I agree that there’s an ultimate limitation; that games are never going to feel like Façade at the fullest expression of its potential, or like you’re videoconferencing. A game’s cast of characters will never pass for human. But that’s the case with every other kind of fiction, too. I don’t think a character in a movie is a real guy, it’s clearly an actor who’s also on the cover of People magazine, I don’t believe that the protagonist of a novel has a life outside of the book, and if I’m watching a play I can be pretty sure that if I throw something at the actors then they’ll give up the illusion pretty fast.
And I can love and hate characters in all those forms of media, and I don’t think the verisimilitude of video games is intrinsically less believable. It’s harder to maintain, for sure, because of games’ capacity for interactivity. And games are way behind right now, because they got a late start.
Also, the failure of games to reach that hypothetical standard (which we both agree is likely impossible) doesn’t matter very much. I guess it has something to do with your perspective on writing, but I firmly believe that the purpose of a character is to support the story, or the atmosphere, or the gameplay. I never read or see anything for “the characters”. When a companion character in Fallout 3 can’t react to a critical plot twist or a crazy emergent event, then they’re undermining the credibility of the world and that’s where I think the problem lies. They’re not mechanically sophisticated enough to react appropriately and so they call attention to the game’s limitations. That’s my concern. Not that they’re unconvincing as humans but that they’re tripping all over their lines and crashing into the backdrops.
Alyx in HL2 can basically pull this off, but only because HL2 is so on rails. But ten years ago, games couldn’t have made Alyx work with a game of HL2’s comparatively narrow scope. I do think this is a technical thing, largely. Valve will get better with Alyx as they go along, and developers will eventually create characters complex enough to support a game as broad as Fallout 3. In the meantime, you can see these occasional gaps between reach and grasp (the “uncanny valley of ludonarrative dissonance”, as some call it.)
In regard to your last paragraph, right off the top of my head KOTOR and Mass Effect did exactly the things you suggest. I would bet there’s a lot of developers that have been thinking along those lines for some time.