Opinion: Making games pay attention
As the first of the entire series, it was a boilerplate mission. Break into the manor of Lord Bafford, steal his jewelled sceptre, and sneak back out. Yet this one particular Thief player approached it with a unique flair. One by one, he hunted down each of Bafford’s hired guards, smacked them with the blackjack, and dragged their unconscious bodies into the dining room where he arranged them in colourful poses: propped up back to back, legs poking out from the dumbwaiter, or draped across the length of the dinner table. Then he collected every available wine bottle and tossed them haphazardly in with the comatose crowd. Here was the perfect crime. When the Lord returned home, he’d find his precious treasure missing and his security staff in severe disarray after what was evidently a night of debauchery while the boss was away. What would his reaction be?
Of course his reaction was nothing, at least not in the context of the actual game. I used this story in my GDC presentation about player expression because it’s my favourite example of what it means, and where games could respond better. To me, player expression happens any time the player makes a choice from a range of equally valid options. It includes customising your avatar, like in Rock Band, or your car in Need For Speed. It includes building insane creations in Minecraft, picking your character class or choosing dialogue. I’m drawn to cases where players select something suboptimal for gameplay, like my friend Rob who dresses his Fallout 3 character in a ghoul mask and an Abraham Lincoln top hat, even though he could be wearing a combat helmet. And I love the Bafford example because players weren’t even prompted toward that type of expression in Thief. This player took it upon himself to craft a perfectly appropriate continuation of the narrative using the tools available.
So what should we, the designers of that game, have done? Should the player have received a special mission debriefing describing how Bafford had fired all of his guards in a fit of rage? Let’s talk code for a second: would that even be possible to detect? That exact instance, sure, but what if the player had missed a few wine bottles, or only knocked out two guards? What if the player had posed them scandalously in the boudoir instead? Or hung all of Bafford’s unmentionables in a hallway where guards patrol? Could you detect those cases, and any others the player might invent? The problem is the semantic meaning carried by the objects in question: guards, wine, dining rooms and lacy undergarments all mean something specific in the context of a secure manor when the owner is away for the evening. This is more meaning than game code has historically responded to, but that’s what makes it an exciting area for innovation. The sound effects in Thief were tagged with semantic data relevant to stealth: how suspicious they were, or whether they could plausibly have been generated by a nearby ally, general information the AI could reason with categorically. You could imagine extending this approach so the game knows about what behaviour is appropriate for guards, or what makes Bafford angry, or what combinations of things and places are especially perverted. This is clearly a lot of risky R&D effort for something explicitly off the path of expected gameplay.
And anyway, is a custom debriefing really what the player wants? It reminds me of an Achievement, a hypothetical ‘Made It Look Like The Guards Got Drunk’ Achievement. And the problem with Achievements in cases like this is that when the player feels imaginative, you show them that you already thought of the clever thing they did. That crushing disappointment of being just another sheep in the herd is the bad part, but the good part is the little thrill you first feel when you realise the game has noticed and acknowledged you. Which implies the crux of the challenge: respond to the player’s expressions without invalidating their sense of ownership. Stats maybe work better, like GTA’s ‘Fires Extinguished’ or ‘Longest Flight in Dodo’, which serve as yardsticks the player can use to measure accomplishments of their own invention. If you had the tech that could notice the Bafford case, you could respond not with a debriefing but with a new stat: ‘Guards Fired’. Maybe then the player would feel noticed and also intrigued. How can you max that out?
The problem can be simplified by shifting to a context with less semantic meaning, such as a cave on Mars filled with unearthly lifeforms, which describes Tiger Style’s in-progress project. Whether the player is able to share their amazing discoveries with humanity depends on which objects they are able to bring where, and most combinations aren’t meaningful and therefore don’t have to be supported. I hope this turns out to be interesting, but like the ‘Guards Fired’ stat, it’s far from a perfect design. Maybe it’s at least a step into uncharted space where games work differently to collaborate with their players.
Randy Smith is a game designer and principal at Tiger Style, which has its first game, Spider, in the App Store now. See and follow all his columns on his topic page.
Illustration: Martin Davies