Connect the Dots

Let’s talk about stories.

The Archival Story

Imagine this: you’re a historian in the far future. All of the records of our era have been lost.

One day, an archive full of records on population, births, deaths, hospital records, industrial output, temperature and rainfall records, election tallies, price records and migration records is found. It seems like a historical windfall. There’s only one catch: it’s all just raw data. Numbers, categories, and names. There isn’t a written sentence anywhere. No emotions or idea. No record of cause and effect. No interpretation.

Is this archive a story?


A story is a sequence of related events. All games have sequences of related events, thus all games have stories. Even abstract games like Tetris have stories.

That time when you were about to lose and barely survived when that long skinny piece showed up? That was a story. What’s special about this story is that it wasn’t written into the game. It emerged dynamically.

There is something fundamentally different about experiencing an embedded, pre-scripted narrative versus an emergent narrative. When I’m playing Half-Life 2 and Alyx comes and saves me at the critical pre-scripted moment, it’s cool, but it doesn’t feel like it actually happened. I know that these events were pre-ordained. It’s similar to the difference between watching a war movie and watching a war documentary. Even if the events are the same, the fact that one is real makes it much more poignant.

When I’m playing Halo, though, and I’m gunning from the Warthog and we get hit by a missile and fly through the air upside down, and I manage to shoot a guy using a gun attached to a jeep which is spinning upside down through the air, that is really cool. Bungie didn’t script that experience. In a very real sense, it actually happened.

This is why I love emergent narrative. They happen within an artificial space, but they aren’t forced or invented. They’re true stories, and they feel that way.

The problem is that there is something very important missing from emergent narratives.

A Ledger is Not a Story

Let’s go back to the archive. Does it tell a story? The information is there, buried in the billions of data points. But a story is more than data points. The historian’s job is to sort through all this data and try to identify cause and effect relationships between the data points.

This is the kind of problem that games have with emergent storytelling. We have no historian. We don’t interpret anything or pick out relevant data. Emergent narratives in games tend to be shallow, confusing, anticlimactic, meaningless. The narratives tend to be short action vigniettes or data-heavy strategy reports. Never does a game emergently generate a clean narrative arc in the traditional sense of the word.

If you’re playing Civilization, your faction can conquer many enemies, resist invasion and cultural domination, trade, gain allies and betray them, and ultimately rule the world. But the game doesn’t emphasize the importance of key battles. It doesn’t tell you that a particular trade agreement needed to be preserved so that you could maintain your uranium supply and continue your plan to build an unstoppable nuclear arsenal that you can use to blackmail a particular neighbor. These causal relationships are all there, but all the work of interpretation is left up to the player.

There’s a really fascinating site called that has a huge database of written reports about video game matches. Each report is a written and illustrated story explaining the events in the match. The cool thing about it is how much more interesting these matches can be with a decent interpretation than when simply viewed raw.

Connect the Dots

I think we should start trying to thread this interpretation into the game design itself. The game should tell a story together with the player so that the player understands what’s happening in the game. Every action and reaction should be imbued with a sense of meaning and emotional gravity, and should be connected to everything else inside and outside the game. We need to embed an artificial historian in the game,who can pick out meaningful data and cause-and-effect relationships.

Doing this with past events is called storytelling. If you are playing Civilization, and you are at war, and your enemy captures your uranium source just before you finish building your nuclear weapon, which leaves you defenseless against their onslaught, the game should tell you that you lost the war because you lost the uranium, which deprived you of a key strategic resource. As it stands, the cause and effect are there, but the player is the one who has to extract all the information.

We can also connect the dots into the future. This is about identifying important events as they happen. This is hard because the same event can be either world-changing or completely unimportant depending on context. In Starcraft, if a Marine dies in the middle of a massive, pitched battle, it’s not a very significant event. If that Marine is the only defense against an early-game rush of two zerglings, however, his death is a significant event because we know that it will leave the base defenseless, which will allow the zerglings to do significant damage before they can be destroyed, which will place the Terran player at an early disadvantage which will be very hard to reverse, likely causing him to lose the game.

Imagine a Civilization game where one of your allies comes to you and says, “You need to send troops to help me in my war against the Germans because if they capture my uranium supplies they will build a nuclear arsenal and end your nuclear world domination.” Imagine he begs or cajoles or threatens the player. That’s emergent storytelling, and I’m looking forward to it.

6 thoughts on “Connect the Dots

  1. Aubrey

    Really enjoyable post.

    In a sense, though, there’s no “correct” story for the historian to tell. It’s like the fallacy of an objective documentary: The simple act of choosing one fact over another results in a bias. Only the raw data is objective.

    That could work to ones’ advantage, though: It would be quite interesting to see multiple historians, each with a built in bias (kind of like Civ/SimCity advisors), each using the set of game events to try to push their personal world view. The Right Wing lobbyist, telling you that you lost because you didn’t enage in the arms race properly, or a Liberal who says that you should have tried to improve relations with the country that attacked you. At that point, the player is allowed to choose which take on the history that they enjoy the most.

    Choosing one story out of the data makes it sound a little like there was a degenerate strategy in the game? Perhaps not. Perhaps everything, during the course of the game, packs up and weighs heavily on one event in a purely emergent fashion. I guess that could happen. In fact, it’s not hard to contrive it to happen without resorting to straight-up scripting. You could do it with a systemic pattern which inevitably builds up to this point, just as rain inevitably reaches the sea. I just don’t know that, if you want an emergent game experience, it would be wise to make that event something systemically inevitable. Or do you? I dunno. Maybe there’s no right answer to that.

  2. Tynan Sylvester Post author

    Indeed, I thought about noting the alternative interpretation aspect of it. I think this ambiguity is what makes the problem so hard. It seems to require the type of judgment that machines lack. I think that for now, we can start with baby steps. Right now I’m willing to accept any interpretation of game events at all, seeing as we currently have basically nothing.

    The closest I can think of are the newspapers at the end of missions in Hitman: Blood Money, which were very simple in their interpretation of mission events.

    Interestingly, I hadn’t thought of this before, but part of what happens in Blood Money is inevitable, just as you say. At the end of the day, you always kill the target somehow. The article is structured around that assumption.

    Most games have a few clearly definable states (won, lost, tied, lived, died, etc) with many small variations within those categories. It might be worth organizing an AI storyteller around those broad game states, seeing as you will inevitably be inside one of them.

  3. Aubrey

    I actually had an idea like that a while back on a project I can’t mention. The game had squad mates. The player vocabulary was wide enough to allow a range of approaches to given missions.

    Squad mates had fairly polar personalities, personifying that idea of sets of subjective points of view, as opposed to having one didactic voice telling you “you dun good” or “bad player, no buscuit”.

    That allowed the player to carve their own expression into the game, and have the narcissistic payoff of being recognized for being, say, sneaky, as opposed to blood thirsty.

    Hitman (and other stealthy games) tend to heavily reward a “perfect” run, which for me is a bit of a missed opportunity. If Hitman’s actual combat were as fun/challenging as the stealth/adventurey aspects of Hitman were, you could absolutely give equal reward and credit for taking a mission in the most psychopathic way possible. As it is, you’re only rewarded for playing your role… which is fair enough, I guess, but I think that I prefer it when I’m not made to feel like I’m doing the “wrong” approach, even though this “wrong” role playing of the character might take just as much skill as the “right” one.

  4. Aubrey

    For give the typos – exhausted right now.

    What I was trying to get at is that morality in games is often in the eye of the designer, and when it’s a mismatch with the player’s take on the world, the message may come across as overt propaganda. KOTOR has some pretty clear signs of this – occasionally you’ll think you did the right thing by someone by telling them a white lie, but end up with dark side points because it was still technically a lie. In positions like that, who is to say who should be the arbiter of morality? Really all you can do is side-step the issue:

    If you have enough facets of morality represented in the game, each espousing different world views, players always have someone they can identify with, and always have a way to feel like their individual, unique expression (their emergent story) is being recognized and fed back from game. A news paper might only tell the gruesome facts (“If it bleeds it leads”), but perhaps some other foil could reflect the positive things you did (like the cutscene for saving the prostitute in the ChinaTown mission). You need to give that treatment on every axis which feels like a legitimate form of player expression. But that, in of itself, is a subjective call – the bigger the verb set, the more likely it is that you might, due to limited resources, or un-tameable emergent complexity, have to miss some of the more niche approaches (speed runs? play throughs without using health packs? It’s hard [practically impossible] to give every single approach its due).

  5. Tynan Sylvester Post author

    >”If you have enough facets of morality represented in the game, each espousing different world views, players always have someone they can identify with, and always have a way to feel like their individual, unique expression (their emergent story) is being recognized and fed back from game.”

    That’s relativism, still a worldview in itself.

    Anyways, I don’t think it’s really possible to satisfy every viewpoint. Hell, I usually can’t find people to identify with in real life, much less a work of fiction. And real life is the product of billions of minds interacting over years.

    In a game, we’ll still need to leverage the player’s own powers of interpretation as much as possible. I just think we don’t need to completely depend on them, as we generally have up to now.

    Also, if we cue players as to what kinds of behaviours the game will interpret, it would help them act out stories within the (broader than usual but still present) limitations of the story interpretation system. Have characters talk about doing things fast, slow, without health, without being seen, and so on.

  6. Aubrey

    “That’s relativism, still a worldview in itself.”

    I’ve been out-post-modernized!

    You’re quite right, of course.

    And yes, I love a game whose story aspirations are born out of the pure mechanics of the game, rather than working on some untouchable layer. There’s Mr. Levine’s excellent example of how he wanted to make the story for a Thief Mission be about the Rope Arrows themselves. I find that sort of thing is an excellent approach, and at the core of what it means to fundamentally adapt yourself to become a good game writer.

Leave a Reply

Your email address will not be published.