Still Logged In: What Social VR and AR Can Learn From MMOs

Updated on

This talk was delivered on Friday, March 3rd, at the 2017 Game Developers Conference in San Francisco. It is, as the title says, about the many lessons from virtual worlds that carry through into any sort of persistent space, including the ones that we think of geotagged web services.

Slides

Video

A video of this talk is available for free on the GDCVault.

A condensed version of the presentation (just a half hour long) was delivered at the Intel Buzz Workshop Los Angeles 2017.

Download

A PDF of the original slides is available here for download: Koster_Raph_Still Logged In

Recommended additional reading

This presentation was based in large part on the material in the following essays also found here on my site:

Transcript of the full GDC version

RAPH KOSTER:

Hello. Are you all exhausted?

No? Well, okay, then it’s your job to keep me awake, because, oh boy, I’m exhausted. I’m also somewhat losing my voice, so I need you to be the loud ones this time.

How many of you have been to my talks before? A few of you. I’m known for giving these incredibly dense, brain-blast talks with 150 slides and tons of references and facts and whatever. This is not that talk—not only because I’m losing my voice, but because I agonized over how to give this talk. To be honest, I’ve rewritten it more than almost any talk I’ve ever given, and guys, I’ve been doing this a long time.

It’s something that I feel really deeply and passionately about, and something that I take very, very seriously. This is co-listed with the advocacy track and not just the design track, and it’s for a reason.

The other thing about this talk is that it isn’t entirely going to be about games. “Social” is in the title of this talk, and as soon as we’re dealing with people, we aren’t necessarily entirely in the realm of just entertaining.

Here’s the thing. I’m just going to say it bluntly: social spaces aren’t just games.

From the get-go, there are a few of you in this room with whom I remember sitting and having discussions—almost twenty years ago now—about whether we were even making games when we were making the early virtual worlds and MMOs. When I was making Ultima Online, I would have arguments with people like Eric Zimmerman about this. I told him bluntly, “I’m not sure that’s what I do. I’m not sure I make games. I think I design places and worlds and societies, and not just games. They just happen to have a lot of games in them—things that you can stumble across and play.”

A huge amount of the work was going into simulating things like tailoring, rabbit hunting, and running a shop—not necessarily what we think of as games. So I used to say I was making worlds: virtual worlds.

I really do believe that when you start working in the arena of online social spaces, that’s what you’re doing. Even though those days are long gone, and MMOs are not the new hotness anymore—they haven’t been for quite a while—if you are making a social VR space, or a social AR space, as a lot of people are trying to do, you aren’t just making a game. You are designing a society.

When you choose to take up those tools—connectivity, persistence, identity—you are going to design that society either on purpose or by accident. And, sorry to say, I’m seeing a lot of people doing it by accident.

So I’m here to try to persuade you to do it on purpose.

We’ll start with the obvious thing, because this should be a terrifying responsibility. The problem is that we’re all blinded by Snow Crash. There are still people out there—I know we have some folks from Oculus in the audience, and there are people in leadership at Oculus outright saying, “Yes, I dream of building Snow Crash.”

No. Snow Crash is not where we are really headed. It’s a great book and a really compelling vision, but we need to be aware of what we’re actually working with.

The very first time I met Bruce Sterling, a different cyberpunk science-fiction writer, I had already been working in online worlds. I asked him point-blank, “How did you get virtual reality so wrong?” It wasn’t really a fair question.

I want to tell you an anecdote that happened while speaking with an extremely powerful and famous executive at a very large blue company whose last name starts with Z and ends with “uckerberg.”

I was interviewing for a job there. I had talked with people all day long, and the last interview was with him. He said to me, “Well, I’ve heard good things from everybody all day, and everybody’s been asking you questions for eight solid hours. So what questions do you have for me?”

I said, “Well, I’d like to know your opinion. What do you think are the social and ethical implications of social virtual reality and connected augmented reality?”

[Pause.]

I may have ranted at him for a little while and then not gotten the job. But I also would not have taken it at that point, and probably will never work for them again now that I’ve said that. That’s okay.

Let’s get real for a minute. We are talking about companies that are trying to create world-spanning social VR and AR structures and haven’t done basic work on what online communities are. I often say I’m not entirely sure Facebook understands people. Adding you willy-nilly to groups—is anybody a fan of that feature?

The idea that people are out there building constructs like this without doing some basic lesson-learning is a little bit terrifying to me, given the way societies develop and the amount of power that has ended up in the hands of large internet companies. It should be a little bit terrifying to you, too.

I promised some lessons. How many of you here are actually working on building social virtual-reality experiences? Awesome. How many are working on trying to cash in on Pokémon Go and build social AR experiences? A few. Okay, good.

Let’s get real and look at some specific lessons before I go all lofty on you.

I’m going to call out a particular company, but I think its response to this incident was actually really good, so I want to preface it that way. Do all of you recall the case of the virtual groping in QuiVr? Good.

A woman logged into this game, which apparently is about doing VR archery, and a stranger named, I think, BigBro442 walked up and started virtually groping her.

The company was not logged on at the time. They did not know about the incident until an article was written. When they found out, to their credit, they immediately put in a bunch of features so that users’ bodies could not see one another anymore when they came into close proximity. They expanded what they called the “personal bubble,” and they added a new gesture that they would like to add to the vocabulary of social VR: one where you can make an exploding blue force field that turns the harasser invisible and prevents them from interacting with you.

They were very upfront. They immediately posted a postmortem about what had happened and encouraged everybody: “Hey, we need to take this stuff seriously.” Major props to them.

But let’s be straight about what should have happened.

What should have happened is that the people who were logged on 24/7 monitoring behavior were called by a harassment report, appeared with the wrath of God, turned the offender into a toad, and banned him for life.

This is actually best practice in online community spaces.

I understand that we have a lot of small companies that don’t have the resources to maintain customer support on call 24/7, that don’t have all the tools, background, and technology to be able to do this.

I don’t care.

If you host an online community where people can cause harm to another person, you are on the hook. And if you can’t afford to be on the hook, don’t host an online community.

I wish there were a different answer. I don’t think there is.

You’re the government of that space. You’re hosting real people. You should want to be doing this.

I understand that you may not quite know how. That’s fine. Let’s talk about ways you can do it.

The first thing people need to understand is that you actually need to build all of your moderation structure on due process. You need to institute a rule of law. That means starting from the very beginning with a code of law—a code of conduct. You need some kind of code of conduct, end-user license agreement, or terms of service written in plain English that every player affirmatively signs up to the first time they log in. They need to swear, in effect, “I will abide by this.”

It needs to be posted everywhere, and you need to be reinforcing this as a cultural thing every day. This is kind of what we do in the real world, and we’re not even entirely satisfied with how well it works here. But you don’t get to abdicate doing it.

It has enormous benefits, not only as a cultural thing, but also for clearing up edge cases, because you are going to be fighting an awful lot of edge cases.

The most concrete practical tool is one that has been in use for—oh gosh, I went asking, and the earliest virtual worlds I could find doing it were games on America Online and GEnie. For anybody here who doesn’t know what those are because they’re too old, this is a concept called the circular buffer.

This is a very simple thing. If you weren’t using voice chat, you could all go and implement it right now. Unfortunately, you’re using voice chat, so you’re screwed.

Simply record and keep track of the last one to five minutes of every interaction each client sees. Store it, discard the stuff that is more than five minutes old, and always enter the most recent material. Then, when something like this happens and no admin is online, when the user hits the report button, the report is attached to impartial evidence that can be adjudicated.

The QuiVr story—I absolutely believe the woman. I absolutely believe that story happened. There’s corroboration and all the rest. At the same time, believe me, you will have cases where it turns into “one person says, the other person says.” If you don’t have tools like this, you are literally going to spend your entire monthly profit on the first call, all standing there mediating between two people who are yelling at one another.

It is actually cheaper to ban every single user who calls for help. I know this sounds horrendous. Banning every person who calls for help is cheaper than helping them, because they cost so much money to support.

So you need to have incontrovertible evidence.

I have no idea how small shops are going to do this with voice logs. Worse, I think you need to do more than just voice logs. I think you need to record every single gesture, so that when somebody does this, the customer-service person can replay it and verify who started the fight.

It’s not going to be cheap. But you need to do it.

The gesture was a great touch, but banning wasn’t mentioned in the article.

A key lesson from all kinds of social spaces and virtual worlds is that offenders repeat. The number-one player killer in Ultima Online was personally responsible for murdering over 4,000 people.

The people who engage in this behavior do it over and over again. If your social VR or AR system does not have a persistent account system with persistent identity that players invest into, you are effectively letting every offender get away scot-free, because all they need to do is create a brand-new account every single time they log in.

Actions like this matter more in VR precisely because VR embodies us and makes us feel like we’re really there.

There’s a lot of talk about, “Oh, well, we’ll just block. We’ll just mute. We’ll do a personal bubble.” It’s wholly inadequate.

Let’s take a simple thought experiment. One avatar interpenetrates another avatar in some way that somebody finds uncomfortable. They do something nasty and start dry-humping the other avatar. Great. So the personal bubble kicks in and the other avatar can’t be seen. In theory, that person is now okay. It’s the same logic as, “Well, I’ll just block them on Twitter and everything will be fine.”

What about the third party who is still watching?

You are going to be just as angry about this incident when you get told about it by the third party.

Block and mute have never been adequate enough tools for managing these sorts of problems in online settings.

How many of you have read the original “A Rape in Cyberspace” article?

I should ask that more precisely. How many of you working in social VR and AR have not read the original “A Rape in Cyberspace” article?

Okay. Homework item number one.

This was an article that came out in The Village Voice in 1993 regarding the first major documented case—because I’m positive there were many more—of one avatar raping another one.

Actually, he raped two in public, in the living room of the main house at the main entrance of the entire virtual space, with a knife—as in, raped with a knife.

Here’s the interesting thing. He didn’t even do it to their actual bodies. He did it using a technological tool available in the game called a puppet. This was a text world, and all the puppet did was make it look like somebody else was doing something. The victims were actually just standing there.

They could mute and block him all they wanted, but it didn’t matter, because it looked to everyone else in the room like Mr. Bungle was getting these two poor people to do horrendous things to themselves.

In 1993, this became a bit of a cause célèbre because people started arguing about whether or not that was really rape. “I mean, it’s online. That can’t be real, can it?”

We know better now, I hope, although we’d have to see. I think there are plenty of Twitter trolls who probably still think, “Well, what’s the big deal?”

But I’m pretty sure that right now somebody is going, “You know, it would be awfully cool to do user-generated content in social VR,” and somebody is probably inventing the VR version of that puppet right now.

Damion Schubert, another fellow virtual-world guy going way back, has this absolutely wonderful phrase: TTP. We use it as an index with which to measure the success of virtual worlds’ social design.

It stands for “time to penis.”

How long until you find the user-generated-content penis that some player has managed to stick in your game?

In Ultima Online, the first one I saw was drawn by dropping fish on the road. It happened on the first day.

And, of course, the most famous incident—I encourage you to go look up the video on YouTube; it’s astonishing in a surreal kind of way—was when poor Anshe Chung was literally attacked by a flying swarm of thousands of penises that swirled around her in a tornado.

That isn’t entirely on the troll. If you run that world, that’s on you.

Every feature you make should be looked at first as a weapon. It’s a dark way to look at humanity, I know. I’ve grown so cynical.

How will a player abuse someone else with this?

Odds are excellent that you aren’t actually evil enough when you playtest. Otherwise, something like groping somebody would have come up in the first fifteen seconds. It’s almost an obvious scenario. What could they do with the actual bows and arrows? We haven’t even gotten to that part.

It wasn’t long ago that we saw an experienced AAA team miss one of these basic things. How many of you played The Division?

Here’s this MMO shooter on the console with a AAA team. They made all the players collide with all the other players, and then they had a single quest-giver sitting in the central lobby. So there were lines running out the door, and people would barricade the door to prevent other people from getting quests.

My personal experience with making that mistake was in the alpha for Ultima Online, sometime around February to May of 1996. We allowed people to spawn in the inn, and then they would walk outside. They were collidable, of course.

People would gladly let them out the door, but then they would stand at the door and not let anyone in. In the pursuit of realism, we had made it so you could shoot arrows through the windows of the inn. So they would let you out, rush to the windows, and shoot you in the back as you walked away. And, of course, you couldn’t get back in to defend yourself. You couldn’t see who was in the building.

We basically proved all of the fundamental premises of medieval defensive architecture.

That was 1996, and a game in 2016 can still make the same mistake because we’re not necessarily very good about sharing the history—or reading it, rather. The stories have been told. We laugh about them. They’re fun, until you’re a pincushion with a hundred arrows in your back.

This brings us to another issue.

Those of you who are building social VR: you are here listening to this talk. Excellent. You are good people. You are not going to do these things.

But if you are successful, the people you hire will hire other people who will, in fact, do things like consider it a perk to go over to people’s private chat spaces and watch them as they engage in very private things.

The very first guy that I banned and fired as an admin on a text MUD was somebody who thought it was the perk of being a god to be able to spy on same-sex couples who were having cybersex in their private rooms.

You have to hard-code limitations into the architecture of your world: permission bits, absolute logging of every kind of extraordinary command, the ability to grant or revoke any kind of capability granularly per individual, and only grant the powers that are specific to the scope of that individual’s job.

You have to do this because, in the early days of MMOs, we had issues where game masters in Ultima Online were having hot-tub parties with players and trading sexual favors for in-game items.

They needed the ability to spawn those items for their job. Unfortunately, we didn’t have logging in the hot tubs. Probably fortunately, actually. So it was hard to prove that these situations were happening.

The thing is, the odds are near certainty that, if your virtual world is successful, your digital objects are going to have real-money value.

There is no escaping it at this point. It is absolutely hopeless to try. Just assume that anything you create as a digital object has real-world value, and people will be doing things like sleeping for it, paying for it, and stealing it.

Go in the door assuming that this is part of the field.

Fundamentally, you should not be opening a virtual world unless you think in terms of this kind of legal infrastructure.

The easiest way to think of this is: you are the gods of the space, but you are also the government of the space.

We have drifted away over time from having in-game admins with visible avatars in the space. The reason is to minimize the possibility of personal relationships forming between customer service and individual players.

I miss the days when you could hang out with the players, and you were the god and they weren’t, and you could hang out and do things for them. But the reason we have anonymized everybody is because corruption happens.

You also need to be aware—and this is just a home truth—that the people who use virtual spaces are very often using them for therapeutic purposes.

That means a higher incidence of individuals inside your space who are dealing with psychological stress of various sorts. It’s just how the numbers work out. You need to be aware of it.

There is not a single large-scale commercial virtual world that does not have the FBI and suicide hotlines on speed dial.

The number of runaways that we tracked down for the FBI and the police while at Sony Online was easily in the triple digits.

I would add to this: be prepared to deal with manifestations of mental illness. Those folks will frequently be calling for assistance within the virtual setting.

You have gone and built their home away from home—the place they use to cope, the place they are using as their support system. Often, the friends they have made in your space are the ones who literally keep them alive.

You are holding their lives in your hands, so you have to take it seriously.

Have I scared the hell out of you yet? I hope so. Right? I hope so.

Don’t worry. It won’t all be a downer, I promise.

The flip side of this power is that you can shape this environment in ways that are amazing.

The original MUD—Richard Bartle actually designed it as a blow against the British class system. That is why you can level up and be anyone and anything. He felt like he had grown up in a very restrictive world where, “Oh no, you have the wrong accent. You’ll never get a doctorate.”

That’s amazing. That’s super cool.

In fact, a lot of the earliest virtual spaces were safe spaces for queer people, safe spaces for the asexual, safe spaces for people who preferred the idea of having hundreds of little tentacles for genitalia. They were all colors and shapes and sizes and ages—all of those identities.

One of the huge, huge mistakes of social VR is that people are forgetting the promise: be someone you aren’t.

That has always been a fundamental premise of engaging in a virtual world. Even in normative modern MMO worlds, cross-gender roleplay is still incredibly common. We all know this.

Identity exploration is one of the fundamental things that alternate realities let us do.

When you choose a class or a race or whatever in one of these online worlds like WoW, you’re effectively almost engaging in a personality test. People tend to choose archetypes that resonate with and relate to how they visualize themselves.

This is why I hate voice chat, because it’s one of the many things taking these tools away from people.

It extends beyond voice chat, though. How many of you are short like me?

All right, short people. You guys know that you have lower lifetime earnings than tall people, right? You get fewer promotions and lower salaries.

It turns out that halflings and gnomes in fantasy role-playing games earn less XP per hour and level more slowly, because we import real-world biases into these worlds. Not intentionally. Just the fact that we made halflings short results in them earning less XP, because people have biases.

So, as we chase realistic avatars, we should consider whether realism is really a good idea—whether those are behaviors and beliefs that we actually want to reinforce.

One of the beauties of the early internet, and of virtual worlds in particular, was that nobody could tell you were a dog. Only now we can hear you barking in Discord, and we know.

That’s an issue.

The most obvious of these tend to be related to gender presentation, but race presentation is another one that is frequently overlooked. I have lost count of the number of social spaces that don’t have race as a customization option.

That is embarrassing.

From the get-go, you should be thinking about it.

It was controversial in Star Wars Galaxies for us to have a breast-size slider. Fortunately, female players pitched a fit when they found out it was controversial, because it’s important. People need to feel like they can be what they want to be and be who they are.

We had to rename it the “torso slider,” though, because LucasArts was very touchy about the terminology.

Probably the biggest bias present in social VR is around empathy.

There are multiple challenges here. As people enter a space, if they don’t have a stake in the space—if they don’t already have persistent identity in it—it is very easy to behave as a virtual sociopath. They don’t have any community ties whatsoever.

So you have to start creating investment in an identity really early on.

Those of you who are making systems that are drop-in: don’t. You have to start doing persistent identity, because if you can abandon an identity, that’s basically a license to do whatever you want and not get stopped.

You can’t use things like credit cards to block people. The average U.S. household, I think, has thirteen credit-card numbers. That’s an awful lot of shots at harassing somebody without getting caught.

The real way you have to do it is by causing personal investment in the identity they are creating, so that they don’t want to lose it. They don’t want to lose the investment they have made.

The other big thing you need to do is increase channels for conveying emotional content between avatars and between players.

Again, this is a place where voice kind of cuts both ways. There are two classic ways of doing this. One is gameplay that involves teamwork. The other is an emote system.

Simply put, best-practice emote systems involve automatically parsing text chat and pulling out the emotional cues that people are already putting in there, like smileys. You know this because Facebook Messenger does it, chat does it, Twitter does it—everything does this now.

Giving people manual puppeteering structures—“Oh, I have to hit this combination of buttons to smile”—results in people not smiling, I’m afraid.

Yes, there is a subset of people who are good at puppeteering. Most people are not.

Having the system puppeteer for them based on cues has been shown to be dramatically more effective. Dramatically more effective.

I don’t know how you’re going to do this with voice. Maybe it can be a really cool research project for somebody to start developing technology to do natural-language processing and analysis of voice stressors. But I’m not expecting it anytime soon.

So, good luck to those of you working in that field. Voice does convey a significant amount of emotional content, but particularly once you get into fictional social VR spaces of any sort, you can’t expect people to be trained actors who know how to convey emotion.

It’s going to be challenging. And if what they need to do is convey real emotion, they will prefer to use Skype, where they get all the cues, rather than settling for an inferior system.

My actual advice to social VR people, bluntly, is to sidestep as many of these problems as possible.

Don’t have large-scale worlds with a bunch of strangers interacting with one another in unsupervised fashion, because I don’t think you have solutions at the current level of technology for some of these problems.

Maybe Facebook or Google will be solving these for you at some point. But right now, they just don’t.

Stick to small groups—people who mostly already know each other. Make the avatars relatively unrealistic so nobody cares about the fact that the emotional cues are wrong.

Play to the strengths. The strength of VR today is that VR is clumsy. Play with that.

I think that is part of why something like Job Simulator is one of the most successful VR experiences we have. It embraces the limitation.

Now, how many of you played Pokémon Go?

I want to say in advance: Pokémon Go is an amazing achievement. It’s phenomenal and brought joy to millions of people.

So I am now going to shred it to pieces.

All of these issues get dramatically worse once we’re talking about social AR.

From the technology side, people have started to figure out that a lot of people made this mistake: they thought the technology of VR headsets would naturally lead into the technology for AR headsets. They were wrong. It’s kind of a speed-of-light thing. It’s just not happening.

But if you are talking about social things, they are the same technology in this very particular way.

I want to take you back in time to 2006, when a whole bunch of us got invited to Stanford for an exercise with the Institute for the Future, where we engaged in a futurist envisioning project called the Metaverse Roadmap.

This was in those days when people thought it was a good idea to have Twitter feeds behind speakers. I’d love to see the one for this talk right now.

This is what we said virtual worlds would become.

We said there are four kinds. At that time, almost all of them were virtual worlds—which is to say, fictional environments that you connect to through a client: maps of places that aren’t real, where you go run around and do stuff. That box holds EVE, Second Life, Habbo Hotel, and World of Warcraft.

But we said, “You know, if all you do is swap out the data set of those maps, you instead get virtual worlds that are built using the real world, and you can log into those instead.” Those are called mirror worlds, and it’s a term that has been around for quite a while.

This is also the time when there were a lot of people mounting backpacks with computers on their backs, putting a webcam on their heads, filming everything they were doing, and having to swap out hard drives about every twenty minutes. This is called lifelogging, and yes, we saw that as being part of virtual-world structure as well. This is the identity piece—the identity component.

Lastly, and chronologically last as well, is what we called augmented reality, where these things would start to come together: where you might have a virtual world or a mirror world both overlaid on top of the real world, and engage in lifelogging as one of the ways to interact with it.

Well, all of these are now here. All of these are now, in fact, here.

The thing is, even though the operators of all these different services don’t know this, really, they are all running against what I would call virtual-world servers—virtual-world back ends.

It’s just that Twitter only bothered to build the chat part, Waze only bothered to build the moving-around part, and Yelp only bothered to build region markers.

One of the key lessons of virtual worlds—and it’s one that a lot of virtual-world players refuse to accept, and I think a lot of people who work primarily with headsets refuse to accept—is that rendering is by far the least important part of this entire field.

Sorry if you’re a graphics programmer. It is easily the part that matters the least.

Every single graphical world I have ever seen has offered a plain-text client. Why? Because the admins need it to go look at logs.

The worlds are on the server. They are representation-independent.

We might have a canonical client, but the representation is not the part that really, really matters. What matters is the action going on on the server, which is where you actually track things like avatar profiles, histories, the location of objects, how they interact with one another, and what their behaviors are.

So all of those things are actually virtual worlds.

I call this the goggles fallacy, because it is distressing—perhaps inevitable—that the current boom in virtual reality is being driven by the hardware.

If you actually go look at what is going on in these various settings, they are all about clients moving on servers, interacting with objects on servers.

Sometimes we drive around on those servers by driving around in the real world. That’s Waze and Google Maps. It’s actually every Android phone. You guys know that, right? Android phones are part of how Google detects where there are traffic jams. Your phone not moving tells them.

Yelp and Zillow are all about teleporting around on the virtual side but not the real side. It doesn’t matter either way. There’s an avatar moving on the virtual side.

So it’s the same technology.

You can literally run down the kinds of technological elements required to build a virtual-world server.

Does it simulate a space? Yes. Check, check, check.

Where is the data from? Is it real or fictional? Doesn’t matter. Who cares? It could be half and half.

Does it track clients? Are they connected? It doesn’t need to be a constant-on connection. It could just be some of the time. Check, check, check.

Are clients positioned relative to one another spatially?

The main difference is that in WoW and Pokémon, we actually bother to spawn bots in there and let you interact with them in some fashion. But that’s about it. And, of course, these days we allow people to spawn geotagged markers into Google Maps, so we’re kind of letting people spawn bots there, too.

I’ll say it bluntly: an old-school text MUD from 1992 is closer to Snow Crash than anything that has been built with goggles, by quite a lot.

A lot of defenders of VR start talking here about how the sense of presence that exists inside VR is what makes the difference.

They’re wrong.

Telepresence was defined back in the days before we had pictures. People reported those sensations in pure-text environments.

In fact, if I had to rank the level of presence, I would probably put VR first, then the text games, then all the ones on monitors, because text games actually embody you almost as powerfully as VR does.

If you are looking for more specific, concrete lessons about how to deal with avatar-intersection issues and all the rest, I would look at how text MUDs handled things.

For example, consent systems for paired emotes were a very, very common design trope in text MUDs—one that also appears in modern-day LARPing—where you can’t just hug somebody or grope somebody. There’s almost a two-factor authentication process players must go through in order to engage with others.

That stuff popped up in the text worlds first because, in the text worlds, there was no distinction of distance. Everybody was standing on the head of a pin in each individual text room. So all it took was one word for them to be interpenetrating with your avatar.

The takeaway from this is—how many of you have heard crazy stories about EVE, or the plague in WoW, or any of these things? You all have heard crazy stuff like that? Awesome.

In a post–Pokémon Go world, expect that stuff to start happening in the real world.

All of those things will happen in the real world because all of these technologies are really just virtual-world servers.

It is not going to be very long until there is a template ID associated with your washing machine so that, when you look at it, you will be able to pull up the manual. You can do that crudely right now with Google, of course. But we are not very far away from it actually having an individual instance ID—your washing machine—and it will have its own character sheet, with its repair history.

All it needs to do is be an object in the runtime database of our world.

That’s not far-fetched. That’s not science-fictional.

The problem is: we never trust the client.

What happens when you are the client? Because you are going to be the client. Well, your phone is, anyway.

This is most relevant because of Pokémon Go. I’m going to pick on it. I can’t help it.

I already did a very detailed mathematical breakdown of why its advancement system was utterly broken, but I did that back in Austin. You can find it on my website. It’s called “Systems Design for Retention.” If you like graphs, it’s full of them.

The real question for us in this room is around things like: there are no player-killer switches in meatspace.

If you start designing your social structures in a way that incentivizes players to commit virtual violence against one another, they are probably going to start doing it for real.

All of these things can be quite unintentional.

Pokémon Go, I’m sure, absolutely did not intend for its map of spawns to be racist and classist. And yet it’s really, really hard to find Pokémon in poor Black neighborhoods and in rural areas, because its user-generated-content map happened to be built out of kids who were going to Stanford and Berkeley—technologically advantaged college students.

What starts happening? You all heard: “Pokémon spawn in the Holocaust Museum. Pokémon spawn in Arlington Cemetery.” Sure.

None of this happened on purpose. I don’t ascribe any malicious intent. But nobody ever means to do this stuff.

Nobody ever means to damage somebody’s actual equity in their home by accidentally putting Pokémon spawn points or gyms in their privately owned backyard, thereby causing an endless stream of trespassing. And yet it happened.

We do have to think about these issues.

It might be that someday legislation will appear that goes right alongside water rights, mining rights, air-traffic rights, and virtual-layer annotation rights. That might actually be the only solution to something like Pokémon spawning in your backyard.

And, of course, then somebody will license Pokémon rights to a company for money.

“Yeah, I struck oil. I struck Pokémon.”

It starts raising questions like: can or should a company like Niantic strike a deal with Coffee Shop A as opposed to Coffee Shop B, thereby tilting the business balance in that area?

There are no rules or regulations around stuff like that. They could direct traffic somewhere it isn’t wanted.

What if somebody creates a really cool social AR game and says, “Wow, it would be really fun. You know, I’m actually a secret GamerGater. I’m going to start spawning Pokémon on top of a Black Lives Matter rally.”

It is trivially easy for an operator of a virtual world to pull a stunt like that.

We’ve seen the early harbingers of things like that. Anonymous got its start by holding a flash mob in front of a Scientology facility in New York City.

We kind of trained people to do this.

All those Anonymous kids—we kind of built them. We’ve been training them how to do this stuff.

Phrases like “geotagging” mistakenly give us the impression that we’re sticking a little bit of virtual onto the real, when what is happening is actually the exact opposite. We are sticking a little bit of real stuff into what is already a primarily virtual environment.

The harbingers here get real.

I said earlier that a character choice in an MMO is effectively a personality test. But so is a collection of your likes on Facebook, because Facebook is really just an extended character sheet.

And sure enough, that’s why there’s a company called Cambridge Analytica that data-mined that data on 220 million Americans and used it for targeted micro-advertising in order both to suppress votes and encourage people to come out and vote, and, you know, get us Brexit and Trump.

Because if you run a virtual world, that is the kind of panopticon power you have.

Fake news is an instanced dungeon.

It’s an instanced dungeon personalized and tailored just for you. You get your own spawns, and there’s Alex Jones as the boss at the bottom of the dungeon, and you get to play lizard people.

That’s the power you have when you’re running a virtual space.

You all saw the mobs for Pikachu. I have no idea whether it was a conscious decision in that game to make it so that each spawn of a Pokémon was instanced. But if they had not been—if it had been zero-sum that day in Central Park or around the Santa Monica Pier, and there was only one Pikachu to have—I think we would have seen Pika-murders by now.

It’s human nature.

This sounds crazy, but it is not. I know it because it already happened in Ultima Online.

When we launched in Hong Kong, the Triads joined the game. They would go and do their player-killer war in the game, and then they would take it to the streets.

By designing our game in such a way that it encouraged tribalism, encouraged othering other players, and then encouraged violence between the groups, we encouraged them to do it everywhere.

What happens when the real world is the playing field?

My local gym is a locked church. They’ve had a lot of problems with people parking in the parking lot at 3:00 a.m. and getting as close as possible to the stained-glass windows to try to catch a signal so they can zap a Pokémon inside.

How long until they break in? I don’t know.

The phones have all of this stuff built in. They do. It’s built into the phones. But the phones at least have privacy chips. The phones have private storehouses. Even Apple can’t break into its own phone, right?

But just yesterday, a cloud-connected stuffed toy—I don’t even remember the name of it—turned out to be storing the voice-chat logs of all the kids who ever talked to their plushy on an AWS instance in a database that was not password-protected.

We don’t even need the actual user. It’s enough just to get enough users to see the perturbations of what’s around them.

Are we prepared to community-manage downtown San Francisco when we have better history tracking of what each individual does than the police or the government do?

Do you want to manage community for downtown San Francisco?

We couldn’t even handle a bunch of kids angry about supposed ethics in video-game journalism. We’re supposed to deal with San Francisco?

But those are the tools we’re taking up.

It doesn’t take away from the achievement of Pokémon Go. It’s possible to achieve great things and still open Pandora’s box.

But we do have these lessons. We do know that these are the things we can be thinking about.

I realize I’m here at GDC and I’m talking to you about stuff that does not seem very game-like at all. But I’m talking with you about it because you’re my peeps. You’re the ones who will listen.

I can’t get a Google or anybody to listen to this. Well, if there’s anybody from Apple here, call me.

We’re going to be playing logged in whether we signed up for it or not. That’s your takeaway. That’s really what this is about.

We usually call entities like that governments. It’s not an accident that we talked about governance clear back in the text MUDs.

So, yeah: monopolistic giant governments.

There’s a little bit more awareness of that since I submitted this talk months ago. Zuckerberg actually said, “Oh, maybe this news thing is an issue.”

And we do have our first gold farmer in the White House. Steve Bannon used to run IGE, the gold-farming company, and then he turned around and built a gaming website network. Those were his training wheels for running Breitbart.

So some people are paying attention.

“When a time comes that new modes and venues exist for communities, and said modes are different enough from the existing ones that question arises as to the applicability of past custom and law, and when said venues have become a forum for interaction in society for the general public, regardless of the intent of the creators of said venue, and at a time when said communities and spaces are rising in popularity and widely exploited for commercial gain, it behooves those involved in said communities and venues to affirm and declare the inalienable rights of the members of said communities.”

August 27, 2000.

To my knowledge, exactly three virtual worlds in history have bothered to adopt that Declaration of the Rights of Players.

But the need for it is now greater than ever.

I really wish Pokémon had adopted it, and I sure hope that those of you who are going back to your companies will seriously consider it.

Please update it. Revise it. But start thinking in terms of the rights that players should have in these spaces.

Soon enough, the Internet of Things is going to hit. This is also not fiction.

I had this conversation with Zuckerberg, too. This is why he bought Oculus: to turn the world into a unified database of objects.

I’m probably not getting any more consulting gigs there. Oh well.

We’re actively collaborating in this process. All that’s missing is unifying the data: Facebook owning identity and chat, Amazon the object-template database, Google the runtime-template database, Apple the client.

We’re collaborating in this process, and what comes next—I don’t know. I honestly don’t know. That’s why I’m here.

I need your help. We need each other’s help if we’re going to deal with this, because this is not hypothetical.

Do not let our world be designed by the people who prefer the lobby of the W to the IGF.

You guys don’t go to the lobby of the W, I hope. Right?

Technologists in this room: you guys know how much of this we can build right now. Bluetooth LE beacons. Some simple mesh networking. We could build this now.

So I wish HoloLens and Magic Leap had that emblazoned on the eyeballs instead of demos, because the chip that matters is the networking chip, not the graphics chip.

That is what keeps me up at night.

I’m sorry. GDC is not supposed to be a downer. If we’d all gone to the IGF, I’m sure we’d all be leaving inspired and psyched up.

But this is the advocacy track.

I’ll say this: this room is the one that gives me hope.

How many of you have played a game in the last twenty-four hours with a stranger?

How many of you feel games are for connecting?

How many of you feel games are about fairness? They’re supposed to be fair, right?

How many of you think games are for being smart?

And games are for everyone?

Yeah. Good.

That’s what gives me hope, because entertainment has always led the way on this stuff, and that is a pretty good ethos with which to start.

This is a group that is willing to read giant strategy guides. So please go read the work. Read Julian Dibbell. Read Nick Yee. Read T.L. Taylor. Read Thomas Malaby. Go read all of the endless stuff that has been written about this.

Because I’ve got to tell you: virtual worlds are the best-documented part of all of video games, and the most studied, probably by an order of magnitude. The material is out there.

Frankly, I’m leaving our collective fate in your hands here.

These are your tools. Now you know what you’re actually doing. You are not making cute archery simulators. That is not what this was ever about.

Now you know better, and that’s good.

So it shouldn’t be too much of a downer. You’re the best-qualified people on Earth to actually build this future.

Go out there. Let’s have some fun, right?

Seriously. Let’s explore. Have some fun.

Just get your homework done first.

Thank you.

[Applause.]

Q&A

RAPH KOSTER:

We probably have some seconds for questions. Five minutes for questions, says the glowing yellow sign.

There are microphones here if anybody has any. But you don’t, because I’ve persuaded you completely. That’s awesome.

There we go.

AUDIENCE MEMBER:

Hi. I thought it was really interesting what you said about how, essentially, you have responsibility, and in other fields you also have legal responsibility. If someone wants to open a taco stand, they have to deal with the health department and all those things. If you run a bed-and-breakfast, you have to do that.

So all of these things are not unreasonable for developers to take into account.

And I guess, dealing with human nature—at Turbine, we had to take out the [unclear] emote because it was a Nazi thing. You don’t always expect that at the beginning, but it’s bad stuff.

RAPH KOSTER:

Yeah. So if the questions all turn out to be MMO veterans coming up and saying, “Yeah, that’s fine”—

AUDIENCE MEMBER:

Thanks.

ANOTHER AUDIENCE MEMBER:

By the way, the inventor of TTP—

RAPH KOSTER:

Actually, it was Jeff Freeman who came up with that.

AUDIENCE MEMBER:

There’s one thing I wanted to say just to pile on top of this.

Over the course of WoW, we saw something really remarkable—something that makes me really sad that people aren’t making MMOs anymore. MMOs started as the Wild West of gaming and ended up being the single safest genre for women to play in gaming, because everything else is unmoderated voice chat, whereas MMOs have persistent identity, persistent community, logging—all the stuff you talked about.

That’s something we fought for, and Blizzard led the way and made it real. You can actually see the gender split of the audience of WoW go from 20/80 to 50/50 over the course of a decade.

I think it’s important for us to realize that took a lot of work.

I don’t have a question. I just want to be sure that we all acknowledge that.

RAPH KOSTER:

Yeah, I think we should.

It’s even more depressing when you realize that the starting point of 20/80 was already way the hell ahead, because at the time virtual worlds were starting, the gender split for typical mainstream gaming was more like 95/5.

So virtual worlds were already dramatically more inclusive spaces than gaming was in general.

Now we’ve made a lot of progress, and along comes Xbox Live. Oh my God.

There are definitely a lot of lessons to learn there—a lot of lessons to take on.

Let’s go over to the side.

AUDIENCE MEMBER:

I think you said that for the ideal virtual space, the avatars would be non-realistic and not moving. Can you elaborate on the “not moving”? Because not moving in VR, especially with hand controllers, seems kind of bad.

RAPH KOSTER:

Not moving very much. I actually mean locomotion—physical locomotion.

VR is really good at clumsily moving arms. It’s really good at that. It’s pretty bad at moving your legs, as we know. Most of the locomotion methods involve short-hop teleporting and whatnot because of the nausea issues, vestibular issues, blah blah.

But the other side of that issue is that allowing one avatar to move onto another is one of the starting points for these problems.

If I were making a social VR thing that I wanted to be a commercial success, I would go clone IMVU.

You sit at opposite sides of a table. You’re seated. You can wave your hands and your head all you want. You can build crazy-ass avatars, because it’s all about personal expression and getting to know the other person.

It’s limited to two people at a time, so if you bail or block, you never have to deal with that person again. It just sidesteps all the issues.

Whereas if you say, “No, what I really want to build is WoW, only in a VR headset,” I go, “Then you’re going to have to build all of the infrastructure that Damian just talked about.”

And you can’t. Frankly, I’m not sure anybody can.

I literally do not know a way for people to solve some of those issues with voice chat at current levels of technology.

Will we get there? Sure. But not today.

So that’s really what I meant.

Thanks.

AUDIENCE MEMBER:

I am old enough to have played MUDs and traditional MMOs for decades, and people tend not to take it very well when they get banned, even in text MUDs.

What happens when the form we are projecting into a virtual world is our form? Are people going to get more addicted to that experience, such that we could expect severe reactions from people we ban from those worlds?

RAPH KOSTER:

No, because most people who want to project their real forms aren’t going to bother to use virtual worlds to do it. Virtual worlds convey real forms really, really badly.

AUDIENCE MEMBER:

That’s not what I mean.

RAPH KOSTER:

I know what you mean. I think I do.

You’re basically asking: will people take banning and punishment more seriously when what they projected is their real self?

AUDIENCE MEMBER:

Right. They put more of themselves physically into it. With MMOs, it’s a lot of text. Maybe there’s chat. But in virtual reality, you are there. Your avatar might not look like you, but to you, you are in that space.

So what happens when somebody kicks you out of that space?

RAPH KOSTER:

Yeah. I think there’s probably a physical reaction.

But in practice, I literally do not buy that you are more there in the VR case than you are in the others.

The building up of identity is the thing that actually stings. The loss of that identity is what ends up being really painful.

They’re signaling me that we have to go to the overflow questions room, so, really fast—

AUDIENCE MEMBER:

To add to your list: Life on the Screen.

RAPH KOSTER:

Yes. Sherry Turkle.

If I started listing books and academics, we’d be here all day.

I’ll head to the overflow room—or just follow me. We’ll find a spot out there.