« State of Play III | Main | SoP3 Catch-up »

Oct 08, 2005



We've had diseases in virtual worlds before, of course, but not in virtual worlds wih 4 million players.

In MUD2, players stand a chance of catching a cold if they sleep outside in the open and it starts to rain. When they have a cold, there's no cure (no magic is that powerful) and they'll occasionally sneeze and cough, reducing their stamina (hit points). If you sneeze or cough in the presence of another humanoid, they could catch your cold too (unless they only recently had it, in which case they're immune).

Most colds were spread between players, but occasionally a humanoid mobile would get hurt in a fight, sleep (to recover stamina) in the open, and catch a cold. This mobile would then cough on players and perhaps infect them.

Naturally, players took (well, take, as they still do this) this cold and use it as a weapon. The usual tactic is to catch a cold, visit a realm inhabited by some 52 (NPC) dwarfs, and infect one of those with it. They'll pass it to other dwarfs, until eventually most of the dwarf population has a cold. Dwarfs have high strength and dexterity, so they're tough in fights, but their stamina isn't so high; losing 3 stamina to a cough is therefore more worrisome for a dwarf than for a player character. Players will let the infection spread, then weigh in and attack the dwarfs that have reduced stamina or, better still, that are sleeping (no defence against an attack!) to recover the stamina they just coughed away.

One more from Irrelevant Anecdotes R Us...



Wow, that was beautiful, Richard. I could see the imagery in my head as I read the text.

Some have been kickin' the idea around in SL of a zombie invasion that escapes an island estate only to break out on the mainland. We'll need av decapitations first, tho.


Torley Wong>I could see the imagery in my head as I read the text.

There was imagery?

After seeing so many SLers partying in New York when I dropped in on their get-together half an hour ago, I think I can say you're going to get plenty of zombies tomorrow morning anyway...



I would guess that it would be pretty easy to make a VW that could test this kind of mentality. You'd have to do a few things though:

Give the test subjects good reasons to avoid harm. For example, if you hire the subjects and tell them that injury means temporarily lowered pay and death means their job is gone.

Don't encourage tempting death in the first place. It would be a bit silly to test a community for their reaction to dangerous situations if that community was made of people that actively hunted dragons, for example.

Remove the kind of surety that game VW's offer. If I see a level 20 avatar survive the threat, and I am level 50, then I can be fairly certain that I will be fine.

I don't think a game VW would be able to offer a good field of study... or at least, not games that center around dealing and defying death in the first place.

P.S. Since we aren't touching the permadeath issue here, I won't mention that the best analogy for it is that of people who insure their property and people who don't.


Oh, one more thing you'd need to do in this test VW. Either allow the subjects time to recuperate from the disasters, or run the test through cycles, hiring new subjects each time. The rationale here is that you would desensitize the subjects pretty quickly once they realized that the whole point was to survive "catastrophies".


Yes, one more comment...

Make sure that social connections were of high importance, or you'll never get the self-risking and self-sacrificing actions or the actions would have different meanings. Subjects would be more likely to risk whatever carrot you've dangled for them if they're helping a friend/aquaintace save their carrot. The lesser severity of the consequences on the physical world would probably enourage people to watch out for their own skins. After all, who cares if some stranger loses their carrot?

I don't know that this problem could ever really be overcome. People might risk something to preserve a relationship, but then they might just expect to get in touch in the physical world (or a different VW) as well.



Long-time reader, first time poster. :)

Very intersting comments. I love Richard's anecdote from MUD2 on player strategy and the common cold, and Jim's comments provide some solid starting points from a game design perspective.

I humbly submit, however, that to truly use a virtual world as a viable test mechanism for transmission, you have to hide the test from the players. When we were designing Glympse (which sadly died on vine), we never trusted the players. "How can this be power-gamed?" was the mantra. We would vet design with power gameers who would come in and take great joy in breaking things. That was obviously tremendously helpful for dev, but if there's no need to go through the process in the first place ...

Don't treat a study on transmission as a game design challenge. Create a small group of insiders on the live dev team. Randomly seed the disease in accordance to the test criteria you've set up. Let the entire thing unfold in a double-blind roll out, and only allow data analysis after the predetermined time limit. That would be fascinating.

This would be the way to go to simulate thransmission. You want the players to go about their business. If, however, you then wanted to simulate deeper reactions, then it would indeed become a game design issue.


Dave C.> Let the entire thing unfold in a double-blind roll out, and only allow data analysis after the predetermined time limit. That would be fascinating.


Yes, this is an approach you might take if your only interest was science unconstrained by any ethics... but doesn't this smack of "research on unconsenting human subjects?" Is it acceptable to experiment on characters without the knowledge or consent of the players who run those characters?

Of course Bad Things can happen to any character in a game as the impartial computer rolls the pseudorandom dice. But we're talking about going beyond that -- this is about letting the live team deliberately and secretly cause Bad Things to happen to particular characters.

"But isn't that exactly what a live DM in a D&D seesion would do?" Not quite. There's a difference between making a Bad Thing happen to your character as an expected in-game challenge for your entertainment, and making a Bad Thing happen to your character for some out-of-game reason such as research.

When your thief fails her saving throw and gets injured by the trap pre-planted by the DM in the chest she just opened, you accept it because you understand that it's part of the game. When your thief gets injured by a gnoll because her vitality was sapped by a fever that the DM only gave her to see whether she'd infect other players, you're likely to object because you had no reason to expect that being a real-world test subject would be part of the game.

Whether playing as characters in a D&D session or as avatars in a computer-based virtual world, the actual test subjects are real human beings whose personal psychologies are being explored. This suggests that adherence to a human research code of ethics is appropriate.

The American Psychological Association has published an ethical code of this kind. Sections 3.10 (Informed Consent) and 8.02 (Informed Consent to Research) are applicable, but in particular, Section 8.07 (Deception in Research) is relevant to a covert "viral" transmission study.

Please don't get me wrong: I'm not opposed to research, or to research on humans, or even to covert research of the kind proposed in this thread. But I don't think human science happens in a vacuum, either -- if it's to be done to real people, then there are ethical concerns that should be addressed along with the scientific methodologies.

In a MMOG, I'm not sure how you satisfy those concerns without giving some thought to them as part of the game's design. The ethical stance of a game's operators toward their players shouldn't be an afterthought.

All of which said, I still wonder: What if the "virus" didn't injure a player's character in any way? What if the payload was merely an invisible tag that researchers could track as it passed from character to character? Would that still be "deceptive" research? When a player agrees to a EULA that says something like, "We own your character and can do anything we like with it," does that constitute informed consent to being treated as a psychological test subject?

I definitely think developers need to consider the ethics of experimenting on player characters. But I don't pretend to have clear answers to such questions.



Hi Bart,

I should have been clearer: I was suggesting that this not be introduced as a dynamic of game play at all. I.e., that as you suggested, the "virus" was just a tag that could be parsed to generate data sets.

I submit that if you inform the players that you are running the test, you lose your ability to have a control set as you can't truly know whether or not the announcement has or hasn't affected their behavior. If the idea is to use virtual populations to simulate these types of phenomena in meatspace populations, then you need to recreate those conditions. I'm not a viroligist or epidimologist, but I believe that at some point in the rise of a full-blown epidemic there is necessarily a time where the situation is not understood. If this is true, then you would have to hide at least the beginning of a virtual epidemic from the players.

Hmmm ... I guess by my own argument, true authenticity would involve notifying the players and providing them with the game play mechanism to protect themselves, but at least you'd be able to simulate fatal transmission resulting from zero attempt to halt the epidemic. ;-)

You further point is still a great one, that is, would it still be "deceptive research" to conduct this in a "harmless" way in-game? This could indeed be covered by the EULA (e.g. "Customer is paying for service, publisher reserves right to anonymously track usage data of said service for R&D purposed, blah blah blah").

Does that make it "right"? I don't know, but I do believe it's a philosophical issue at that point: you have a blanket statement that absolves you of secrecy and you're not harming their paid-for experience in any way.


Give the test subjects good reasons to avoid harm.

I think this is the key if you want to study any realish approximation of the dynamics at play. Fear is a huge behavioral factor. Imagine if WoW agents were invoked to corral and quarantine anyone known to be a carrier. How would this alter the virtual society?


Re: Bart's post on ethics.

You probably can't do it in a current virtual world, but there's nothing stopping future developers from adding a clause saying such things may happen and should thus be expected. Players will still go ballistic (See Lies, Heist, and Social Emergence), but it will be a part of the game, and wouldn't it thus be ethical to do it?


Incidentally, a virus (Why Pox) was deliberately unleashed in the kids' 3-D world Whyville a couple of years back in order to teach kids about infectious disease. Yasmin Kafai presented their results at DiGRA this year.



I would agree with you that such research would raise an ethical issue. I would not agree with Dave on that particular point simply because it is the psychology that is the heart of the ethics and not the server-side information. Sure, you have the right to give PCs a disease, but do you have the right to cause panic in the meatbag behind it? I would say that would be a violation of the spirit of research ethics if not the letter.

I still don't see the problem in having a MMOG with research. Sure, inform the subjects, but you don't have to tell them *what* research will happen or when it is happening. If you are doing studies similar to what Nick is doing with WoW at the same time, all the better. Letting the players know that you will be studying "lots of things" would be vague enough to keep your secrecy and still meet the ethical requirements. I would expect some players to try to break the experiment still, but within a reasonable margin. Especially if the game itself is played up to distract them.


Talk about a lot of attention given to games lately in the real world, I heard about this and the very next audio in the news on NPR was about the Nobel economics prize.

Gamers are given more credit than they should, irrational behavior is common in gamers whose actions have no personal consequences.


I feel the same way most commenters do. The needs of ethical disclosure can probably be met with a simple statement such as, "You acknowledge that we may, at our discretion, collect data on your in-game actions for research purposes." As Jim pointed out, such a statement needs to be broad enough to adhere to the spirit of the ethical code, but shouldn't be so specific that player behavior is affected.

(Note: It would be nice if a notice included a privacy statement -- something like, "This data will be aggregated so that the privacy of individual players is not compromised" but that's not necessary to meet the burden of notification. It also might not be desirable if you want to associate character actions with aspects of the player, such as age, sex, ethnicity, etc. for those kinds of analyses, or if you ever want to do in-depth follow-up assessments with the player.)

What I'm wondering now is whether any notification that a player's in-game actions might be scrutinized for research data could lead to behavior that distorts the results.

At a gross level, how many people simply won't ever play a game that might "spy" on them? Sure, it's irrational; CSRs in many games can not only watch players, they can do much, much worse things... but there's just something about the idea of being studied against one's will by some invisible lab-coat type that some people simply won't accept. That would alter your sample population.

Then there are those who'll play but who'll constantly be looking over their shoulder, wondering who might be watching them. To some extent, these players could self-censor, skewing behavioral results.

Finally, what if players somehow learned that information about their gameplay was being tracked in a viral way -- that is, by tags passed among players through some kind of contact? Whether out of concern for their privacy or a perverse need to grief the researchers, it's easy to imagine some players choosing to limit their contacts with other players. Not only would that affect study results, isn't it (in a practical sense) contrary to the idea of a massively multiplayer world for players to avoid each other out of fear of being "infected?"

Viral propagation is interesting, and more research on what players really want would be helpful. A multiplayer world seems like a ready-made place to do both. But I wonder if caution is advised on both counts, and for the same reason: how will knowledge of them change player behavior?

Galrahn> irrational behavior is common in gamers whose actions have no personal consequences.

I'll agree, if you'll allow me to amend it to: "Irrational behavior is common in any persons whose actions have no personal consequences."



First to address Bart, and the ethics issue. It has already been decided, and anyone who plays WoW has already agreed to it. An excerpt from the WoW Terms of Use:

Blizzard Entertainment may change, modify, suspend, or discontinue any aspect of World of Warcraft at any time. Blizzard Entertainment may also impose limits on certain features or restrict your access to parts or all of World of Warcraft without notice or liability. You have no interest, monetary or otherwise, in any feature, content or availability of World of Warcraft, any Game Data (defined in Section 13(J), below) or in any terms or conditions of access to or use of World of Warcraft.


Now that that's been said, we can discuss the effectiveness of virtual experimentation, as I believe was this post's original intent -- whether studies can be directly performed in a virtual world.

Ultimately, I'd say no, they cannot effectively be performed in a virtual world, because there is nothing at all in virtual worlds that even comes close to simulating the fear of death. I know, I've been there.

But there is something I believe that might work nearly as well. Fear of someone else's death. Or, at least in this case, virtual death. In WoW, as well as in all other RPGs/MMOGs/etc., a fundamental aspect that keeps players playing is friendship and community. Undoubtedly within a very short period of time after having begun a new game, we meet other people in that game. That's the way they're designed. Then we chat with them, chat with others, get to know others, learn their names, add them to our friends list, join in parties with them, join their guilds, perhaps even exchange email addresses, and so on. These people we never really see become our virtual friends. Our pals. Honestly, many of us spend more time with them than we do with any other Real Life person.

Let's just take that as a given. I'm sure there are arguments that there are "lone wolves" out there, but I think for the most part, we will agree that the majority of players fit the more social description.

A potentially realistic reaction could be obtained from players if a sudden contagion started spreading throughout the land, and... the contagion is based upon a correlation: essentially, those who spent the most in-game time chatting with each other would be selected and divided, approximately half of them catching the virus; those who caught the virus would progressively become "unlucky" in the game -- they'd roll bad on everything, their stats would begin to lower, and their gear would not sustain them. The other half would be fine. Nothing would happen to them. Then the worst part begins. The infected start to "die." Once they are "dead," when they relog onto "their" server with the "dead" characters, they'll be just fine... but on a mirror server where their old friends cannot come. Their old, uninfected friends will still maintain healthy characters on the old server. The devs would examine the chat logs (yes, within their rights), as the experiment starts, and progresses, to see how people react.

Of course, then everyone involved would cancel their accounts.

But, in the process, some answers might be gleaned from the uninfected: what were you feeling when your friends got infected? did you feel sorry for them? were you glad it wasn't happening to you? which feeling was greater? how did you feel when you couldn't communicate with your friends? when you couldn't even see them in game any longer?

An interesting experiment that's sure to evoke some real feelings, and probably some feelings and reactions that would mirror those caught in real-life similar situations.

However, I still think it's bunk.


Blame. It's easy as heck to point a finger at Blizzard. Hey, quit that! Fix it!

In other words, we "know" there's a solution to this problem. With epidemics... we don't know. Sure, we can pass blame. The government didn't find a cure fast enough, or didn't distribute it effectively, or inform the public in a timely manner. Yes, it's easy to place that kind of blame. However, it's different because (unless you're a conspiracy theorist) the government can't just press a button and make it all go away, and we "know" that. We might not want to admit it, but that feeling of helplessness and impending doom is the essense of epidemics... i.e., what would our reactions be if we were in the path of an unstoppable train?

In the real world, we start praying. Even non-believers start praying (just in case). We beg, we plead, we cry, we wonder why, and most of all, we sit there, helpless. Helplessness is the crux of the problem, and I don't believe (at least in this point of virtual technology) it can be recreated in a game.

Because in a game, we always have Alt-F4.

P.S. Yes, there's always RL suicide that could be argued to mirror Alt-F4, but the end result is obviously vastly different. Why'd you even think that!




I would agree that your scenario would skew the research. On the other hand, IMHO players would lose themselves enough in the game to not care what data was being gathered, especially if it was guaranteed to be non-intrusive and to not include chat logs.


I would also agree with you that true fear of death can't be provoked in players of a MMOG, but the behavior could. Also, my personal experiences in gaming have led me to believe that players pass through the same kinds of mental stages as in RL when confronted with an unexpected disaster: shock, comprehension, action/inaction, etc.

Your argument that the emotions wouldn't reach the same level is totally valid (speaking generally), but I wonder if gathering the data on the behavior wouldn't be enough? It's hard to quantify emotions anyway, RL or otherwise; behavior fits much more nicely into columns. *grin*


I also want to suggest, for anyone who's considering actually doing this, that you use a third party journalist to do your interviews, with the promise that the journalist will give you copies. =P

Oh, and you're required to inform us about the results.

The comments to this entry are closed.