May 17, 2006
Social dynamics in MMOGs
It's a common belief that Massively Multiplayer Online Games (MMOGs) are more interesting than their single-player counterparts because of the ability to socialize in the game. A paper presented at this year's ACM Computer Human Interaction conference, "Alone Together?" Exploring the Social Dynamics of Massively Multiplayer Online Games, offers a different spin on that. After installing /who-bots on several World of Warcraft servers and watching people's play habits, researchers from PARC and Stanford University concluded:
"Our observations show that, while MMOGs are clearly social
environments, the extent and nature of the players' social activities
differ significantly from previous accounts. In particular, joint
activities are not very prevalent, especially in the early stages of
the game. WoW's subscribers, instead of playing with other people,
rely on them as an audience for their in-game performances, as an
entertaining spectacle, and as a diffuse and easily accessible source
of information and chitchat. For most, playing the game is therefore
like being "alone together"— surrounded by others, but not
necessarily actively interacting with them."
Some other interesting tidbits from the paper:
Players who never grouped tended to level up about twice as fast as those players who grouped more than 1% of the time. (The paper doesn't mention this possibility, but this makes me wonder whether these anti-social players are actually farmers working in a virtual sweatshop.)
Median guild size was only 9 (6 if you include "one-person guilds"), and the 90th percentile of the distribution is only 35 active members.
Guilds tend to be sparsely-knit social networks, with a guild member tending to ever see only one in four other guild members and only playing in the same zone as one in ten. (Again the paper doesn't say, but I imagine this statistic is influenced by people playing multiple characters in the same guild, which already forces some exclusion since people can't play more than one character at a time.)
Guilds tend to have one or two groups of tight-knit "core" players who play together regularly and are all of roughly the same level. This is probably a result of the level treadmill and the fact that people of radically different levels can't really adventure together — which means people who get out of synch with other guildmates can't adventure with their friends anymore and are more likely to quit the game or find a different guild.
(Thanks to Amy Bruckman for the pointer!)
Posted by bug to Culture at May 17, 2006 9:57 PM
I'm glad CHI is _finally_ publishing real papers on games, but I thought this one was fatally flawed. In particular, I think the authors don't give enough weight to the mechanics of the specific game. Drawing social conclusions about MMO gamers in general is not warranted on the basis of the data they have.
For example, just to address two of their points: WoW encourages soloing by biasing the rewards towards soloing. It would be completely possible for them to scale the rewards with group size, but in fact they don't. XP, gold, and quest drops are all sub-linear with group size. Compare that with, say, City of Heroes, where rewards scale the other way. You get _more_ rewards for fighting in group there.
Likewise, CoH has a "sidekick" mechanism that permits a low-level character to accompany a higher level and be an effective participant. WoW lacks such a mechanism.
In effect, the paper says "balls tend to roll this way" without actually pointing out that the particular hill they studied is sloped in that direction.
As a fellow student of this kind of thing and a WoW player:
I agree with the findings, and with most of your interpretations of them. Some quibbles, though:
Median guild size was only 9 (6 if you include "one-person guilds"), and the 90th percentile of the distribution is only 35 active members.
Yes, but the largest guilds are hundreds of players, and most people who are 'serious' about guild play belong to one of them. The median is less important in this case because most people looking for a guild look for one of the "big three" or "big five" or whatever it is on their local server.
Guilds tend to have one or two groups of tight-knit "core" players who play together regularly and are all of roughly the same level.
Yes. This is also why you don't regularly see most of your guildmates -- it's very hard to play with somebody who's more than five or six levels lower or higher than you are. With guildmates you'll occasionally see them essentially as a favor - they need somebody to run interference in a hard area so they can get some difficult goal. But that's basically you doing them a favor - in an area that you can defend them properly in, you'll get almost no money, treasure or experience to recompense the time spent. You do it as a way to bolster the strength of your guild (often by getting desirable items for the battlegrounds, to improve your group's player-versus-player ratings).
But it's very hard to do that with friends since some of you are essentially in the position of getting constant favors from the others, which gets to be a strain quickly.
(The paper doesn't mention this possibility, but this makes me wonder whether these anti-social players are actually farmers working in a virtual sweatshop.)
Some of them are. But speaking as one of them, not all of then are. In general, players will optimize extremely efficiently for the fastest gain of money, items and experience. Running a MUD is a surprising little exercise in that, it turns out -- players will often find much better scenarios than you thought could exist, because they spend all their time looking for tweaks and optimizations. Since you gain more experience when soloing, even non-farmers will solo most of the time. The exception, in my experience, is 'support' characters (mid- to high-level mages and priests primarily) who will generally join a guild and do fairly profitable group activities like instance runs because for *them* that's the most-optimized path. Other characters will cooperate because there are a few items and privileges that essentially require grouping to get, even if the moment-to-moment money and experience doesn't outweigh the risks. Again, there's an interaction with the battlefields on
this one because being very powerful compared to your level is a huge advantage there, so people will do riskier and less profitable activities to get that advantage.
I read the paper when they first started talking about it, several months ago, and brought a number of issues about it up with the author. I would like to second the 'fatally flawed' assessment of the survey.
First and foremost, they were able only to analyze the public chat spaces. That is to say, they did not have visibility into any conversational behavior during any multi-player activity. From a statistical perspective, that's a miniscule fraction of the social interaction in the place. Even stranger-to-stranger interaction in the game is predominantly private, directed between individual characters.
Following up on that issue, they have a very limited data set - they examined one instance of five different server categories, out of more than one hundred servers; players report that there is significant server-to-server variation, and I'm both suspicious from both an anecdotal and statistical perspective that they're drawing conclusions from an effectively very limited data set.
They also have no way to observe alt behavior. Most players now maintain a number of different characters on each server that they play on - generally one or two primary characters that they are playing, and then one or more characters that are placed in specific game locations (notably near the cities with auction houses) to perform trading activities. These special purpose alts are only leveled enough to allow them to actually reach the city, and then are left at that level indefinitely.
Related to alt behavior, they monitored how fast characters level, but those metrics become unusable at level 60 (because of WoW's level cap). This is important because it perturbs the way that players behave in guilds, and the fundamental item and experience economics of being in a guild. In particular, the fastest way for low-level characters to level is for a high level friend (or guild-mate) of theirs to run around with a group of low level characters running them through level-appropriate quests. Of course, this (paradoxically) can lead to a circumstance where a party of players are playing together, but *are not grouped* (in a game mechanic sense), because optimal xp is for the low level players not to be grouped. That behavior, called power leveling, is reported to be common, but can't be observed by their bots, which will (incorrectly) report that each one of those players is playing alone.
Similarly, one of the properties that guilds bring to the table is that they can organize higher level characters power leveling lower level characters (including cases where players will trade off between their own characters, in one case playing the high level character to farm gold and items, or chaperone lower level characters, and playing lower level characters and being the recipients of items and chaperoning).
Given the author's inability to observe and analyze group and guild communication, private player-to-player communication, and grouping behavior beyond the most limited set of information, I recommend taking many of their conclusions with a large tablespoon of salt.
Oh, some quick numerical responses to their specific observations:
- Players who never grouped tended to level up about twice as fast as those players who grouped more than 1% of the time. (The paper doesn't mention this possibility, but this makes me wonder whether these anti-social players are actually farmers working in a virtual sweatshop.)
Again, their analysis methodology hides power leveling behavior from their analysis tools. Characters are able to level fastest when they are being chaperoned by a high-level character but not grouped with them. Their analysis tool had no way to detect that behavior, which is common.
- Median guild size was only 9 (6 if you include "one-person guilds"), and the 90th percentile of the distribution is only 35 active members.
Yes... but the guilds above that 90th percentile have hundreds of members. So what's happening is that there are large numbers of vanity guilds, formed by a single individual who is essentially playing just with his existing tabletop gaming group, and small number of guilds that have very large numbers of members. And, again, because the author's tools have no way of determining the mapping between individual players and characters, they can't see behaviors like creating characters whose sole purpose is to play with a specific group of friends (again, this is a common behavior - people routinely create multiple characters, and those characters will be created on specific servers so as to play with their friends who already have characters on that server).
- Guilds tend to be sparsely-knit social networks, with a guild member tending to ever see only one in four other guild members and only playing in the same zone as one in ten. (Again the paper doesn't say, but I imagine this statistic is influenced by people playing multiple characters in the same guild, which already forces some exclusion since people can't play more than one character at a time.)
Yes, you are spot on - also, keep in mind the distorting impact of median guild size. The largest guilds have hundreds of members, at which point it is not entirely meaningful that you never group with more than ten percent of the other guild members.
- Guilds tend to have one or two groups of tight-knit "core" players who play together regularly and are all of roughly the same level. This is probably a result of the level treadmill and the fact that people of radically different levels can't really adventure together — which means people who get out of synch with other guildmates can't adventure with their friends anymore and are more likely to quit the game or find a different guild.
I would be very interested to see additional detail about this. It makes sense for smaller guilds, but large guilds (tens of players and above) have enough characters in play that there are significant benefits to lower level players to be in the guild (as recipients of items gained or made by other players, and even just hanging out). Also, the authors seemed to fail to comprehend that for most long-term players, all of the play up to level sixty is 'low-level preparatory play.' (This is also noteworthy, because your subsequent characters tend to level dramatically faster than your first one or two).
Oh, one other thing that just rankled when I read the paper:
Reaching the 40th level opens the possibility of riding a mount -- a way to travel across the world 60% faster, and a significant social status marker of being a "high level" character.
Um, yeah. Kind of like having a car conveys high status in the real world. Sure, if you're a fifteen year old.
I agree, the flaws you've all mentioned so far are pretty significant. Most especially, their numbers don't really support their conclusion that the social aspect of the game comes in large part from the having an audience, a sense of social presence and a sense of spectacle. I think there's probably some truth in that statement, but it in no way falls out of their results.
At various points of reading the paper I also had the feeling of "fatally flawed", however upon consideration I still recommend it. My reasons for the feel of "fatally flawed" is the paper doesn't acurately describe World of Warcraft as it exists today. The paper describes WoW-2004/5 and either through Blizzards design or inevitability WoW-2006 is an entirely different entity.
A WoW-2006 server, at least in the US is primarily composed of persons who are level 60 or playing an alt. Although I can only provide anecdotal evidance from the server I play on, I believe that 99% of persons have over 15.57 days played (time cited as average to level a 60) and easily over 90% of players have a 60.
I suspect this is inevitable over time on a server. Servers have a # of player cap (approx 20K) and Blizard tries to discourage new players from joining servers that are nearing or have reached their cap, instead rerouting them to new servers (which will inevitably reach this state.)
This has a dramatic effect on the social structure and game play structure of WoW. First, there is the concept that "WoW begins at 60". Many persons look at Levels 1-59 as training levels. Another is the seperation factor the authors discuss based on different levels disappears once everyone is 60. The other differences can be seen by how the game is played at 60.
At 60 WoW gameplay breaks down into four major activities:
1) Grinding: This is done solo or with whoever is available, example kill 12,000 Furolgs.
2) 5-10 man instances: There are 5 5 man and 1 10 man instances. Players run these repeatedly some people with pick-ups others play consistantly with a group of strongly bonded companions
3) 20-40 man raids: The game after 60 is focused on getting all players to raid. Raids have by far the best loot and are also of considerably higher difficulty. These are done with your guild almost exclusively
4) PvP Battlegrounds. These are actually raids due to their size (10,15,40) but are mostly with pickups.
So how does this different game affect what the paper had to say? First, I'd say that the numbers given don't really provide any evidence or indication of how WoW is played today. A prime example is using their numbers I should only know 1 in 4 of my guildmates and play with 1 in 10. This is not true or even close for me, anyone in my guild or any 60 in any raiding guild. I play with 95% of my guild (85 persons) and know a similar number (know to the level that as a guild officer I can a do rate guild members performance and provide specific comments.)
Another issue that I have to bring up is the authors request for tools to manage/report/support guilds. These tools exist and I'm not sure how the authors missed guildportal.com a hosting service that provides all the features the authors were requesting. Every raiding guild I'm familiar with uses guildportal. The same goes for Teamspeak/ Ventrillo a crucial element when discussing sociability in MMORPGs raiding play.
So considering I feel they're talking about a phase of a game that no longer exists, missed the current gameplay and overlooked the major tools used in the environment, why would I recommend the paper? Several reasons. First I found the infomation insightful about the training version of WoW. Secondly, I believe their concept of "Alone Together" has merits, both the concept of presenting to an audience and the fact the feel of WoW is that of the great good place. Finally, WoW has managed to do things with their Skinners box that deserves attention and that this box exists in WoW-2006 as well as WoW-2004/5.