Understanding Video Games text-book
Trust, Cooperation, and Reputation in Massively Multiplayer Online Games

Date posted: November 16, 2006

By Tony Tulathimutte

Given the genre’s staggering growth and diversification over the last decade, the trust issues surrounding massively-multiplayer online games (MMOGs) are becoming as diverse and complex as those found in real-world systems. MMOGs like Second Life and The Sims Online created environments where real-life social phenomena are encouraged and replicated, while games such as World of Warcraft, Lineage, and Everquest, in virtue of their role-playing and fantasy settings, create new social dynamics with few practical real-life analogues, which in turn create new bases for trust.

User Demographics
As of June 2005, there are an estimated 9,250,000 active MMOG subscribers, with the games Lineage, Lineage II, and World of Warcraft comprising 67% of the market share (Woodcock). According to an online survey of 30,000 MMOG players, the mean age of users is about 26, and ages range from 11 to 68; weekly use averages 22 hours (Yee, “Demographics”). Though in most MMOG populations male players outnumber females by a wide margin, gender proportions are steadily converging, and in many respects (e.g. guild membership) females tend to be more dedicated to certain aspects of gameplay than males (Yee, “Norrathian”).

Massively Multiplayer Online Games – Background
In addition to an initial software purchase or download which costs around 50 dollars, MMOGs typically charge a monthly fee of 10-25 dollars, excluding one recent game (Guild Wars). Players are encouraged to meet, cooperate, and socialize in the game environment; users in my survey reported that they meet and play in a group with new players every time they play. Common tasks include informal adventuring for the sake of gathering items and completing predefined mission objectives, meeting to socialize and role-play, and creating and exhibiting player-created content such as items, furniture, character models, organized performances, and so on. Often times, tasks are designed such that they are too difficult to realistically complete with only a single player. Most MMOGs have a form of “guild” system which allows players to organize into a semi-hierarchical group with fellow players, and there is data to suggest that the majority of players belong to a guild (Yee, “Norrathian”).
The games are offered as entertainment, but many more serious uses and abuses of MMOG systems have since emerged: “farming” characters for retail (Loftus), real-world attacks prompted by in-game actions (Levander), and high-profile allegations of virtual underage prostitution (Ludlow). Since MMOGs are subscription-based services owned and maintained privately, players are subject to strict end-user license agreements and terms-of-use policies, as well as less formal game etiquette standards established both by the game companies and the player communities. However, the extent of repercussions for transgressive in-game behavior has thus far only amounted to account suspension or cancellation; there has yet to be a criminal investigation arising from actions between in-game characters. This may have to do with the regular patrolling of game environments by company-employed officials, or “GMs”, who have the ability to move undetected, observe remote exchanges, and eject any players from the game at will; moreover, most game actions and dialogue are recorded in server-side logs. The lack of privacy makes the use of MMOGs for illicit legal conduct risky; however, the otherwise lax repercussions make more minor behavioral infractions prevalent, such as verbal harassment and item stealing.

Trust Issues and Benefits in MMOGs
In almost every sense analogous to the offline world, trust serves numerous functions between MMOG players. Trading and bartering of equipment, items, and property occur much as they do in real life, and cooperative tasks such as exploring dungeons and defeating enemies form the bulk of gameplay in games such as World of Warcraft. As such, MMOGs share many trust issues with online transactions, such as those found in e-commerce and online auctions like eBay and craigslist, where participants are mutually anonymous and direct retribution for fraud is difficult. Similar to those sites, then, MMOGs have implemented reputation systems of their own; however, the entertainment-oriented environment of MMO worlds makes certain abuse and fraud issues all the more salient for their ease of execution (Appelcline). Corritore et al. cite risk as a defining factor of online trust (241), and since online play environments are typically designed to be risk-free, people are more willing to trust more quickly and on weaker grounds.
Naturally, players have found many ways to exploit reputation systems in MMOGs. Since certain actions will enhance one’s trustworthiness according to the conventions of the game, players can write “macro” programs to repeat these actions ad nauseam, or simply invest time in performing the actions themselves, artificially inflating one’s reputation score, and thereby their perceived trustworthiness. Moreover, since new characters and identities are easily created, it is easy to falsify positive reputation from many different sources, which is a common basis for judging overall trustworthiness online.

Finally, the anonymity and lowered stakes of the MMOG environment have spawned a category of players known as “griefers”, who take pleasure in the intrinsic appeal of annoying others, going to great lengths in-game to cause slight-to-major annoyance to other players; this is less common in the real world, where such people might incur severe consequences for this behavior. Griefers confound the motivations for evaluating trust and trusting reputation scores, because some griefers will build reputation for long periods of time simply to grief more effectively, and they are not motivated by self-interest where game standing or welfare is concerned.

MMOG groups share several similarities to temporary systems and virtual organizations in the real world: like temporary systems, groups can easily be described as “a set of diversely skilled people working together on a complex task over a limited period of time” (qtd. in Meyerson et al. 168). Players often interact in highly transient, lightweight situations, and many users report that they play with different players nearly every play session, and often only once. As in the real world, this pattern of play makes it difficult to form long relationships upon which one would otherwise base trust; rather, players must employ swift trust (167). Furthermore, player-created groups lack the kind of authoritative “institutional mechanisms” into which team members in real-world teams invest their trust (187); there is often no “leader”. An effective reputation system is therefore critical for providing a surrogate basis for trust and facilitating cooperation.

Reputation Systems in MMOGs – Background
Reputation systems of all sorts have been in widespread use in online games ever since the first mainstream MMOG, Ultima Online (UO), was released in 1997. Most often, reputation systems have been criticized for being “gameable”, or capable of being exploited, allowing a player to either artificially inflate his own reputation or defame another player’s. Raph Koster, one of the lead designers of UO, had this to say about his experiences with reputation systems:

…the game system attempted to detect good and bad actions, and adjusted a stat on the character based on their history of actions. It led to all the bad guys having sterling reputations and all the good guys with terrible reps because they were willing to sacrifice their good stats in order to take down the bad guys (who had great reps through abuse of the system). I suppose that in some ways this is an accurate simulation of real life.
After that failed we moved on to one where transactions were assessed by a human, rather than by the computer… Each murder you committed gave the victim the choice to report you, and to submit cash towards a bounty on your head… Numerous tricks had to be put in place in order to curtail people working off the murder counts over time (we believed that people needed to be able to reform, which led to people “macroing off murder counts” in their homes… (Koster)

In addition, players criticized the system because it was unclear to them what types of in-game behaviors would lead to gaining or losing notoriety; for example, looting corpses or slaying non-player characters (NPCs) would cause one to lose points, but looting other players and trespassing in people’s houses would not (Fitzpatrick). Interestingly, although players could give other players positive karma (by forfeiting 5 points of their own), the development team described the system’s intent as “to make this into a roleplay thing–it has no real gameplay consequences” (ibid.). Rather than a system intended to indicate trustworthiness to other players, it was only intended to govern interactions with NPCs.
Other notable instances of online reputation system implementations have been World of Warcraft’s “Honor system”, which rewards players who fought with other players of comparable experience levels with access to special titles and items; the idea is that players who fought fairly would be more trustworthy. However, as one user pointed out, one’s honor ranking typically has more to do with how much time is invested in fighting, rather than exactly how honorable a character is. The socially-oriented MMOG Second Life allows players to rate other players with positive or negative feedback, for a fee of game money. Though the fee has reportedly served as a deterrent to exploitation, it also means that rich players have greater leverage—which is even more problematic due to the fact that game currency can be bought offline with real money. Finally, The Sims Online’s “Relationship system” (shown at left) consists of a visualizable network of everybody the player has made a transaction with; friends are indicated by green links, enemies by red, and the length of the links indicates the depth of the relationship between two players, as measured by the number of positive or negative transactions shared between them. This system provides a quick means of assessing not only how reputable a character is, but who the source of the reputation is. Unfortunately, this aspect of the system is not as useful if the user does not know who those sources are, which is often the case. Furthermore, TSO’s system has been subject to one of the most well-publicized abuses, in which a group of players calling themselves the “Sim Mafia” accepted payments of game money to gang up on a player and perform a “hit”, bombarding the player with negative ratings. This was highly disruptive to the target of the hit, because TSO links a player’s access to game features with his reputation, ironically, in an attempt to encourage goodwill.

A Proposed Implementation of Reputation in MMOGs
I propose a general design for reputation systems in MMOGs which, although not ironclad, hopefully resolves many of the loopholes and vulnerabilities of previous attempts at encoding trust into a system operated by the population of players. In doing so, I have attempted to identify the bases of trust that apply specifically to MMOGs and apply theories of online trust accordingly; I will enumerate these after describing the proposed system.
In my system, which I will refer to as “RS-Tag”, ratings are based on a “tag” system similar to one proposed by a poster on the TerraNova game development blog (AFFA). A player (player A) can assign another player (player B) up to one negative or positive rating, which can be modified at any later date if the player changes his mind. Each rating is accompanied by a mandatory 30-character comment “tag” which describes the rationale behind the rating. Although the ratings are initially valued at either +1 or -1 reputation points, if another player C also gives player B the same rating, and either C is on A’s friends list or A is on C’s friends list, then RS-Tag count their two combined ratings as only one point. That is, each of their ratings are divided by the number of friends giving another player identical ratings. So, for example, if players A, B, C, and D were all friends, and they all rated player E positively, then each of their ratings would only be worth 1/4th of a point, so the “voting bloc” is restricted to a single point. The relationships between A, B, C, and D would be checked whenever the E’s reputability was assessed, so that the friends could not temporarily remove one another from their friends list when assigning the rating and then simply add each other later. Furthermore, the database of all tags would be publicly available, such that if you checked player A’s public profile, you could see what all other players have said about player A, with links to the other players’ own profiles and trustworthiness. Finally, multiple characters from the same account could only form one rating of another character, and all characters on a single account share the same rating.
RS-Tag focuses on improving two elements of MMOG reputation: removing the incentive to game the system, and preventing factions of players or high-level players from inflating their own ratings (as in UO) or driving down other people’s ratings (as in TSO). First and foremost, all ties between trust scores and game content have been severed; as soon as there is some tangible benefit conferred by a high trust score, there is a huge motivation to game the system. Unlike UO’s reputation systems, RS-Tag ensures that the reputation system’s only purpose is to assist players in forming judgments of trustworthiness. The text tags emphasize this by giving specific details about the character’s trustworthiness, and their mandatory provision simply adds another deterrent to making artificial ratings .
Furthermore, the “bloc voting restrictions” prevent groups of friends from performing hits on players. They also cause slower gain/loss of reputation than the one-player, one-vote system; as a result, the player’s reputation score is more trustworthy. In order to have a score of +5, a player would have had to cooperate with at least 5 distinct groups of players, which is considerably different than cooperating once with a group of 5 players. This slow growth of trust could be effectively used as a surrogate for “slow trust”, since it would take a player quite a lot of cooperation to achieve any significantly high score, and conversely, quite a lot of grief to many different people in order to earn a low score; thus, there is an adequate basis upon which to base swift trust.
There are fallibilities to RS-Tag, but hopefully the cost of exploiting these vulnerabilities would be too great to appeal even to dedicated griefers. First, the bloc voting restrictions could simply be circumvented if a group of friends all remove each other from one another’s friends lists; however, the benefit to this group would not be very significant (adding or subtracting a few reputation points to some player), but the loss of communication between them would be very inconvenient, as friends lists are becoming more vital for managing in-game communication, so hopefully this trade-off will deter this behavior. Another possible exploit might involve a player opening up several distinct accounts, but this would require acquiring many subscriptions with distinct credit cards, and few players would consider this practical. Also, players who give other players positive ratings in order to receive one in turn might later change their minds out of spite; this is easily remedied by a notifier which informs players of when other players have changed their ratings, so they can respond in turn. RS-Tag also prevents players who prefer to play solo or always with the same group of friends from earning a high reputation score; but then, such players would not have any use for trust systems at all. If anything, RS-Tag encourages players to meet and cooperate with as many separate groups of people as possible, which indeed is one of the underlying tenets of the MMO genre at large.
RS-Tag is a synthesis of previously postulated ideas, coordinated in order to provide players with a basis for trusting other players. It has nothing to do with role-play; that is, it is not meant to represent a character’s trustworthiness, but the trustworthiness of the player controlling the character . However, in the future, it might be interesting to study source-orientation effects to see if reputation assignments are influenced by the appearance or in-character behavior of an avatar, even when players are explicitly instructed to rate the person controlling the avatar. If the effects are significant, then this might be another potential failing of RS-Tag, which assumes players are able to distinguish between in-character and out-of character behavior. However, the alternative would be to put reputation in the hands of automated behavior monitoring algorithms, none of which have yet succeeded in resisting exploitation by any player with enough friends or time on his hands.

Works Cited

AFFA. “TerraNova: Reputation.” 21 December 2003. 6 August 2005.
http://terranova.blogs.com/terra_nova/2003/12/reputation.html

Appelcline, Shannon. “Future Memes, Part Four: Community and Reputation.” 24 January 2002. Skotos.net. 10 August 2005. http://www.skotos.net/articles/TTnT_58.shtml

Corritore, C.L., Kracher, B., Wiedenbeck, S. “On-line trust: evolving concepts, evolving themes, a model.” International Journal of Human-Computer Studies. 58:6. 2003.

Fitzpatrick, Rob. “Ultima Online: Social Accountability for Good and Evil.”
(Presented 2/22/05 to Georgia Tech Game Seminar in the EGL)

Koster, Raph. “TerraNova: Reputation.” 21 December 2003. 6 August 2005.
http://terranova.blogs.com/terra_nova/2003/12/reputation.html

Levander, Michelle. “Where does fantasy end?” Time Magazine. 157: 22. 4 June 2001. http://www.time.com/time/interactive/entertainment/gangs_np.html

Loftus, Tim. “Virtual worlds wind up in real world’s courts.” 7 February 2005. MSNBC.com. 3 August 2005. http://www.msnbc.msn.com/id/6870901/

Ludlow, Peter. “Evangeline: Interview with a Child cyber-Prostitute in TSO”. 8 December 2003. Second Life Herald. 7 August 2005. http://www.alphavilleherald.com/archives/000049.html

Meyerson, D., Weick, K.E., & Kramer, R. “Swift trust and temporary groups.” Ed. R. Kramer & T.R. Tyler, Trust in Organizations: Frontiers of Theory and Research. 1996.

TSOMania.net. “Game Guides :: Relationship System (Friendship Web).” 2004. 10 August 2005. http://www.tsomania.net/gameguides/relationship_system.php

Woodcock, Bruce Sterling. “An Analysis of MMOG Subscription Growth” MMOGCHART.COM 12.0. 29 November 2004. 1 January 2005. http://www.mmogchart.com

Yee, Nick. “The Demographics, Motivations and Derived Experiences of Users of Massively Multi-User Online Graphical Environments.” Diss. Stanford University. http://www.nickyee.com/daedalus/archives/pdf/Yee_MMORPG_Presence_Paper.pdf

–. “The Norrathian Scrolls: A Study of Everquest.” Diss. Haverford College. May 2001. http://www.nickyee.com/eqt/home.html


Appendix A: Sample Questionnaire

Below is one sample response to the questionnaire I sent to a dozen people. Salient comments are in bold.

Answer these questions about your MMO(s) of choice. Be as detailed or as concise as you please, but answer completely.

–Which MMO games do you play most often?

World of Warcraft

–How often and in what cases do you cooperate with people you’ve met in-game and have never met face-to-face?

All the time. Pick up groups require at least 5 people—I have only been in a group of 5 people I know from RL twice maybe. Meanwhile, any sort of end game raid (40 people) definitely requires cooperation with people I do not know face to face. Essentially every time I sign on there is some cooperation required with people I do not know from real life.

–How often and in what cases do you cooperate with people only once or for brief spans of time?

Mostly for 5 man instances, it is possible to cooperate with somebody only once. While there is no guarantee that the cooperation will only occur once, there is no assumption of further interaction in many cases. For the brief span of time one it is either when someone asks for help “briefly”, if we notice we are working towards the same simple quest, or if the group sucks and it falls apart.

–Do you tend to play with people you’ve played with before, or do you tend to play with people you’ve never played with before?

A little of both, and often both at the same time. The class structure in wow requires a variety of classes to be successful in an instance. If people I know of a certain class are on, I try to query them first, but if they are unavailable I just take anybody who responds to my LF “this class” messages in chat. Sometimes if I am particularly bored I’ll also join groups in a similar scenario—they are looking for somebody and I fit the bill. This is often the case as I play a healer class and they are in demand, so there are great swings in the familiarity I have with my group mates.

–How well do you feel you typically get to know people that you’ve met in-game?

This is a difficult question. I feel you can get to know a lot about their personality and their playstyle, but that unless you really go out of your way, you won’t find out much about their real life undertakings (work/age/etc). The exception being, if they have “off-hour” jobs, leading them to be at work 2-10 pm Saturday and Sunday, at which point it becomes common knowledge they are waiter or something. Once you get on to a voice chat server with your guild for more complex raids, the amount of familiarity with individuals increases. Also, while I have certainly put my time into the game, I have played a lot (/played 20 days), but not as obscenely much as others (/played of 40 or more…)

–Are you in a guild?

Yes.

–How/why did you join?

Pretty damn necessary in WoW. Very few 60’s are not in a guild and they are usually Chinese gold farmers, or people who were dissatisfied with their guild or whose guild was dissatisfied with them. I joined because I was grouped with an individual who seemed nice enough (and skilled enough) and they asked if I wanted to join.

–Do you regard your fellow guild members as trustworthy? Why?

For the most part, yes. Partly from having grouped with them over time—they pass the Turing test of trust, if you will. Also, because I know they have more to gain over time through cooperation than by defecting. This is not the prisoner’s dilemma—word gets around in the guild and by working together they can be more rewarded than by stabbing me in the back. Once again, especially since I am a healer and in short order. I leave the guild, they start having a lot of trouble .

–Do you prefer playing in a guild / with a team of friends, meeting people on your own, or playing alone? Why?

All of the above but the last. I think I enjoy playing with friends the most, but also enjoy the socialization and potential “human capital research” derived from playing with new people. You don’t get more skilled, cool friends by not meeting new people. The last option I don’t go for too much—the game is all right solo, but the complexities and challenges only emerge in group play. The AI is cheesy and boring—it’s working with people that is interesting. Also, being a healing class while people really need me, I also really need people. Killing things on my own is extraordinarily slow.

–Do you consider in-game reputation systems effective and/or reliable?

While the game does not have one in place, except for, arguably the PvP honor system, I am wary of in-game reputation systems. You have a bunch of maximizing nerds with a fair amount of time they dedicate to the game—the system would have to be rather robust to withstand attempted cheats or reputation would not have to be rewarded enough for cheating to be worthwhile. If reputations were publicized it would be impossible to control how much individual players reward positive reputation, thus making the goal of a robust, impossible to game system more important. I doubt whether I would believe people’s abilities based on their reputation score.

–If applicable, in what cases have you rated a person positively / negatively?

Not applicable. On the extremes are the only two personal options available—adding them to your buddy list, or deciding never to group with them again. I have added about 30 individuals to my buddy list, and have decided to never group with, I’d say, about 5. Most of those are from personality and not skill disagreements though…

–What does it take for you to trust other players, both in low and high risk situations? How long does that take?

Well, risk is actually never that high, but I guess the time loss can be huge. You know within the first 5 minutes of a group how skilled the players are (are they fulfilling their needed roles?), and if they are doing something wrong you will know after the first 10 minutes if they are willing to be open and work as a team. Or usually you do. Meanwhile they are many systems in place to make sure you don’t have to completely trust individuals either. Synchronous trading, master looting system, hierarchical guild powers all prevent people from having too much ability to abuse trust.

–Do you feel that you can easily distinguish between in-character (IC) and out-of character behaviors?

There is basically no role-playing that goes on on the server I play on. Everybody knows they are playing a game with people on the opposite keybord—I would feel comfortable with saying everything is OOC. Your character says what you want to say—thus night elf, humans, dwarfs, and gnomes (from the alliance) will all talk the same about how 1337 their crits are. Haven’t played on a role-playing server though, so not sure what it is like there.

–A priori, do you consider other players generally trustworthy or generally untrustworthy?

Generally trustworthy. Doesn’t mean I’ll trust them though.

–Which modes of communication do you typically use to communicate with other players (i.e. text-only, text + avatar, text + avatar + voice chat, etc.)?

Text for people I don’t know. Avatar chat is usually used for trying to do silly things while waiting for something. (Or occasionally doing some taunting at your cross-faction opponents—who can’t read your text chat). Voice chat for in guild (and some of my RL friends) cooperative efforts. Once voice chat enters the text screen becomes rather muted.

–What are the most desirable traits in an MMO partner/companion? The most undesirable?

Patient, nice, competent, fair. Hurried, distracted, belligerent (they always think they can do no wrong and anybody making a slight mistake is just the worst thing in the universe—often these people spend way too much in game/have too much invested in achievement in game), unwilling to be flexible with their play to benefit group. And no fucking ninja looters.

–What do you consider a successful in-game partnership? A failed one?

A successful one is any partnership that keeps me entertained and not frustrated. I have done many failed instances with funny people. Ultimately though the surest measure of PvE success is the number of “wipes” (or everyone in the party dies) while seeking out whatever goal was had in the instance. No wipes is good, one is pretty much expected, 2 is reasonable. Above two and I start questioning my commitment to continue. Even that can be okay though, but that would probably be considered a failed run. (on the other hand, for some more complicated things, if the group performs better, a learning experience can be considered a success). A complete failure is a group that fragments before it reaches the instance, where somebody has to leave in the middle, or where a total self-centered ass wastes my time with his douchity. And ninjas.

Share and Enjoy:These icons link to social bookmarking sites where readers can share and discover new web pages.
  • del.icio.us
  • digg
  • Shadows

RSS feed for this page
since June 2007

RSS feed | Trackback URI

3 Comments »

No comments yet.

Name (required)
E-mail (required - never shown publicly)
URI
Your Comment (smaller size | larger size)
You may use <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong> in your comment.