The Proteus Paradox

by Nick Yee

  Skinner is well known for his theory of operant conditioning, but in a quirkier study, he induced superstition in pigeons. His goal was to show that complex human phenomena could be entirely explained by observable and measurable behaviors without recourse to internal cognitive variables such as desires, thoughts, or feelings. As you can imagine, Skinner was no fan of Freud, whose psyche (that is, id, ego, and superego) and defense mechanisms Skinner regarded as impossible to observe and requiring further explanation.

  In Skinner’s study, he placed pigeons into Skinner boxes and gave them food pellets using a reward schedule. The pigeons received a food pellet every fifteen seconds, no matter what they did. When the food was released, the pigeon was rewarded for whatever random behavior it was performing. As the bird repeated this behavior because of the reward, the food dropped again, and the behavior was further reinforced. In six of the eight pigeons tested, a clear superstitious behavior resulted. In Skinner’s words,

  One bird was conditioned to turn counter-clockwise about the cage, making two or three turns between reinforcements. Another repeatedly thrust its head into one of the upper corners of the cage. A third developed a “tossing” response, as if placing its head beneath an invisible bar and lifting it repeatedly. Two birds developed a pendulum motion of the head and body. . . . Another bird was conditioned to make incomplete pecking or brushing movements directed toward but not touching the floor.

  These pigeons behaved as if their ritualized actions caused the food to appear, even though this was not the case. The birds repeated the behaviors five to six times every fifteen seconds. Thus, even though only about 20 percent of the superstitious behavior directly led to food, the intermittent appearance of food was sufficient to sustain the behavior. One of the birds performed ten thousand ritual movements without any food reward before the behavior subsided.2

  High school students do not perform much better than pigeons; a fixed-interval reward led high schoolers to perform elaborate compositions on a piano keyboard. Similar superstitions appear quite often in online games. This makes sense because online games are a kind of Skinner box, and often the cause of rare but highly desirable game outcomes is not immediately obvious. When a superstitious behavior emerges, it is often inadvertently reinforced by the game. The abundance of levers and food pellet flavors in an online game makes it easy to confuse the underlying causal relations between actions and rewards.3

  Spawn Dances

  When a player kills a monster, it reappears after some time so that it can be killed again. Otherwise, monsters would quickly become extinct. This reappearance is known as spawning or respawning. Some monsters have fixed and rapid spawn rates (for example, ten seconds after death), while other have random and highly variable spawn rates (anywhere between one and a thousand minutes after death). In the original EverQuest, some monsters had spawn times of an hour or even up to six hours or more. Some rare monsters had placeholder spawns that further complicated things. A placeholder is a common monster that holds the place of the rare mob and thus prevents the rare mob from spawning until the placeholder is killed. To make things worse, many monsters did not have fixed spawn locations but could spawn in one of several locations on the map. This made it incredibly difficult for one player to know whether the rare monster he or she was waiting for had spawned or whether another player had killed it already. This slow, chaotic pacing was not a unique feature of rare monsters; it was a fact of life in EverQuest. A standard monster that took thirty seconds to kill had a respawn time of several minutes—time during which players often had nothing to do. In these many moments of mixed tedium and anticipation, spawn rituals were born. One prevalent superstition was the existence of an “anti-spawn” radius.

  In EverQuest, many players were under the impression that the re-spawn mechanic for monsters/NPCs [non-player characters] took into account players’ positions. So when people were fighting things in dungeons, they’d often leave whatever room they were in for a bit because they felt that the room wouldn’t respawn while they were there. [EverQuest, male, 24]

  It was widely believed that the game designers had implemented an “anti-camp radius” around major spawns, such that the mobs would not spawn if people were within the radius. Of course, no one knew exactly what the extent of this radius was, so more risk-averse people would camp further and further from the spawn point in order to avoid the radius. The developers at Verant found this so funny (there was no anti-camp radius) that they added a comment during some loading screens, “Checking anti-camp radius,” just to mess with these players. [EVE Online, male, 31]

  Another superstition was that the corpses of monsters were placeholders and needed to be looted quickly to speed up respawns.

  In Everquest it was a belief that you needed to loot all the corpses of everything in order for more mobs to spawn. This of course is untrue. The mobs spawn on a fairly precise timer and have nothing to do with crowding around the spawn area. [City of Heroes, female, 37]

  Finally, many players developed ritual “dances” for spawning:

  My favorite rituals would probably be the various “spawn dances” in EverQuest. . . . They varied wildly—some people had special gear sets they used, others had sets and sequences of movements and animations (via animated emotes, spellcasting, terrain), ways to move or not move (must stay sitting, still, as much as possible; or must move continually/every X seconds), etc. [World of Warcraft, male, 23]

  Some players would sit and stand rapidly while strafing back and forth. Others would crouch and run in circles or figure-eight patterns. Jumping seemed also to be a common theme. Seeing a full group of six characters dancing in this manner shortly before a mob was to spawn was very funny. I think that it sometimes was done as a joke, but I knew some players who swore by its success. [EverQuest, male, 28]

  In the same way that pigeons dance for food in boxes, people dance for monsters in online games.

  Dungeon Seeding

  In World of Warcraft, there is a chance that bosses will drop highly valuable pieces of equipment in difficult dungeons. Dungeons in World of Warcraft are known as instances because player teams each enter their own version of the dungeon. So if fifty different teams are running the Molten Core dungeon, the game creates fifty parallel, independent versions of the dungeon, one for each team. Bosses in the game have set loot tables—that is, different probabilities for a predetermined set of loot, of which several may drop if the boss is killed. Because different classes use different types of equipment (for example, rogues can use daggers, druids can use staves) in the game, many players often leave a dungeon run empty-handed. Boss drops are thus low-probability events with highly desirable outcomes that elicit superstitious behavior.

  There is a widely held belief that instances are “seeded” despite lack of evidence and even a direct denial from Blizzard. Seeded refers to the person who starts the group or raid, and it is believed that the class of that person directly impacts what class specific loot will drop. I.E. if a warrior starts the Molten Core raid invites, more druid and warlock gear will drop. If a priest starts the invites, more warrior and mage loot will drop, etc. [World of Warcraft, male, 33]

  Blizzard consistently states that loot drops are completely random. Yet a lot of people don’t believe this because some items drop over and over when under one Master Looter and different items would drop over and over when under a different Master Looter. [World of Warcraft, male, 34]

  A variant of this belief is that certain characters are luckier or have better loot tables if they are allowed to seed the instance.

  We have a particular guildmate who insists that when he enters the dungeon instance first, better loot will tend to drop. Granted, when he has entered first, we’ve received some very nice, even legendary items in World of Warcraft, but to think he’s somehow affecting the loot table by being the first to enter is a bit much. [World of Warcraft, male, 30]

  [There is] the belief that certain classes seed certain loot in Pv
E instances within World of Warcraft and that certain players are “lucky” seeders in terms of an increased high-level loot drop rate. Sometimes, raids have been held up until these lucky seeders or a member of a certain class arrives at the instance entrance. [World of Warcraft, female, 33]

  Silliest is that a particular person provides some sort of luck to getting loot—that one person is responsible for the “seed” being good or bad. [World of Warcraft, female, 49]

  In short, some characters in online games come to be viewed as being inherently lucky.

  Lucky Charms

  In addition to the belief that certain people are lucky, there are pervasive superstitions around items that confer luck. The specific items differ from game to game, but these superstitions take the same general form.

  In World of Warcraft there are 2 items that are said to bring luck to the owner. These are the “Rabbit’s foot” and the “Lucky charm.” These items drop off common mobs around the world. There is a group of players that strongly believes that carrying around one or more of these items increases your luck in loot drops. People often use specific events and strokes of luck to prove that they “work.” I myself don’t believe it has any effect at all but still have a “Rabbit’s foot” in my inventory because you never know. [World of Warcraft, male, 41]

  [Some believe] that carrying or owning items whose names implied good luck (Fortune Egg, Millionaire’s desk, 4 Leaf Mandragora Bud) would increase drop rates despite no evidence to prove this. I’ll admit to doing it myself! [Final Fantasy XI, female, 25]

  In Anarchy Online, some people believed that wearing certain gear was the way to gain certain drops and would spend hours farming gear so that they could farm other gear. [Anarchy Online, male, 33]

  What’s particularly intriguing in this set of narratives is that some players explicitly state that they do not believe in the superstition but follow it anyway.

  Over-Enchanting

  Another type of high-risk action occurs in games in which players are allowed to over-enchant equipment. In many games, players can enchant an item to give it a bonus, whether to combat skills or the character’s traits. In some of these games, the player can enchant the same item multiple times. Once the player has reached a certain threshold, there is a chance that the item may be destroyed in the process. Over-enchanting refers to the process of enchanting an item beyond its safety threshold. The risk of item destruction is proportionate to the number of existing enchantments over the threshold. Given the daunting risk of destroying a valuable piece of equipment, over-enchanting is a high-risk gamble ripe for superstitions.

  In Lineage II . . . enchanting to +3 is risk-free. However, at +4 and above the item has a chance of breaking, causing you to lose a lot of money. Many people have gone so far as to quit the game or reroll after blowing up their ultra-expensive gear. A very prevalent superstition is for people to take the item into a church when attempting to overenchant it. Many people, if they were successful over-enchanting an item at a certain spot, will return to that spot every time they need to over-enchant. [Lineage 2, female, 24]

  In addition to location-based superstitions, ritual behaviors conducted before over-enchanting have also developed in other games.

  In Ultima Online it has been stated many times by the Dev[eloper]s that “eating” does nothing to enhance the characters’ abilities. Many players still choose to eat before they try to do some specific crafting where the risk of destroying an item for example is high. [Ultima Online, female, 45]

  Some go to only a particular NPC—some will not only upgrade at only a certain NPC, but also upgrade ONLY within a certain time period—some do it ONLY while standing on a “lucky” spot, yet others believe that the secret is to wait there patiently till someone comes in . . . then wait for him to fail . . . they believe that their attempt will be 100 percent successful if it follows on the heels of someone succeeding. I personally am guilty of a fairly weird ritual myself—I tend to strip off all equipment I am carrying and log off in between EVERY attempt to refine my gear.:) [Ragnarok Online, male, 29]

  Here’s a final example of a crafting superstition from Final Fantasy XI that also hints at why it is so difficult for superstitions to go away once they begin.

  One of the most persistent superstitions (and for all I know, it might be true) was that facing in certain cardinal directions would affect how your crafting came out. It was the perfect superstition, because it took so little effort to follow that even if it wasn’t true, you didn’t lose anything by acting as if it was true. [Final Fantasy XI, female, 23]

  Treasure Negotiation

  Superstitions, pervasive across online games, develop wherever a high-risk or low-probability event leads to a highly desirable outcome. This scenario is common in online games, whether it is valuable loot from a boss, over-enchanting a weapon, or having a rare monster spawn. Many of these superstitions persist despite limited or anecdotal evidence or even direct refutation by game designers.

  In Dungeons and Dragons Online, diplomacy is one of many skills that a character can learn. Game designers intended for players to use the skill on computer-controlled characters, allowing for alternative conversation paths as well as the distraction of enemies during combat. A programming error made it possible for players to use the diplomacy skill on treasure chests, although doing so had no impact on the game. Heather Sinclair, a member of the development team, has publicly discussed the aftermath of this programming error.

  From beta all the way through months into launch players were CONVINCED that if you used the diplomacy skill on a chest it would improve the loot you got. . . . This was SO widespread that you literally could not get in a pickup group without them querying about the diplomacy skills of the party and someone forcing everyone to wait while the highest diplomacy skill player cringed before the chest sufficiently.

  This superstition became so pervasive that the game developers decided to debunk it publicly. The public statements, however, had the opposite effect:

  No matter how many times we posted on the forums that this was a myth and it doesn’t do anything, they kept doing it. It got so bad our community relations manager even put it in his [forum signature]. Finally we made chests an invalid target for the diplomacy skill, then players whined that all the points they put into diplomacy were worthless because we “nerfed” the skill!

  Not only are superstitions prevalent in online games, but some are also incredibly resilient to debunking.4

  The Social Reinforcement of Superstitions

  In several of the player narratives, people who claim they don’t believe in the superstition nevertheless carry out the superstitious behavior, just in case. Social factors also help sustain these superstitions. The most significant is the relative low cost of the ritual compared to the relative high value of the potential reward, especially in situations in which the team members have nothing else to do to fill the time. After all, if you get to run a difficult dungeon only once a week, what’s the harm in trying something that takes just thirty seconds?

  Generally the experimentation is harmless enough that it is at least permitted by skeptics of the theory. [World of Warcraft, male, 24]

  There is also the relative cost of trying to debunk a superstition. In a typed chat setting, it takes much more time and effort to argue and attempt to debunk a superstition than to simply follow along, even if you don’t believe in the behavior.

  If the potential outcome is negative rather than positive, risk aversion comes into play. For example, there are superstitions that make a boss easier to kill and thus decrease the odds of a raid wipe—the obliteration of the entire team by a tough encounter.

  “Hey kids, don’t use curse of weakness on Gandling, because he starts teleporting people a ton faster . . . .” But nobody wanted to try it out; I remember actually offering to pay people a gold each to let me try . . . and they refused; . . . people are very pious when it comes to respecting these technological taboos. [World of Warc
raft, male, 23]

  In Skinner’s pigeon study, superstitious behaviors persisted even though they did not produce food 80 percent of the time. Even a low contingency rate was sufficient to sustain a superstition. The same is true in online games. After all, a ritual that produces highly beneficial outcomes 20 percent of the time is still worth performing. Indeed, probabilistic superstitions are hard to debunk without a large experimental data set, which few players would have the time or tenacity to collect.

  If it worked some of the time, it was enough for the group in question to continue to think that the process they were following was crucial to the success of whatever it was they were doing. [EverQuest II, male, 36]

  With a group of five people, the likelihood that the superstition has recently been true (that is, reinforced) for any one team member is very high. This secondhand reinforcement also creates the illusion of a much higher success rate.

  Old Dogs and New Tricks

  Superstition in online games reveals something very important and fundamental about how people interact with new technology. To help us unravel this, let me describe a study that changed how we think about human-computer interaction.