Skip to main content

Information Cascades in Runescape Polls

Or really, the lack thereof.

To put it briefly, Runescape is a MMORPG (massively-multiplayer online role-playing game) that no doubt a vast number of college students have at the very least heard about, or even played for some period of time. Earlier this year, Jagex released the “Power to the Players” poll system, to allow Runescape players to vote on the content of game updates. The way the polling system works is that players may choose among multiple options in a poll. After voting, players will be able to see the percentages of total votes each option receives. Finally, once the poll ends, the option (or options, if the poll allows for multiple winners) with the highest percentage(s) of votes wins.

What I found interesting about the game’s polling system is that it appears to be completely immune to information cascades, whereas many other online polling systems frequently do end up exhibiting information cascades (in much the same way as in problem 4 in the most recent homework).

As we saw in the most recent homework, we are able to delay information cascades by introducing different payoffs for alternative options. We can model the polling system (as in section 16.5 of the textbook) as  having two states G and B, where voting for a particular option is either a good or a bad idea, with payoffs for a player voting for or against a option as either being zero for rejecting and either positive or negative if they vote for a good option, or negative if they vote for a bad option. Finally, we have each player’s individual private signals that dictate whether a player believes any given option is a good or a bad idea.

Perhaps not surprisingly, the structures of the game and the polling system make it incredibly difficult for any individual player to deviate from his/her own private signal, so they will, even in the presence of opposing signals, choose to vote in alignment with their own private signals, which ends up preventing information cascades from ever happening.  Players tend to be incredibly well informed about what they want to see happen in the game. So in our model, it is as if the payoff for accepting an option they deem as “good” is significantly higher than the payoff they receive for accepting an option they deem as “bad”, even if there is a sizable portion of the playerbase that chooses the “bad” option, because they feel very strongly in favor of their own private signal. That being said, it is often the case that the “bad” options in these polls aren’t even all that “bad”. Rather, it is often the case that players will be selecting among several options that do end up resulting in a positive payoff to the player, but they are simply picking the option that results in the highest positive payoff to themselves. And perhaps because of the fact that many polls are actually choices between multiple good options, players tend not to feel any pressure to actually follow any kind of “majority”.

Hence, looking through the history of polls, it is very rare that in a poll between multiple beneficial options that there is a clear majority; the vast majority of such polls are actually very even, where the winner is selected by a margin of just a few percentage points over the second-highest option.

So, as one might deduce, the key to a poll where each individual is influenced solely by the his/her own private signal, rather than tagging along in an information cascade, is a poll where each individual is well-informed about their options and is able to convince themselves many times over exactly why the option they have voted for is truly the option with the highest payoff no matter what anyone else says.

Poll data accessed from


Leave a Reply

Blogging Calendar

November 2014