Our class discussion on information cascades and pluralistic ignorance reminded me of a situation that occurred several months ago that exemplifies these issues.

Towards the beginning of Winter Break, a high school friend used Facebook to invite around 140 people to a party at a beach house. The invitees included the entirety of our high school class, along with 25 or so of the host’s friends, who belonged to other, separate, social networks and were generally not known to those in our high school class.

Facebook, along with Evite and other social event planning sites, has a very interesting feature for events, which is that, by default, the attendee list and RSVPs are public. All guests are listed in one of four categories: confirmed guests, maybe attending, not attending, and awaiting reply (the person hasn’t yet RSVPd). Because the party was to be held at a beach house about an hour’s drive north of the city, it was essential that guests decide in advance whether to attend. And because, as was said in class, when it comes to a party, political protest, or any other network decision-making problem, “If you don’t know how many people are going to show up, you’re not going to show up,” the guests at my friend’s party relied on Facebook’s RSVP data to see who else was planning on attending.

Normally, when deciding whether to attend an event, people ask only their own circle of friends whether they plan to attend. However, there is more than a strict information cascade at work here: people may also try to convince their friends to attend if they would like to attend, pressuring entire clusters of the network to come. However, Facebook, Evite, and similar sites break this pattern: in addition to the high/low signal from one’s close friends, one receives a signal from the entire network in the form of RSVP data. If few people have indicated that they plan to attend, then one may not even bother to talk to their friends and attempt to convince them to come. On the other hand, if many people RSVP that they will attend, a cascade of attendances could be created, and a large number of people will accept the invitation. Essentially, the public display of RSVP data makes the first few respondents crucial, because the responses of the entire network are visible to all invitees. If the first few responses are unfavorable to the other guests, a negative cascade will be created and few are likely to attend, or if the first few responses are favorable, will may attend.

So what happened to my friend’s party last December? Unfortunately, a negative cascade quickly developed. Initial responses generally fell into two groups: negative responses from those who would be out of town or unavailable, and a small number of positive responses, almost entirely from the host’s friends who belonged to separate social networks and were unknown to many of the guests. As a result, invitees viewing the event page saw a number of familiar names not attending, and a small number of unfamiliar names attending. Given this, most concluded that they were uninterested in the party, and avoided responding entirely: a negative information cascade. When I asked my own circle of friends on the day of the event about their plans, nearly all expressed ambivalence, in most cases wondering outright whether anyone we knew was coming. Many seemed to want to attend, but Facebook’s display of RSVPs led them to receive a negative signal and avoid attending. Relatively few people actually attended the party.

In this case, the additional information provided by Facebook and the public display of RSVP data actually resulted in less information being available. Rather, the RSVP data created a sort of pluralistic ignorance where attendees were afraid to commit publicly to attending or not attending, leading everyone to assume that no one was attending. This created a cascade of negative signals and therefore few attended. This lesson should be heeded by social networking sites and their users in deciding whether to publicly display RSVPs on the site.