Wednesday, December 10, 2008

Fuzzy Logic on Facebook

I just clicked yet another "Maybe" button in response to a Facebook event invitation:

RSVP box with 'Maybe Attending' selected

The event shows 60 definite yeses and 57 maybes. For someone planning an event, that's a pretty big amount of uncertainty to take into account. You don't know how many of those maybes are "it's a weeknight, so probably not, but you never know" versus "I'll totally be there but can't commit in case someone offers me Elton John tickets at the last second."

The interface should have a slider input representing the user's best guess as to the probability of attendance. It would still show up as "Maybe Attending" to the event creator and any fellow invitees, but the actual likelihood of your attendance would be stored. Then that could be used to predict the number of attendees with a little more accuracy (three people 40% likely to attend plus one person 80% likely to attend equals two expected guests).

In addition, the software could recognize events that overlap. So if you give a definite "Attending" to one event and it sees that you've already said "Maybe" to another event at the same time, then it could prompt you to change the "Maybe" to a definite "Not Attending" (and if you demurred, the software could quietly downgrade your probability of attendance, so you could be on the record as "Maybe Attending," with all the social courtesy that implies, without distorting the expected guest count).

Actually, what would really be the most helpful would be to randomly send out surveys to people who create or are invited to events. Netflix does a similar thing when it asks you when a certain movie arrived, or when a return was mailed back. So every now and then you would get an email, "you said you would Maybe Attend event x; did you end up going?" or, "we predicted 73 attendees; was that too high, too low, or about right?" And then they could revise their attendance predictions, even taking into account things like whether an event is on a weekend or is really late, and how many conflicting events each invitee has.

Of course, maybe empirically people who are "Maybe Attending" just show up 40% of the time or something, and you can just use a heuristic and get just as accurate a result. But in principle, you know.

3 comments:

DU said...

Actually, it shouldn't even have to ask you if you are attending or not, or what the probability is. I should just get an email like this:

"Melissa is having a birthday party this Saturday. Others like you are attending at a rate of 95%. Her birthday has been appended to the top of your social queue. Black tie."

Or this:

"John's kegger is 87% popular with the general public but only 3% with people with your partying history. Here is a list of libraries and art museums in your area."

DU said...

Another thing that could be automated is the follow-up comment box.

Dave said...

maybe is a terrible option. Just tell me if you're coming or not, jerkoffs! Perhaps an I don't know option that sends you reminder emails weekly a month before, and daily the week of.