Do you really know your friends?

Think of someone you know. It can be anyone, just pick the first person that comes to your mind. How close are you to them? Do you know whether their political beliefs lean more towards conservative or liberal? Do you know how they would answer the question "Do you feel the U.S. spends too much money on assistance to other countries?"

During the 2008 primaries, Sharad Goel, Duncan Watts, and I released a Facebook application that asked questions like these about the users' political beliefs (and some fun, non-political questions like "Would you pay $100 for jeans?") and then let them answer the same question about a randomly chosen friend. If the friend had agreed to share his or her answer, they could see whether they were right or wrong. Most people who played this guessing game enjoyed it, because they could see how well they knew their friends and see how well their friends knew them.

There were 80 unique questions, 47 of which were about politics. We had over 2500 people use the application, who together answered a total of 12,000 questions about their friends. The first thing we found was that, as you would expect, friends are more similar to each other than they are to strangers. On average, people answered the same way as their friends 75% of the time, relative to strangers who answered the same way only 63% of the time This indicates substantial homophily—that is to say, "birds of a feather flock together."

Now, you may notice that 63% agreement is pretty high. This tells us that the people who used our application are more like each other than two people randomly selected from the U.S. To adjust for this, we fit a model that allowed us to answer what we would have gotten if the questions we asked were perfectly divisive; that is, if random strangers only agreed 50% of the time.

Real and perceived similarity of friends

We also asked whether they discussed politics with their friends, and by using the number of mutual friends the two people shared, got a rough estimation of how close they were; we said 20 or more mutual friends indicated a "strong" tie and 3 or fewer mutual friends indicated a "weak" tie. As you might expect, people tended to agree with their strong ties more than their weak ones. Relative to the 50% baseline, weak ties agreed 63% of the time (so still a fair bit more than strangers), but strong ties agreed 73% of the time (on average, 67%).

However, people thought they were even more similar to each other than they actually were. As you can see in the above figure, the difference between perceived agreement—that is, how often people guessed that their friends would answer the same way they did—and their actual agreement is about 10% on average, with people overestimating their similarity to strong ties by about 7% and to weak ties by about 13%! Thus, although people are fairly similar to their friends, they consistently underestimate how different they can be.

The accuracy with which friends guessed each others' beliefs

Another question to ask is how often people correctly guessed their friends' beliefs. On average (again adjusting for the 50% baseline), we found that people were correct about 74% of the time. However, much of this could be from lucky guessing, and in fact, this is basically what we found. Looking only at the questions on which the friends actually agreed ("sensitivity") they were right 90% of the time. For the questions they disagreed on ("specificity"), people correctly predicted their friends' answers only 41% of the time—significantly worse than chance (see the above figure).

What this means is that people were generally unaware of when their friends disagreed with them! Some might say this is unsurprising, as it is impolite to discuss politics... except that these results hold even for people that said they discussed politics with each other. It is simply that people believe their friends agree with them more than they actually do.

Real and perceived agreement for majority and minority opinion holders

There are two potential reasons people might do this. One is projection: they are simply projecting their own beliefs on their friends. Another more subtle explanation is stereotyping: they believe people of a certain "type" believe certain things, and it just so happens that their friends are the same "type" that they are. To figure this out, we used the model to look at instances where people were in the minority (only 40% of the population agreed with them) and instances where they were in the majority (60% of the population agreed with them). We found that while people in the minority guessed their friends would agree with them less than the people in the majority did—reflecting an awareness that their beliefs were in the minority—they overestimated how similar their friends were more than the people with majority beliefs (see the figure above).

Relationships Essential Reads

10 Ways to Spot a Love Bomber

Relationship Resentment

What all of this suggests is that when people are trying to figure out what their friends believe, they tend to rely on what kind of person they are and what they themselves believe, and not very often on real knowledge. This implies that although friends actually have people around them with differing opinions, because of this assumed similarity there seems to be little active discussion that actually has an impact on their beliefs. So now you have to ask yourself: do you really know your friends?

*The work behind this blog post and the figures in it will appear in a forthcoming issue of the Journal of Personality and Social Psychology.