Gender, Age and Relationship Predictions

With our latest beta release, we’re proud to introduce machine learning predictions for the age and gender of users and contacts.  What we learned early in studying text messages was certain demographics like age and gender tend to correlate with the way we text. 

The chart above tracks average characters per text by age across several thousand Americans.

For example, guys tend to use “hey” while girls tend to say “hi”, and the older you are, the longer you text messages tend to be.  With this version, Mei is also able to predict the type of relationship you have with the person you’re texting – whether they are a friend, family member or crush (someone you’re romantically involved in).

Looking at men and women separately, there doesn’t seem to much of a difference between genders up until the early 20s.

Why Do We Do This?

The purpose of Mei, our AI assistant, is to help you with your relationships.  But since different people tend to communicate differently, it’s helpful for Mei’s algorithms to know certain things about the people in the conversation.  That way, they could try to adjust for biases that are simply a function of demographics when she encounters them.  In subsequent releases where she will give users advice, having this information will help better tailor that advice.

We don’t want to ask you for all kinds of personal information up front, so we use the data that we already have to make predictions about age, gender and relationship type.  Our algorithms aren’t always going to be perfectly accurate so by answering these questions, users will help us get better over time.

So, don’t take offense if Mei thinks you’re 25 when you’re actually 47.  You’re just young at heart (in your texts) ☺️