A Smart Algorithm Looked at 16 Billion Emails, And Here Is What It Learned
When you respond to an email, you mirror the sender’s email style
/https://tf-cmsv2-smithsonianmag-media.s3.amazonaws.com/filer/b5/e0/b5e0feea-61c1-4257-bca9-2d3c223784d4/42-63668445.jpg)
If you look at your inbox and feel doom and gloom, know that you are not alone. The feeling that you have too much email has an official name: email overload.
To better understand how we're dealing with the digital onslaught, a team of scientists from University of Southern California and Yahoo Labs used a machine learning algorithm to peer into the inboxes of 2 million Yahoo users. Over the course of a few months, the study participants sent 16 billion messages in total. The algorithm winnowed down that pile of digital messages to a few million sent between human beings participating in the study.
In addition to confirming the email overload is real, here's what they learned:
1. How you deal with email overload may correlate with your age. Older users tended to deal with the onslaught by replying to a fewer number. Younger users replied faster.
2. However old we are, we're clearly all glued to our computers and phones. The median reply time was 13 minutes for teens, and 16 minutes for young adults. Adults are barely slower, at 24 minutes. And those over 50 take a whole 47 minutes.
3. Looking for a substantial reply? Send a message in the morning. As the day goes on, emails get shorter in length.
4. Mirroring someone's body language and tone can make them like you more, psychologists say. Whether consciously or not, we mirror in the virtual world, too. Over the course of a conversation, email styles become more and more similar.
5. But on the other hand, reply times and reply lengths between pairs of people start out in sync, and then desynchronize over the course of a conversation.
Using that information, the researchers created a model to how long it would take a user to reply to an email. The model was accurate 58.8 percent of the time. And looking at an email chain in progress, the model was able to predict which would be the last reply of the email thread to an accuracy of 65.9 percent. A model like this one could help rank emails in order of important in a user's inbox, the researchers say.
(H/t MIT Technology Review.)