I recently spoke with Stanford Linguistics and Computer Science professor Dan Jurafsky, co-author on the paper Making the Connection: Social Bonding in Courtship Situations. (See more details about that study, along with comments from my interview with co-author Professor Dan McFarland, in a previous blog post.)
I was interested in what can be applied to online dating based on their research on speed dating, which analyzed participants’ speech to note features that predict whether the two people “clicked” during their 4-minute date. Professor Jurafsky said we should be able to apply these same types of feature analysis to text, and since more and more dating sites and apps include chat (or texting) as part of the process, there will be a lot of data to look at.
How might this work? Let’s say you’ve chatted/texted with a number of people, but are wondering whether there’s any chemistry there, before you decide to meet up for a date. By running your chats through an algorithm that looks as lexicon (e.g., verbs discussing judging, feeling, etc; pronoun usage, such as the woman using more “I” and the man using more “you”) it can give you a chemistry score.
But I’m really interested in what we might be able to do BEFORE you chat with someone. What about using prior chat history to make predictions? It could look at your chats with people you’ve gone on dates with, and correlate those scores with whether or not you “clicked” when you met the person. For people you clicked with, it could compare their texting features with other people’s texting history to see if there would be a good match.
Some people might argue that there’s no way to predict chemistry using this method because it is only through the interaction WITH the other person that the information becomes accessible. Perhaps, given the rich set of data we’re building through chat, but it’s still worth exploring.