Text messages are the currency of modern-day courtship. But, when the texts start to become scarce, many of us search for signs of a relationship’s demise.
Instead of passing around your phone over brunch with friends, how about getting an app to overthink and analyse your text messages for you? Crushh is an algorithm-driven app that reads your text messages and according to its creator can interpret how much the other person likes you and detect any possible shifts in a relationship.
Creator Es Lee says the idea for the app came when he was sitting in a park with a “confused” friend who was new to the New York City dating scene. “He’d gone on a date that he thought had gone really well, but the woman hadn’t responded to his last text,” says Lee.
“I flipped through the text exchange and I could tell that she liked him from the ‘body language’ displayed in the texts. But that hadn’t been too obvious to my friend,” he continued. This textual body language refers to the punctuation, emoji and language used in a text, as well as the frequency of the messages and the time it takes to respond.
Lee decided to convert this analysis into an algorithm that works in a similar way to a human brain, minus the memory problems. The app takes the role of a friend and decodes text message chains using data and algorithms developed from analyzing more than 200,000 relationships and consulting with sociologists, psychologists and dating experts, says Lee.
To use the app, you need to select a contact and specify your relationship to that person. It then analyses your text messages already in your phone.
The app analyses sentiment, punctuation, emoji usage as well as message length and response rate to give you a score out of 5, telling you if that person likes you more, less or the same as you like them.
Machine learning algorithms are used to adjust scoring for personal behaviour, taking into account factors like age, and personality. “It takes into account personal behaviours,” says Lee. “Things like, the features he or she does and doesn’t use. It looks at individual patterns prevalent in a person’s messaging behaviour.”
The app also charts your relationship over time, and can point out any changes in messaging behaviour. Very useful if you’re trying to pinpoint the moment your relationship began turning sweet or sour.
Does it actually work, though?
Can an app really tell you how much a person likes you based on a string of texts? We contacted two experts and neither were entirely convinced.
Stephen Pulman, a professor at Oxford University’s Department of Computer Science, says the app looks like a “straightforward application” of sentiment analysis and emotion detection techniques, which can be “reasonably accurate.”
“The problem is that nuances like sarcasm and metaphor are still difficult for this kind of technology to detect. Problems may arise in the interpretation of the contents of a message,” says Pulman. “The only technical problem I can see is determining whether the topic of the text actually concerns the person or the relationship, rather than being generally warm about something else,” Pulman continues.
Furthermore, the technology needed to analyse a relationship via text messages is very advanced and likely still at the academic research phase.
“What I can say for certain is that the app doesnt understand the messages and it’s making a decision based on the syntactical attributes of the messages without understanding the semantics behind it,” says Mark Bishop, Director of the Tungsten Centre for Intelligent Data Analytics at Goldsmiths, University of London.
“We’re crowdsourcing intelligence about how a person texts” – Es Lee, creator of Crushh.
The app is only available on Android because iPhones don’t allow apps to access iMessages. Lee is hopeful that Apple will change this. The app also doesn’t pull in messages sent on apps like WhatsApp, Messenger or Snapchat, so if your convos are happening on multiple apps, your score and history won’t give an accurate reflection of your relationship.
Lee regards Crushh as a “diagnostic tool” for relationships, and a way to use data to pinpoint problems and potential areas to work on. “We’re crowdsourcing intelligence about how a person texts,” says Lee. He says there are no plans to sell that data and wants to convert that intelligence into advice and tips.
Who knows, maybe someday we’ll be able to use data to figure out why we’re being ghosted? Alas, if you’re an iPhone user, it looks you’ll have to keep going to brunch with friends for your answers.