Pin It
Artificial intelligent crush app Mei
Still from Ghost World (2001)Via IMDb

Using artificial intelligence to find out who fancies me

One writer tests two apps that decode your texts and analyse how you and all the fuckboys in your phone match up

“I’ve been meaning to message you,” opens a break-up text from an ex. “Sorry for my ridiculously slow reply,” reads another. And my favourite, begins: “Yo sorry my bad, been a bit of a wasteman. Not parring you off.” He was, definitely, parring me off.

These exchanges are searingly familiar to most people who find themselves on dating apps in the current hellscape – the headscratching, soul-searching, and analysing of old messages to find out where it all went wrong. Now it turns out these stunted back-and-forths – spread across a three year period – could have all been avoided had I utilised artificial intelligence. Well, maybe. 

Messaging app Mei, which debuted on iOS earlier this month, uses AI to scan your WhatsApp messages and tell you how likely it is that someone has a crush on you. Originally launched on Android, the full app – which isn’t available on iPhone because of Apple’s privacy settings –  also enables users to analyse conversations in real-time, offering live advice on how best to communicate with someone.

“There’s so much potential for miscommunication over text,” the app’s creator Es Lee tells Dazed, “meaning you often end up ruining a relationship, or not giving it a full chance. People don’t realise that others have different ways of communicating, and Mei is a reminder of that. It aims to help you avoid misinterpreting things.”

The Android app came out in September last year after Lee offered his friend advice based on the sub-text in flirtatious messages they were receiving. Lee realised “there’s a lot of things you can measure with algorithms”. As well as live-scanning text conversations, the full version offers a personality profile, automatically generated by AI from your messages, that shows users how they come across in texts. Dialogue is scored based on: openness, conscientiousness, extroversion, agreeableness, and neuroticism. Outside of its optional AI assistant, Mei is simply an alternative app for messaging. At the moment, the iOS version can only do crush-scanning, which costs $8.99 (£7.20) a pop. Each scan adds to the artificial intelligence’s capabilities – if you’re feeling dubious about swapping that sweet, calculated knowledge of your ex’s true feelings for your data privacy, X out now.

As a loyal iPhone user, I wasn’t able to test the full version of Mei, but – for entirely journalistic purposes – I let the app analyse how likely it was that the following people had a crush on me: three close friends, someone I’m currently dating, someone I thought might fancy me, and a couple of exes. I also did a bonus test on my mum. As well as offering a percentage of crush likelihood, Mei offers one line of advice about how to improve communication to get the other person to be DTF (mum excluded here, obviously). 

Admittedly, the results were pretty accurate based on what I’d predicted: a friend who has a long-term girlfriend had a zero per cent chance of fancying me, someone I used to sleep with had an 81 per cent crush on me, and the guy I thought might fancy me scored the highest with 86 per cent (sadly, he’s not for me). I was refunded 50 credits for my current fling because apparently we’re ‘communicating well’, although the test on my ex also returned this result, despite the last message in our chat being him dumping me (LOL). My mum absolutely savagely only had a 29 per cent crush on me, but the app attests that if I was more myself, she would supposedly “like me for me”. Huh?

The only erroneous scores came from the tests on two of my close friends, which suggested both of them fancied me – something I know for a fact isn’t true (or do I?). Though, given the intimate nature of our messages and the frequency of which we text, it’s hardly surprising the app predicted a close bond. When I questioned Lee about Mei’s ability to differentiate between romantic relationships and close friendships, he said: “We think of this when we programme these algorithms, but with AI labelling things, it’s kind of out of our control. Basically Mei will ask the person: ‘This seems like a close friend?’ They go, ‘yes’ or ‘no’, and eventually the algorithm will start picking up patterns, developing its accuracy.”

“The more we end up just texting and not talking, the more necessary it is for something like this – something to sit on your shoulder as a guardian angel” – Es Lee, Mei creator

Mei isn’t the only app employing artificial intelligence to enhance relationships. With the aim of making its users “communication Jedis”, Keigo helps people navigate social situations, offering recommended behaviour based on personality types. The app can use your Twitter account or a portion of text you write to analyse your personality, before doing the same to anyone you input, ultimately providing advice on how well you match. Unlike Mei, Keigo isn’t specifically for relationships, instead intended to be used with any ‘conversation partner’.

With just three free goes, I could only analyse a friend, an ex, and the guy I’m dating. All of them had similar personality types to me – we’re all ‘idea-rich visionaries’, who knew? – but annoyingly I matched highest with my ex, who is 93 per cent compatible with me (even though I literally inputted a break-up text). Each match is compared to you based on four personality types: quality-mindset analyst, action-oriented driver, harmony-seeking diplomat, and idea-rich visionary. Keigo tells you what behaviour is most typical for that person vs you, before outlining your chance of misunderstanding, and suggesting how to behave in conversation with them.

Although Keigo was less fun than Mei – and for some reason felt less accurate – it definitely offers more practical advice for users, and because it doesn’t just focus on romantic relationships, it’s broader in its reach.

With both of these apps, you have to share significant personal data – namely highly personal messages – which, at a time when there’s a new data breach almost every week, raises a serious concern about users’ privacy. Though Mei’s website asserts that the app has no plans to sell users’ data. “This is the most private information you have,” says Lee, “so Mei doesn’t ask for your identity – we don’t need your name to help you. We’re not going to hoard the data, we make it easy within the app to delete the conversation if you want.” 

As well as security worries, the impact of these apps on IRL communication is brought into question. Is a real-time conversation booster really helpful, or just a distracting hindrance? Given a lack of physical connection is one of the main reasons for our current loneliness epidemic, you would think the logical solution is to spend less time analysing texts and more time actually meeting up. “We can’t change the fact that as a society we’re becoming more and more digital,” argues Lee. “The more we end up just texting and not talking, the more necessary it is for something like this – something to sit on your shoulder as a guardian angel.” Apps are increasingly being used to have real-world effects, from finding love and friendships, to organising protests and getting mental health counselling

Although it might feel like widening the gap between the digital and the real, lessons learned via AI advice could eventually improve human interaction. Plus, Mei’s logic of embracing this digital adjustment as opposed to fighting it can actually make positive change outside of the complete app’s intended uses, for example spotting the warning signs of suicidal tendencies. “Not only has Mei identified when a user is being more negative,” Lee reveals, “but it goes into a whole dialogue where it asks if you’re OK, and searches your entire text conversations to find the exact right person to help you at that time.”

These apps are bound to be met with a mixture of excitement and fear; sure it’s fun to humour the idea of your mate having a crush on you, but it’s also terrifying to think that the more information we offer AI, the more intelligent it’s going to become. There’s also the sweet and hilarious irony of needing to use artificial intelligence to understand each other after spending so much time improving digital technology so we didn’t have to speak to each other. Anyway, in conclusion: the world is stupid, and loads of people fancy me.