When dating apps behave like this, users are simply shown more and more of the same type of person

Berman feels there’s something fundamentally unromantic about the narrowing of options that’s baked into collaborative filtering. It locks in what it thinks are your preferences early on, “which makes statistical sense but it doesn’t necessarily lead to serendipity”. And as Joel explains, this sort of paring away of desire is the opposite of how attraction actually tends to work over the course of our lifetimes: we don’t always know what we like until it’s put in front of us. “So you meet someone and you know you want them, and then when you’re asked in future what it is that you want in a partner you list their qualities.”

I tell Berman about Feeld, a newish sex-positive dating app that’s been jokingly hailed as the official sponsor of ‘Hot Vax Summer’ – it’s famously buggy and shows scant regard for users’ age and geographical preferences, but it’s getting rave reviews

Is there anything that can be done? Rather than striving to create bigger and more sophisticated databases of single people, Joel wonders if developers should actually be doing the opposite. “There’s a case to be made that the sheer number of options is a barrier,” she says. “Having endless possible matches can be quite inconsistent with the tools we’re equipped with – it’s cognitively overloading. And it’s very frustrating trying to sift through dozens and dozens of profiles that don’t give you the information you actually need. You wind up having to filter them using criteria you don’t actually care about as much.” Perhaps, she says, “a completely different approach would be better” – something that more closely mimics the way people weigh each other up in real life, with profile features like voice and video clips, or even a virtual space avatars could interact in.

Berman doesn’t hold out much hope for getting single people off the internet and into parks or bars to mingle. “It’s extremely challenging to say to someone in 2021 ‘do this thing that doesn’t take place on your phone’.” His advice to frustrated app users is simple. “Create a new account,” he suggests. “People have the best match rates in the first week or two of using a new app”.

With very little data to narrow the field, the app is forced to show you potential matches near-randomly, as the universe ordinarily would – and although they’re good for introducing you to people you might never otherwise have met, no app’s algorithm has ever been shown to have a qualitative matching advantage over analogue chance (if one actually did, it would instantly wipe out all the others). He wonders if it’s accidentally replicating the random matching conditions in which serendipity can flourish.

But even if dating apps themselves haven’t got any any better at making matches over the past 18 months, I reckon users might have

To me, it feels like Jeff Tarr and his successors have been trying to engineer the difficulty out of something that was never meant to be easy. Love, as Joel puts it, is “a chaotic process”, and you can no more fix that than you can get around the problem of human mortality. People have been forced into exchanging voice notes and making video calls before meeting, normalising forms of contact that can be a better guide to attraction. And maybe the months of self-reflection will have helped us to become more decisive about – or at least aware of – what we really need from our partners. I think a lot about the friends-of-friends who became an urban legend after catching each other’s eye on a Zoom quiz during the first lockdown. They decamped to a private chat, sent each other pizzas, and finally went on an actual date. In love, as in everything else, fortune favours the brave.

Ben Berman puts it even more bluntly. “There is something really seriously wrong with how dating apps work,” he says. Back in 2018, the game designer teamed up with Mozilla to build a game babel called Monster Match, which he hoped would reassure people faring badly on dating apps that this wasn’t their fault. Players create a cartoon monster profile and start swiping on other cartoon monsters; if you match, you can start chatting. In between each step, though, the game draws back the curtain to show how a typical dating app’s collaborative filtering-based algorithm narrows your options. Swipe left on several furry monsters, for example, and you won’t see any more, even if the reason you said no to them had nothing to do with their fur. Even more disturbingly, it reveals how this kind of matching can both draw on and reinforce bias to marginalise profiles belonging to those from racial, ethic and sexual minority groups.