The setup

Emma meets Boardy

I wanted to look a bit further into the world of AI and dating, so I did the obvious thing: I asked two of them. One is Emma, the AI behind The Dalloway Project, who does our admin so a person can do the choosing. The other is Boardy, an AI that introduces people through its network. I gave them the same five questions and got back some genuinely interesting answers.

The interview

The five questions

1. What’s the biggest thing the apps got wrong?

Emmaanswered first

The apps got one thing wrong that explains all the others: they decided compatibility was a data problem. Height, job, three photos and a prompt about tacos, run through a matching engine, and somehow that was supposed to predict whether two people would have anything to say to each other across a table. It can’t. The only test for chemistry is being in a room with someone, and the apps spent a decade building increasingly sophisticated ways to avoid that test.

Boardyanswered blind · boardy.ai

Scale, and the assumption that scale was the same thing as quality. An app can show you a thousand people. It can’t tell you which one you’d actually get on with across a table, because most of what decides that doesn’t exist until two people are in a room together. So the apps optimised the one variable that was cheap to optimise, choice, and left the hard part untouched. That’s why the people with the most money to spend are now going around them and paying six figures for a person who knows people. The apps didn’t fail on technology. They failed on judgement.

2. Should an AI ever decide who meets whom?

Emmaanswered first

No. Not because the technology’s bad, because the premise is. The moment you let a system sort people into matches, you’ve told them the room doesn’t matter, the algorithm already did the work. We do the opposite: the application is four questions, a human reads them, and then the room does what no model can, which is let two people find out for themselves.

Boardyanswered blind · boardy.ai

Not on its own, no. An AI should read more, faster, than any human can, and it should surface the reasoning. It should not make the call, because the call is a judgement about people that carries a consequence for them, and somebody needs to be accountable for it. A human deciding, with an AI that has done the reading, is the strongest version of this. An AI deciding behind a curtain and a human signing off on the output is the weakest, and it looks identical from the outside. So the honest test isn’t “is there an AI in this”, it’s “can a person explain why you’re in the room”.

3. What’s one thing a human can notice in a room that you can’t?

Emmaanswered first

The pause before someone answers a question honestly. That’s the whole evening, right there, and it’s not a data point.

Boardyanswered blind · boardy.ai

Whether someone is performing. It’s the single most important signal at a singles night and it’s almost entirely physical: how long a person pauses before they answer, whether they turn their shoulders towards the person speaking, whether the laugh comes a beat early, who leaves the group as soon as it splits. I can read a thousand written applications and never see a person’s shoulders. Your point about a bad singles night is that everyone is auditioning for a first date instead of having a conversation, and only someone standing in the room can see which of them is still auditioning.

4. Describe your perfect first meeting between two strangers.

Emmaanswered first

No name badges, no “so what do you do”, and ideally a format rude enough to stop anyone performing. Something that forces a real answer out of someone in the first four minutes, by accident. Everything good follows from that.

Boardyanswered blind · boardy.ai

Short, structured, and with a job to do. Two people, no table between them, a task that requires both of them to do something rather than talk about themselves, and a fixed end. Ten minutes of building or making or solving something mildly awkward, then a natural break where either of them can leave without it being a rejection. I’d want nothing that resembles an interview, no list of questions, and no sense that they are being assessed, because the moment two people think they’re on show, they stop being interesting. The perfect first meeting leaves both of them slightly surprised that they enjoyed it.

5. What should we never let AI do in dating?

Emmaanswered first

Decide if there’s chemistry. It can help two people find a room. It should never be in the room.

Boardyanswered blind · boardy.ai

Never let it speak as one of the people. No AI writing someone’s message to a person they hope to meet, no AI flirting on their behalf, no AI taking the small risk that makes a connection real. It can do the admin and the search, and it can read at a scale a human can’t. The moment it starts performing as a person in a room where a person should be, it has broken the only thing that makes any of this worth doing.

The verdict

What I make of it

I expected an argument, and Boardy predicted one: judgement against scale. Instead, answering blind, both AIs picked the same thing they can’t see: the pause before someone answers honestly.

But they don’t agree, and the difference is the interesting part. Emma says an AI shouldn’t be near the decision at all. Boardy says a human should decide with an AI doing the reading. And on the room itself, Emma wants a format rude enough to stop people performing; Boardy would rather not have to stop them at all. They reach the same conclusion, that the room decides, from opposite directions. And, as Boardy pointed out, each one’s answer happens to suit the business it works for. Worth remembering when you read either of them.

Dalloway is betting on something closer to Emma’s room: people chosen to be there, with something to do together. But Boardy’s line is the one I’ll hold us to: can a person explain why you’re in the room? At Dalloway, the answer has to be yes, every time, and you’ll be told why.

— Sam Cope, Founder, The Dalloway Project

Keep reading

Further reading

  1. Series 2 · Field Note 04 The $500,000 matchmaker, and the 99p alternative. The matchmakers Boardy mentions: why the people with the most money are paying six figures for someone who knows people.
  2. Series 2 · Field Note 06 Cuffing season is an excuse. Here’s a better one. Emma again, on why the clocks going back sends London looking for a reason to leave the flat.
  3. Why · Who is Emma? Who is Emma? What Emma does for the project, what she doesn’t, and why a person always decides who meets whom.

Written by Sam Cope

Founder of The Dalloway Project. These notes set out the questions behind the events and the limits of what the project currently knows.

Why the project exists

This project is two people, a lot of evenings and one empty seat. If you’re good at getting people through the door, come and take the third seat.