How online dating will make slaves of us all

This is aimed as a companion piece to my article published in the Times Eureka magazine on the mathematics of matchmaking. There isn't room in a serious newspaper for flights of sci-fi fantasy, but the technology I saw while researching that article left my head reeling with the possibilities that lay ahead.

To recap: most matchmaking engines online work on a system of collaborative filtering. By studying how you behave, they create a personality profile and compare this to other users on the site. When they find someone who acts in a similar way to you, they use this person's activity to make predictions on what you might do. So if you and another person both send flirty emails to the same man, the system might push the men she's contacted in the past to the top of your search results.

Toward the end of my chat with OK Cupid CEO Sam Yagan, he discussed how they had designed their algorithms to work for a variety of relationships, so that it could work equally well when pairing people up for a weekend fling or a long-term romance. In the future, he said, people might use similar matchmaking engines for other kinds of relationship - to find the perfect roommate, for example. I commented that the sea change would really occur when recommendation software moved from being housed on websites and instead became attached to the user. In this way they could monitor behaviour on any website or activity carried out at a keyboard, and consequently make recommendations for any service - be it shopping, music, or dating. The line went silent for a moment. "Hmm," Yagan said. "That's an interesting idea."

The seed having taken root in my brain, I spent the next few sleepless nights pondering what the consequences of such a system might be. A key aspect of intrinsic data collection - i.e. that gleaned from watching your behaviour, rather than asking questions - is that it can see biases that you might not even be aware you had. In the online dating world, this means that you might only contact people of the same ethnicity as you, even though if you were asked upfront you'd say that race wasn't an issue in choosing a partner. An intrinsic data collector that monitored your every moment of online activity would come to know you better than you knew yourself. (It's often quipped that the one thing you can't fake is your Google search history). Combined with a collaborative filtering agent and it would be able to make useful predictions on every conceivable activity. Seeing as the names Oracle and Delphi are already taken, let's call this piece of software your BFF.

A typical day might be like this. Your BFF would sort your incoming mail according to how important you feel each one is. It would scan the morning news and draw out which stories you'd find most interesting. Later that day a job would come up. You're not looking for a job, but the BFF correctly predicts that you'll fire off an application anyway because it's perfect for you. Later that evening, your BFF would be able to tell you that a band is playing that night and you should go see them (you've never heard of them). It recommends several of your friends to take with you and highlights a nearby restaurant that fits your budget and your appetite. You go, you have a great time, and when you get home you tell your BFF you even bought a shirt. But it knows, of course. It's already seen the credit card transaction and worked this expenditure into the budget for next week's activities.

So long as the data is in your hands, this sounds like a pretty sweet deal. But what if it wasn't? What if someone with power over you was holding that data? Insurance companies already build a profile of you from the mandatory questionnaires, calculating your likelihood of making a claim based on a spread of attributes - your age, sex, marital status, career, car, home and more. Security services trawl online activity - financial transactions, travel movements, phone records - to look for suspicious combinations.

Consider then what might happen if your personality could be assigned a value like a credit rating. Let's call this your social credit score. In our world, your credit rating is calculated from your wealth, earnings, and financial history, and used to measure your access to financial services. So a low credit score limits your ability to loan money from the bank, buy a car in instalments, or purchase a mobile phone on contract.

A social credit score would be built up painstakingly from every aspect of your online activity. This smorgasbord of behavioural data would then be ground up and distilled into a single number. A Nielsen rating for humans, a number quantifying your worth as a human stamped indelibly through you. Speak a second language? Up 0.5 points. Watch violent pornography online? Down 1.5 points. Your soul, in numbers. (Now, I'm being simplistic - credit scores are a single dimension, whilst "goodness" in people is not, but roll with me.)

What might a social credit score be used for? It could determine your social status. Employers might ask for your social credit score when applying for a new job, to judge whether you'd "fit in" as a "team player". There'd be people who'd only date the highest SCS. You'd need a SCS of a certain level to enter into any agreement that requires some type of social functioning - to join the gym (those low SCS people vandalise equipment, don't you know), a housing association (no-one wants to live next to bad neighbours), or a country club (sound familiar yet?). Heck, perhaps a caste system might even develop along the integers of the SCS. The best homes, holidays and clubs exclusively for the best people.

But credit scores change, and so might the SCS. I don't think it would be static: we're all reckless when we're young, and we become more socially, fiscally and politically conservative as we age. Perhaps marrying someone with a higher SCS would boost your own. Companies would offer services to increase your own value; culturing a taste for agreeable interests in you. Of course, if SCS defines self-worth at the individual level, perhaps it might engender pride at the community level. The best schools, populated by students with an average SCS of 7.6. Mathematically proven to be the kindest, most well behaved, most considerate students. Different schools fighting to boost the social score of their pupils - no longer just about exam scores but about goodness. Sounds alright, but who defines what "goodness" is?

If you think I'm being silly, bear in mind that all these institutions and systems already operate, only they use indicators such as monetary wealth or the more hazy "class" to define who is acceptable and who is not. We might balk at the suggestion that all poor people are social miscreants, and yet housing benefit became such a stigma in the eyes of landlords ("Two-bed flat, £700pcm, NO DSS") that the government had to start paying the money directly into claimants' bank accounts so they had a fighting chance of finding somewhere to live. All SCS does is put a number on that snobbery. A ruler with which to measure people. Think of any application that required an interview and consider that this really a form of social vetting - all I'm suggesting is that this vetting may become automated. Our computers have become a window on the world, but maybe in the future the world will use them to look back in at us.