AI is going to start picking people's significant others. Is that a good thing?

For thousands of years, people searching for love have asked intermediaries for help. Matchmaking is a very old profession, and many social events exist in every culture that are at least in part designed to give people looking for a soul mate an opportunity to find one. But that opportunity has long been constrained by geographical barriers and lack of information. Our perfect soul mate may live two villages away, but because we don’t know that, we will never meet.

This is why dating websites caused an almost immediate sensation in the early days of the Internet. The most popular among them seemed to offer a thick market for love—that is, a market in which a large number of diverse participants increases the chances for success. Eli Finkel, an expert on online dating and professor of psychology and management at Northwestern University, calls the members of this first generation of dating platforms the “supermarkets of love.” They assured that the marketplace was teeming with potential partners. Users liked that, but they quickly got overwhelmed by the effort necessary to find a mate among the masses—to identify “their” needle in the haystack.

In response, dating sites developed elaborate surveys and questionnaires to tease out preferences and help customers identify the best possible matches quickly and easily. Essentially, the sites switched to multidimensional information streams and implemented preference-matching algorithms. It seemed a sensible move, but it failed badly. It took participants hours to answer hundreds of questions about themselves, and the resulting matches were only marginally better than a random walk in the supermarket of love.

Online dating sites reacted as conventional markets might have done. New competitors assumed the problem was cognitive overload and decided that rather than more information, people wanted less. Much as traditional money-based markets have aimed to reduce the complexity of preference matching by headlining price (to give users a sense of comparability), more recent online dating platforms, like Tinder, narrowed the necessary interactions down to a single dimension—desirability.

Swiping left and right is its purest form. By condensing the decision to a single dimension, the process of matching gets easier.

But just as comparing prices doesn’t tell you everything you need to know when making a transaction, reduction to a single dimension does not guarantee a successful outcome when dating.

The problem is that in the attempt to improve mediocre results, dating sites dumbed down their service. Shifting to multidimensional information, it turns out, was the right move, but it wasn’t sufficient by itself. As Professor Finkel explains, the questionnaire process was too simplistic, looking just for similarities (or opposites), and it used the wrong data to boot. What’s necessary, in our parlance, are improved preference-matching algorithms and a better data ontology that capture how people relate to each other. Rather than asking customers to spend hours answering questions, future dating services will use machine learning systems that deduce the necessary relational data from video, photos, speech, and perhaps even wearable tracking devices. They will register when we smile or blush as we interact with somebody we like and know when our hearts begin to beat in sync. If a system gathers our preferences without much effort on the user’s part, and combines them with the right multidimensional flow of information and the appropriate matching algorithms, that will lead to a substantial increase in successful matches. These next-generation dating markets will be far more sophisticated and efficient—although perhaps less exciting.

The necessary techniques for reinventing online dating markets already exist; they just need to be combined correctly.

What’s happening in online dating is happening in other markets as well. Some markets are ahead; others still lag. But it is a change no marketplace that wants to stay in business will be able to resist. In just a few years, we’ll have at our disposal powerful data-rich systems that know us well enough to offer meaningful assistance with our market transactions. In return, having spent less resources and less time to get a better match, we’ll reap a handy efficiency dividend. But that’s not all.

For the first time in human history, we’ll have a choice about whether and to what extent we involve ourselves in certain decisions that are the bread and butter of human coordination, but not exactly the recipe for a satisfying life. We will be able to direct a machine learning system to do the boring stuff and reserve those decisions that give us the most joy and pleasure for ourselves. We will deliberately give up some of our choices so we can focus on the choices that matter most to us. As a result, we will be able to disentangle the need to decide from the pleasure of choosing.

But as we remake the market, infusing it with rich data, we also have to understand, and then rethink, the role of the firm.

#AI