Consider falling in love from the perspective of your phone.

Facebook already knows what it looks like. A search for a name; a new friend request; the growing count of shared likes and comments. Eventually, the opening up of a private chat window. Conversations grow in length and frequency, good time-on-site data and an uptick in ads served. You get tagged in each other’s photographs; other friends are tagged less. The distance between the square in your photos – those that map your face – grow closer and closer together. And then, after days or weeks of your GPS signals and check-ins at the same events and bars, both phones go into sleep mode logged into the same wifi network.

Relationships between human beings are a complicated process of discovery and adaptation. Emotional learning takes time. It demands information about yourself and your partner. You learn to understand each other, and anticipate each other’s behaviors.

When a couple says “we complete each other’s sentences,” well, guess what: so does your iPhone. Machine learning is a process of analyzing and recognizing patterns – then making connections – across sets of data.

Data is the residue of life, told through what French philosopher Bruno LaTour named “an accumulation of traces.” By constantly holding your phone’s hand, it accumulates the most intimate of these traces.

With the  !Mediengruppe Bitnik exhibit and upcoming symposium on human-chatbot intimacy at the swissnex Gallery, we looked at AI as it applies to the “data” of human connection – and if, someday, we could have human connection without other humans. 

Feels Like I’ve Known You Forever

For a machine to identify a bird, it needs data from thousands of images of birds. Whatever can be transformed into data becomes a lesson: shapes, colors, behaviors. An ideal program tells us not just what a bird is, but where a bird might be. Missing values become hidden variables: show the machine a tree and a nest, and there’s probably a bird somewhere.

For dating, social media has created a perfect data set: the collection of the things we tell the world to make it like us. LaTour writes, “It is as if the inner operating modes of the private world have been broken open, because by now their input, and output, are completely traceable.”

Could some online dating algorithm determine patterns in the profiles you respond to, shared interests that inspire you most, the types of actions that draw out the greatest spark? Of course. The big question now is philosophical: do these traces of our lives really have much to do with who we love? Or are these patterns, preferences and choices as arbitrary as a phone number? With enough data, we might find out.

Knowing Me, Knowing You

Kang Zhao, assistant professor of management sciences in the University of Iowa Tippie College of Business, has been betting that your data knows you better than you do. And he’s used that idea to help humans find each other.

He’s designed a test algorithm to find romantic connections by analyzing patterns in the partners people contacted. This would be obvious, but he cleverly ignored any criteria the daters themselves claimed to be looking for.

You say you want to go mountain climbing, but you keep messaging people based on their film taste. This machine knows.

The algorithm increased mutual interest by more than 50%

In another project, Zhao used the two data sets, and connections between them, to determine who the dating service would “recommend” to a user. In this approach, the missing figure is selected through a correlation of data sets: if Ale(x) liked Luc(y), and (Z)ach liked Luc(y) and Katrin(a), then Ale(x) might like Katrin(a), too. Data! 

Scratch the surface beyond the use of data and look at the collection of data, though, and a depressing reality sets in: none of this data comes from successful relationships.

You really have to fail often to generate your data into a recognizable patterns. For monogamous couples, the “last person you date” is a single data point; if it goes well it may never need to be repeated. For the rest of us, a machine can only learn the kind of person we’ve been attracted to before: patterns that have already failed.

R u up?

Ashley Madison was a dating website pitched toward married men. (The site is at the center of !Mediengruppe Bitnik’s exhibition here at swissnex San Francisco). That site seemed to embrace the cynicism of failed matches to unusual ends.

A 2015 “Impact Team” hack revealed embarrassing information not just about its members, but about how the site was managed.

With many more straight men on the site and a statistically nonexistent population of women, the hack revealed that many men were paying to interact with women with a secret: virtually all of them were robots

The Ashley Madison site noted that it was an “entertainment” site, and never promised that you’d connect to anything more than a collection of text files. Bots would send male members messages (“are you online?” being a classic come-on) and men would pay to reply. The messaging is anything but complex: “Hi,” “Hello,” “Free to chat?” etc. Once engaged, they would write something a bit longer:

“I’m sexy, discreet, and always up for kinky chat. Would also meet up in person if we get to know each other and think there might be a good connection. Does this sound intriguing?”

“I might be a bit shy at first, wait til you get to know me, wink wink :)”

After that, men could pay to reply, or purchase “gifts,” but the women would disappear or stop responding altogether.  

It’s unlikely male users were ever really “fooled” into believing these bots were real women. It seems more complex, something like an emotional AR. Ashley Madison created the environment where reality was augmented by fantasy. These men simply filled in the gaps.

What’s NEXT?

Perhaps what’s most disturbing about Ashley Madison’s model is what it says about the market for online dating. The site is a cynical product of a commodified cycle of love and romance. Build recommendation engines for real human connections using data from past failures, then charge money to interact with robots pretending toward an ideal.

In a world that is fast becoming an arms race between loneliness and connection, programmers, soon enough, will scrape enough data to push all of our emotional human buttons. This won’t require mass leaps in current tech – we’re clearly more than willing to do it already, with chatbots perhaps even less sophisticated than the original ELIZA bot, created in 1966.

“ELIZA shows, if nothing else, how easy it is to create and maintain the illusion of understanding,” by the machine, noted MIT researcher Joseph Weizenbaum in his 1966 Computational Linguistics paper on Eliza. He continues: “A certain danger lurks there.”

Where our connections take place through screens and text, much of our connections are already imagined. It’s not how we program bots that define the next stage of this reality. It’s how we reprogram ourselves, and our expectations of what “human” connection looks like.

Photo: Ashley Madison “fembots” given a physical form by Swiss artists !Mediengruppe Bitnik.