Your Eyes or My Words
Feb 7, 2014

Notes from my talk for the Lift Conference, 2014. Part of a panel organized by Nicholas Nova on Algorithmic Culture. Here is a storify of some tweets. (Video of the talk.)


These are images from a video piece by Philippe Parreno, a recording of Pierre Jaquet-Droz’s clockwork automaton from the 18th century at work. It was built to write love letters with the quill and its head and eyes follow along as it scribbles on paper. I think about The Writer often when I come across bots in their natural digital habitat. Or language like we see on products on Zazzle. It can be hard to tell if those are the words of a machine.

I saw these cards in Key West several months ago. And I was puzzled because, why on earth — of all the things DH Lawrence and Edith Wharton have said — why would anyone quote them on this? I want to believe this was a next level joke. I am really confused by these cards. When did DH Lawrence say “Blow out your candles…happy birthday?” Is this a quote from a character in a novel? Did he said this, presumably, with someone with a notebook, scribbling down these profound words?

“We ought to be opening a bottle of wine!” Is this a rare gem of a statement? Has no one else ever said this before? I’m failing to see the wit or wisdom or uniqueness to make it attributable. I wonder if it was an algorithm that searches famous books for keywords then adds the text to greeting cards.

Facebook makes me feel like a bot. Especially celebrating birthdays. The kind of autopilot small talk that makes up a significant amount of social interactions feels even more automated when we have these conversations on a screen. You don’t even need to remember someone’s birthday —just follow the prompt. The reminders are centered in the interface. The convenience of outsourced memory adds another layer of obligation to what should be a warm friendly exchange. And how creative can you get with these comments? “In the words of DH Lawrence, ‘Happy birthday’”

Facebook has a power and interest in engineering relationships that belies these seemingly inconsequential interactions. Every "Happy birthday" comment is another engagement with another user that the website uses to determine the nature of your relationship with that person. A recent paper written by a Cornell researcher and Facebook engineer explains the site can very accurately guess whether two people are in a relationship, not by number or cluster of mutual contacts, but dispersion of friends. If you know someone’s coworkers, friends from middle school, and people in her yoga class, it appears likely you are in a relationship. If you share your life with someone, you see that person in many contexts. The same researchers predicted relationships in trouble because the couples lacked that dispersion.

In the book, The Facebook Effect, it was said the website could predict with 33% accuracy whether someone will be in a relationship or not next week. Things like attending events together and adding mutual friends reveal two people are becoming close. That was a few years ago, so the accuracy must be even greater now. By tracking this information, and tailoring this site so that people you take notice of become more prominently displayed in your newsfeed, the website works to encourage relationships to form.

OKCupid is open about its purpose as a matchmaking service. Facebook is doing all of that and more without the legibility of quizzes or express purpose of drawing your attention to potential love interests. But sometimes we can see the algos at work through the cracks.

I was thinking about this while watching the film Sliding Doors. Remember that one? It is split in two parts. In one half, Gwyneth Paltrow's character catches the train, in the other she misses it, and that leads to consequences of a relationship ending, a visit to the hospital, and so on. I want to see this film remade but instead of missing the train, Facebook algorithms filter a critical post from Gwyneth Paltrow's newsfeed. A Facebook "like" could be life or death!

In December 2010, a writer for the website Shiny Shiny noticed the friends in the sidebar of her Facebook profile appeared not random but possibly the "people who click on my profile a lot." Since then, that post has received 1466 comments. Readers, mostly young women, began experimenting by creating fake profiles and looking at their real profiles to see if the avatars of the fakes would show up in their sidebar. They discovered other places revealed friends that appeared to be based on engagement, like the chatbar when you are logged out of chat and something they called the "Mouse Over 5," shorted to MO5 — not a British algo intelligence agency — but moving your cursor over your own image would reveal the images of 5 friends. They guessed that the MO5 had to be people who clicked your profile the most. What was it that fueled their curiosity to see how the Facebook algorithms — at the time called Edgerank — worked? Most of them just wanted to know if their crushes were paying attention to them or not.

They comforted each other through the hard time of not looking at profiles at all, so that the algorithms revealed one-sided rather than mutual engagement in these avatars on the screen. What you see here is the creation of an Old Wives' Tale about algorithms. It might sound silly, but Facebook is the creation of a college student that was from the beginning a way to stalk crushes and not much more.

I was curious myself and over a period of several months I stopped engaging on Facebook. Stopped clicking profiles, stopped clicking 'like,' did not comment on posts although sometimes I posted things of my own. Over that period of time I did notice the LSB10 (Left Side Bar 10) and MO5 revealed some patterns. One friend unexpectedly appeared in my MO5 so I dropped him an email. He replied to say he was just thinking of me — he was coming to town that week and would I like to grab a drink?

It could be card counting or it could be reading tea leaves. We do know that Facebook wants our engagement. It wants us to click obsessively. The algorithms are designed to deliver you the content relevant to you and that means relevant people. The website The Keesh was able to confirm some suspicions about "MO5" when it released this bookmarklet that revealed the affinity score Facebook algorithms assigned to each friend. Some of the high ranking people were heavy Facebook users that click on every profile nine times a day. Sometimes the information was surprising and made you wonder why that person spends so much time thinking of you. This is information that was never designed for a human to see.

Facebook can form an alarmingly complete picture of your life based on the information it tracks. With only public "likes" a software program could determine someone's sexual orientation with 88% accuracy and whether a person is a drug user with 75% accuracy. That's public information, not even accounting the profile clicks, private messaging, key words in private messages, and other information Facebook has tracked. In another recent story, Facebook Knew I Was Gay Before My Family Did, Buzzfeed interviewed someone who received a targeted ad "Coming out? Need help?" at a time he was about to do just that.

My friend Sara Watson wrote a piece for The Atlantic on a frightening ad from Facebook that appeared after her engagement. It was a picture of an engagement ring and a prompt from the website asking "how well do you know" her fiance. She hadn't shared information with the site about their engagement.

Facebook wants to disrupt compassion. They have a compassion hackday — Compassion Research Day, where they unpack concepts like "awe" and sort out ways to minimize conflict and mediate friction in relationships. One possible initiative to come out of that last meeting is a "sympathize" button for those times you want to passively engage but "like" is an inappropriate response. The website values a more complete self-expression on the website because it wants to mirror your life events, desires, and emotions.

I mentioned the subject of my talk with my friend Kate Losse who wrote the memoir The Boy Kings about her experience working at Facebook. She was Facebook employee 51 in 2005 and stayed on through its major adoption and growth. A point she made was how clumsy the algorithms often are. We are frightened when it correctly figures out who is your fiance, but that assumption is often wrong. An algorithm interprets a relationship as someone who clicks another profile often. But there can be other reasons for this. Maybe there is a mutual friend in a picture, maybe that profile is a shortcut to something else you want to see. It is assuming without asking the user his or her values.

I mentioned earlier how Facebook makes me feel like a bot. I also feel that way using "canned responses" over email —a keystroke shortcut response to email. I feel strange doing it even if it saves time. I guess I'm uniquely apprehensive about feeling inauthentic, because now there is an app Romantimatic that lets you automate "I Love You" text messages to loved ones.

Maybe that is the natural conclusion to Facebook. I encourage Facebook to give us the option to automatically leave a Facebook “Happy birthday” comment whenever it is someone’s birthday. So we neither need to remember someone’s birthday nor remember to leave a comment. That way we always know that users are honored appropriately on their special day.

#

Following the talk Matthew Plummer-Fernandez noted Google has recently patented auto-generating responses to things like congrats and birthday messages. Eduardo Cuducos directed me to this birthday message/thank you Facebook app. And I also heard from Julien Deswaef who has a project called Love Machine that automatically clicks all the likes available in your timeline