I am going to be very busy over the next few months and will have trouble responding to correspondence without a major time delay. Please talk with my chatbot if you'd like to catch up. I've programmed her to speak in a manner very similar to my own. You will see the questions she asks are much like the kind of things I might ask you over drinks. And her responses to your questions are in what MyCyberTwin calls a "warm intellectual" style of conversation engagement.
If you are also busy you might consider setting up your own chatbot and they can speak to each other in lieu of an actual conversation between us. So you know in advance, these conversations will be recorded.
The transcripts of bot-mediated chats I've collected since posting this note explore the boundary between broadcast and confessional styles of online communication. A chatbot has the potential to interact less like a third wheel than an obstacle designed to accelerate intimacy. Naturally this requires a good script and the capacity of the participant to ignore the staging of the conversation. With the willingness of both participants, the outcome is no less valid a conversation format than any other asynchronous communication. (TLDR version at the bottom of the page.)
MyCyberTwin is Australian startup that never quite delivered on its goal to create an Internet full of chatterbot cloned identities. Anyone can set up a free bot and explore the ruins of its now long forgotten social network, which thrives as the company profits as a customer service tool for corporations. In this accelerated age, it too soon to call something just shy of a half-decade old retrofuturistic? Reading the company's mission statement, one imagines a dystopian cyber world of chatbots holding conversations with each other in lieu of actual persons too busy, too lazy, or to indifferent of one another to bother allocating time to talk.
In spite of this, MyCyberTwin is a well written program. You start with a Myers-Briggs kind of test to determine the right conversation style. Then write questions and comments of your own to keep a conversation going. It can easily take over an hour to customize all the possible questions.
Accuracy isn’t the only objective while writing scripts for the bot to follow. Duration of amusement is just as important. I don’t want to bore friends to “brb” after too many canned-sounding answers. So I focused more on creating questions for participants rather than answers and statements of mine.
Composing responses, I was inspired by interactive fiction writing like Zork, which deals with the intuitive nature of participants. A good IF writer knows how to tell a story so a user’s command like "get lamp” is signaled somewhere in the text, likewise, the reader is engaged enough to want to continue.
But no matter how much I write, I never have full control over the bot’s responses and questions. Which means self-promotion sometimes slips in — “Do you have your own CyberTwin? If not you should go to mycybertwin.com and register. You'll have a great time re-creating your own personality online“ — or references to Australian brands or other corny pre-written statements. Yet given the participant’s understanding this isn’t really me, I expected no feelings of embarrassment. I can’t be held accountable for a roll of the dice.
Some of the pre-written text is pretentiously naif . My “warm-hearted intellectual” bot quotes Simone de Beauvoir ("One is not born a woman, one becomes one") and says things like “'All' is very finite. You don;t want to reconsider?” (Complete with typos!) In the tests I ran, it seems to speak in equal parts my writting and MyCyberTwin copy.
That Myers-Briggs kind of test had a dual purpose as this social network, like most bad ones, positioned itself as a dating service. What a preposterously awful idea ("Hey, let my chatbot flirt with your chatbot for awhile and then in IRL we can take things from there" ?) Anyway, this is seductive persona of a “warm-hearted intellectual”:
You would like a meaningful relationship, a partner who can share life's joys with you and join you on the journey. You would rather not muck around with casual, shallow relationships.
You are intellectually open, and like experimenting with new ideas and situations. You are a strong thinker, and are not afraid to put effort into understanding things. You relish good conversation, and anything that expands your horizons. You probably love travel, reading, good conversation, quality experiences.
As a self directed person, you prefer to take your guidance from reason and understanding rather than slavish devotion to an external source of authority, like stuffy traditional values.
Yeah, pretty much. But to rein in questions heading in that direction, I entered the textual equivalent of Sartre’s bad faith weak hand hold as a response: "I don't want to be put on the spot but we can discuss this later." I knew there probably was some way to get the bot to bat her cyber lashes, but doubted anyone I’d talk to would crack that word or phrase.
First the bot offers visitors the choice of a private or public conversation: "Thanks for stopping by. If you'd like to keep the conversation off the record just type OTR." I wanted to give participants the option to keep responses totally private. But even public, there is the option of anonymity. Participants chose to enter real names or screennames. Some of my friends picked screennames I recognize from various online identities, some used aliases that kept me guessing. With a few of these conversations, I have no idea who the participant actually is.
I deliberately chose a vague photo of myself, a screengrab of me on a webcam. I'm a vague apparition beamed from the ocean of Solaris, not a high-res glossy plastic thing here to leverage my personal brand strategy. I have a feeling we trust grainy images of people over crisp ones. That some remoteness makes one feel closer....more casual, less professional.
Most of the questions asked were about how people feel about technology:
- Is there anything about you online that embarrasses you now?
- How much time looking at another person's profile online is unhealthy/too much?
- Do you think it's possible to trust someone you know only online?
- What is your first memory of the internet?
- Have you ever cried while looking at things online?
- Is there something inherently vulgar about social networks?
Lets step back for a moment and consider ways we typically communicate electronically. It is almost always reading and writing. So much can be said in a glance, tone of voice can reveal everything, but generally we stick to text-based communication. Only with comfortable familiarity with a person will I start sending pictures in emails, video chat, or communicate with all the multimedia opportunities that exist in the present age. Going lowercase in an email, removing the “Hi” or “Dear” salutation, or sending a video chat invitation seems like the digital form of the informal “you” in languages with T/V distinction.
Text can create some barriers to intimacy, part of this is a mutual understanding that there is always the possibility of misinterpretation. The burden is typically on the recipient not to take things personally rather than on the sender for writing things that could be taken personally. What’s recognized is words provide many interpretations but everything is okay so long as the general point comes across.
And then there are the inevitable typos and other errors that come up. Ever write “can” instead of “can’t” thus expressing the opposite of what was meant? We are increasingly used to making mistakes in correspondence and sending without realizing.(e.g. “Damn you, Autocorrect!”) So a bot that introduces randomness — an inappropriate response — simply follows in the tradition of so many misunderstanding in text-based communication.
Email is not dying or dead, but it presents advantages and disadvantages like any other kind of communication. It takes discipline and mind free of anxiety to avoid feeling like Skinner lab rat clicking on your inbox over and over, expecting some missive from a work prospect or romantic interest.
Some obsessively clicking users never even expect a reward that great. They click-click-click for any kind of message at all. Email is something you intentionally check, it wasn’t designed to arrive directly to you like a phone call. The way we think of time online has fundamentally shifted due to smart phone market concentration. What is asynchronous now is, as Douglas Rushkoff writes in Program of Be Programmed, is real life. When we check our phones in the middle of a conversation, we are putting real life on hold. But the benefit of email over other text based communication is that you can check it whenever you want.
Another problem with email is coming up with what what to write when the content is not a request or an answer. Some of the worst writer’s block I experience happens when I tried to account for several months of my life to a friend I care a lot about but isn’t a daily, even monthly presence in my life.
With new friendships and acquaintances a different problem presents itself. A problem I call “conversational skeuomorphism,” the redundant nature of small talk when social media already offers up all the answers. Where do you live? What do you do for a living? Do you have a boyfriend? What sort of things do you write about? Small talk is losing its hold as a fundamental first step toward deeper conversations. Is it a bad thing? It’s hard to say. Either way this is the way we live now and no one is scaling back. We need to find alternatives to pre-digital age rote conversation tactics (my suggested alternative: picking cards out of a deck of Oblique Strategies.)
The experiment with My Cyber Twin is an attempt to build an intimacy creating obstacle as an alternative to email. It addresses two common problems with email — the anxiety of waiting for a response and the difficulty in coming up with non-obvious things to write. Ideally, I’ll figure out a language so benign that my bot will fade from the focus of the conversation. But that isn’t unique to the use of a chatbot. Many tweets on Twitter are about Twitter itself. Media tends toward self-referential conversations.
However, my fear as I continue with this is that it may be too cute, too staged, too much of an effort for anyone to bother. I had some difficulty recruiting volunteers without personal letters or seeing people face-to-face with the opportunity to explain my motivation for doing this. It was less about the wtf than simply why?
Remember I don’t have full control over the speech of the bot. I don’t have any access to see what was written by the MyCyberTwin team. What I can do is try to curtail any possible embarrassment is give answers to as many words and phrases of my own. For example I now have a rule that if someone says “facebook” it will say “I don't really use Facebook. Do you like Facebook?” (or two other vague statements about Facebook.)
The unprompted question "Am I your number one friend?" coaxed an interesting reaction from a friend. I did not write that question. Nevertheless it lead to what might be the most honest thing anyone has said to me in my life:
joanne: Am I your number one friend?
Gil-Martin: The side of myself which is cerebral is at odds with my happiness a good amount of the time, and it is this side which you engage with by accident or design. So to answer your question, no not even close.
joanne: Oh no. Is there anything I can do?
Gil-Martin: Does a dog try to be number one? People love dog's friendship. Be number Joanne.
There were lots of unexpected answers to questions:
joanne: Have you ever felt afraid for your life while on the Internet?
joanne: Is that so?
annie: i've had stalkers. and some of the countries i write about are a bit dicey
joanne: How does that make you feel?
annie: but for the most part, no
joanne: Oh yeah? I didn't realise.
joanne: How many times a day do you check social network streams?
annie: there are many countries where information online leads to an invitation to a beheading
joanne: What kind of information are you after annie?
annie: too much
joanne: Perhaps you're right.
Some responses were amusing:
joanne: Do you focus long enough on things to extract happiness from them?
marbles: Once I focused on a cucumber so long, it melted into a puddle of pure happiness.
Some with great insight about the world we now live in:
joanne:How much time looking at another person's profile online is unhealthy/too much?
Brendan Byrne:Depends on what your purpose is. Research, no amount of time is unhealthy. Sexually, probably more than ten seconds.
Brendan Byrne:No. I mean, about the amount of time to realize what you're doing and then stop.
joanne:Is that really what you mean Brendan Byrne?
Brendan Byrne:Yes. It's sort of analaogous to staring at a preteen and then realizing you're doing and then looking away.
joanne: Only kind of?
Brendan Byrne:It's not a perfect analogy.
An early hypothesis — that I wouldn’t feel embarrassed by what the bot said— was quickly disproved. I couldn’t keep all the flirty language under wraps so in the middle of a conversation about technology and society inappropriate MyCyberTwin Mechanical Turk copy questions would throw things off:
joanne: Do you get told that you are attractive often?
s.: that's an odd question.
joanne: I'm trying my best. Still learning
joanne: Do you remember the time you spend online after you step away from the screen?
Far worse was this interaction with someone I've known a while online but haven't met in real life. I was horrified by this transcript. While it seems clear he knew I didn't write this text, I did orchestrate the thing so it is my fault things went so disastrously. At first I thought of the bot as something just slightly more evolved than an answering machine. But that horrible conversation made it seem like something fallible, something almost human. A bad human that is. A slutty, racist idiot of a humanoid. But I can change her.
A successful chat isn’t so much about suspension of belief, but willingness to go along with the experiment. Conversations that started out skeptical or theatrical ended up looking disjointed. But conversations that look most authentic are with people I know and communicate with frequently. “ I think, if we did gchat day-to-day, it might be like. What do you think?” said one friend of mine. And maybe it would. I find when I chat, I often miss questions and lines because I might be distracted by other incoming messages or other things I’m looking at online. Here most participants were hyper-focused on the chat, in ways they might not be over instant messages with a friend. So errors that aren’t too different from a human on the other end, confirmed one’s suspecion about the bot’s incapacity to follow along in the discussion.
I created a fake account for James Franco as that is the one person others might actually enjoy communicating with in this way. I took the Myers-Briggs test and then gave him a “cheeky intellectual” style of responding. Then I answered some of the questions with quotes from interviews and what I think a guy who calls Claire Denis’ Beau Travail his favorite movie would say. But ...it didn’t work. There’s just too much a participant could possible ask him about (Kalup Lindzy, General Hospital Freaks and Geeks, RISD, Hart Crane, etc etc, hundreds of movies...) If anyone would like to take over this account and try to improve on it, you are welcome to it.
Reviews of social networks and communication services are never helpful without a sense of the critic’s preferences. What kind of person I am indicates how I interact with and through technology. What works for me won’t also work for you. I don't mind Twitter, but Facebook always feels like work to me. If I could, I'd only communicate through Gmail, both email and chat, since I've got folder set up in specific ways. Never too thrilled to text or call anyone.
At the time of this experiment I was resting after a two week period of an overwhelming number of social obligations, during which I was meeting friends for lunch, coffee, drinks nearly every night, in addition to conferences and other events that were taking place concurrently. I found myself talking faster, actively more direct, less cagey with people. Not a bad thing, I guess. But this is outside my comfort level. Blustery confidence is not a face I can sustain without feeling tired after a while.
The reason I bring this up is because the female AI persona is usually a sexual aggressor or secretary, nothing like the sort of wallflower I am normally. But unlike real life sexual aggressors and secretaries, I am often called — as an insult or in jest —robotic, which adds another interesting layer to things.
It might also be helpful here to bring up some of the inspiration for this experiment. Attending Kio Stark's magnificent authenticity class at ITP as a guest critic. A conversation with my friend Sarah explaining how people get around avoiding technology on Sabbath (automated timers to switch lights off and on.) My friend Erin emailing me about the movie Catfish and how “the girl in the photos kind of looks like you! OMG! HAS THIS WHOLE THING BEEN A LIE, JOANNE?!?!?!?” Watching The Pillow Book and appreciating the performances as a nod to butoh. Another acquaintance of mine pointing out Tino Seghal's This Progress operated like a computer program. That comment inspired me to enter "What is progress?" among the other questions for my Cyber Twin to ask participants. (The only other question that I nabbed from someone else — "What are your unrealised projects?" —is Hans Ulrich Obrist’s most famous.) Reading about HB Gary and the sockpuppet army creating "persona management" software on the way. Also thinking (and blogging) about Kristin Lucas and Andrew Kortina’s “Identity Swap” conceived at Rhizome’s Seven on Seven event and how badly I want to get into hijinks with my own uh, personal brand.
Meanwhile, I started playing with VYou just before the holidays. Although it is me in the video screen, my identity seems as canned on Vyou as it is on MyCyberTwin. I stutter a little when I talk and have jerky hand movements among other nervous tics, not too dissimilar to the way I tend to act around new people. I’m not like this in front of a mirror, but I can’t feel comfortable talking to myself as a recorded self-portrait. Maybe it’s fear of the anticipated reactions of others, a delayed anxiety. This isn’t me but a robotic version. The identity in the age of digital reproduction:
There is a grey area between public and private communication and that is the unlikelihood that anyone is noticing at all if you “liked” or commented on something. And if someone notices, do they remember it later? Only a handful of people chose to take the chat off the record, which surprised me. But given option of anonymity and the expectation no one will notice, it’s really not a big deal if conversations are shared.
No, I don't expect anyone is going to tell me their deepest darkest secrets on a site called MyCyberTwin. No way. But these questions could spark interesting comments and stories that wouldn’t otherwise be shared. This is a roundabout way of finding commonalities.
I have in the past dismissively called VYou “stalkerware,” as it seems like a trap for people obsessed with you to spend too much time. But “stalker” really isn’t a fair colloquialism in this age. If I like someone as a human being, I’m inclined to want to know what they like and what they do. And so I’ll click around to check out the information on them out there. It is perfectly fine so long as you never create a narrative in the gaps. It’s tempting to jump to conclusions but social media only reveals so much about a person. As Sam Shepard so wonderfully once put it, “The things that I wonder about most are not on the internet, I promise you that.”
A more generous way to explain this kind of communication is "intimacy creating." VYou has more in common with Formspring than Quora as the questions are directed to a single person, giving one a focused amount of attention. I chose an interview format for MyCyberTwin because that puts the focus on the participant over the bot. It’s not just for issues of verisimilitude, but to make the process enjoyable for the participant rather than a tedius exercise in the bot creator’s narcissism.
I like to joke that the perfect Facebook app would crawl through birthday listings and automatically leave "Happy Birthday" comments for others on the appropriate day — without you ever needing to know about it.
But MyCyberTwin is not like that at all. It’s not that a conversation happens without me ever knowing about it. The point is, after someone chats with the bot the transcript is sent to me. It is gameifying email. It also means that a person who prefers chat and a person who prefers email might arrive together while communicating via their preferred methods.
It should be clear by now just how playful this experiment was from start to finish. I’ve been laughing about it all week. It started as a joke but greater truths emerged from there. Chiefly that there is nothing wrong with experimenting with new forms of communication online as each has its own advantages and disadvantages.
We make decisions about what form our message will take. Text message, phone call, email, Facebook message, Twitter direct message, Skype or instant message; each method carries with it a set of habits and expectations. I would find it extremely weird if someone called me the day after meeting me, even if I gave the person a card with my phone number on it. That's too direct. I’m much more annoyed when someone sends me a Facebook message, but I’ll deal with it since the culture at large disagrees with me and seems to find it a valid way of staying in touch (even given the risk of improper delivery.) We don’t all like the same things. Media is no different from film and movies in that regard. Your preferred method of getting in touch might not be mine.
I want to live in a world where people are really listening to one another and feelings are never truncated by media. To get to that place we must continue to develop and experiment. I'm not at all suggesting anyone sign up for a chatbot to keep in touch with old friends, but web developers you could learn a few lessons from it.
This article isn’t about how strange it is to play with a chatbot. It is about a conversation between two people taking two different forms: instant message chat on one end, email-like asynchronous message for the other. It’s obviously an imperfect way to converse, but should offer ideas and inspiration for those conceiving of alternatives to current digital communication tools.
My Cyber Twin chatlogs: http://joannetwin.tumblr.com/
Also, I'll be answering any questions you might have about the experiment on VYou this week: http://vyou.com/joanne
The Tomorrow Museum
Feb. 24, 2011