I don’t have you but you’ve still got my heart

Generative installation, custom code.

“I am now utterly convinced of the impossibility of neural nets ever producing coherent contextually-sensitive verse, yet as language play, and a window into the inherent biases within humans, it is impeccably intriguing and occasionally nourishing.”

-David “Jhave” Johnston

This work explores the intimacy and banality of artificially intelligent predictive language. The language in this work was produced through the generative process of repeatedly tapping a single predictive text option on the smartphone keyboards of 42 people’s smartphones. Your smartphone’s predictive text feature uses machine learning to learn your writing patterns over time, so that if you were to only tap a singular predictive text option–without typing anything in for the machine to predict from–a pseudo-coherent babble of machine/human language pours out to reveal a phantom-like impression of the phone’s user.

This video explains the origin of the text in the piece. This was looped on an iPad in the gallery.


On display in the SITE Columbus gallery at SAIC March-April 2018

How I asked people to send me predictive text “poems” 

Hello friends and family! Do you want to contribute to an art piece I’m currently working on? Now’s your chance! Doing everything involved should take about 10-15 minutes (including reading this), and all you need is a smartphone with predictive text capabilities (which is pretty much any smartphone nowadays).

THE PROJECT: I am creating multimedia installations using text generated by YOUR smartphone.

SOME BACKGROUND: A couple years ago, smartphones introduced “predictive text” features designed to help you text more easily by suggesting words to speed up your communication. The predictive word choices that appear above your keyboard as you type are what your phone thinks you’re mostly likely to say next. How does it know that? Using artificial intelligence, your phone has been training itself over the thousands of messages you’ve sent, or through notes you’ve typed out for yourself. The more you communicate on your phone, the more useful the words it suggests, and therefore the better it can model your speech. I like to think of this artificially intelligent tool as creating a phantom version of ourselves that lives inside of our phones, growing more and more like us every day.

I have found that by pressing a single predictive text key repeatedly instead of manually typing out words when composing a message, the language the phone generates is surprisingly coherent and kind of poetic, if a bit nonsensical. More importantly, the language it generates is unique to each phone, so in that way it is quite personal, and sometimes eerily revealing.

THE HELP I NEED: For the sake of research, as well as for content to be included in the project, I need as much “phantom poetry” as I can get! I am interested in seeing the similarities and differences across multiple bodies of generated text. Specifically, I am interested in what can be revealed of ourselves as well as of the tools we use on a day-to-day basis, and the more source material I have, the better I can understand this technology.

All YOU need to do is generate some text and send it to me!


For iPhone users:

  1. Open up a new note in in your Notes app, a NEW iMessage to nobody in particular, or a new FB message (again, to nobody in particular)

Note: Predictive text is enabled by default. If you have disabled it, re-enable it by going into Settings–General–Keyboards and then flipping on the “Predictive” option.

For Android users:

  1. Open up a new text message.
  2. For some Android users, your predictive words won’t appear until you’ve typed a single letter. Type one letter to make your predictive text suggestions appear. Then, delete that letter. Your predictive text suggestions should still be there.


I need you to generate three different bodies of text! For each predictive text button, generate text for 30-60 seconds! I encourage you to keep tapping to see where things go! It can be a fun rabbit hole to go down. Watch the above video to see how it’s done.

The three ways to generate text:

  1. Only press the middle predictive text button repeatedly for 30-60+ seconds
  2. Only press the left button repeatedly for 30-60+ seconds
  3. Only press the right button repeatedly for 30-60+ seconds

When you have generated your text, label each text block with its corresponding button (i.e., middle, left, right)), and send it to me!


  1. Email it to predictivetextproject@gmail.com
  2. Text message
  3. Facebook Messenger

Note: Sometimes pressing a single button repeatedly leads you into a repetitive, unchanging loop. That’s okay! Just keep tapping to see if you get out of the loop. If you don’t, that’s fine! I find I’m particularly interested in the loops the phantom finds itself stuck in.

Another note: After you send me your texts, I would love if you had any thoughts about what your text says about you, or any observations and interpretations of your text you might have. For example, when I generate text from my phone, the word “tomorrow” comes up frequently, which may allude to my tendency to procrastinate. No need to include anything here if you don’t want to!

Thanks to all who are interested in contributing to this project! Any contributors to the project will receive an acknowledgement in the final output!

As the project develops, I will post updates periodically!