Digital practitioner interview: Tung-Hui Hu on the Last Time You Cried project

by Korey Jackson on August 19, 2010

The dcb blog in the coming months will feature interviews with several practitioners of digital projects on the University of Michigan campus.  This mini-series is intended to showcase the work of digital practitioners across our campus and connect their research endeavors to the publications of digitalculturebooks series.  Our first interview is with Tung-Hui Hu, Assistant Professor of English and one of the minds behind the Last Time You Cried Project.  The project’s website is

digitalculturebooks: Tell us a bit about the Last Time You Cried project, and what the exhibit contributed to it.

The Last Time You Cried was an art project that translated human emotions into computer-generated poems.  For two months, architect Vivian Lee and I asked volunteers to call in and respond anonymously to the prompt: When was the last time you cried?  We then fed the voicemail messages we’d collected through voice-recognition software, knowing that the act of crying or tearing up while speaking would cause the software to mishear words or leave gaps in the sentences.  We wrote a program to fill in the gaps with phrases borrowed from Google and other databases.  Though computer-generated, the resulting poems were affectively charged in their own right: [transcript of “Kalamazoo, MI 5:48 PM,” read by poet Samiya Bashir]


I don’t have any control over what I’m doing and–
and that what is what I’m doing.
If not, what I really mean,
because I feel trapped in my own body …


The exhibition occurred in the hallway of the University of Michigan’s Art/Architecture Building, where passers-by could peer into a listening booth to hear the poems read.


Installation in place in the Art/Architecture Building hallway

We hoped that the listening booth would bring strangers into a shared space, and we often saw several groups listening at the same time.  We’d designed the listening booth using digital waveforms from the participants’ recordings, and the resulting cave-like structure was another way we hoped to turn digital data into sensory experiences.

How does the work you do change when doing it digitally?  Beyond working in a new medium, are there fundamental differences?

There’s a long history of art that draws on participant responses, such as Hans Haacke’s 1970 “MoMA Poll,” which asked visitors to the Museum of Modern Art’s “Information” show to vote on who they might vote for in the upcoming election.  “MoMA Poll” operated in a very simple way: two clear Plexiglas boxes placed side-by-side, into which you cast your folded ballot.  The voting technology (a Plexiglas box) was perhaps the least interesting part of his project, which understood social structures and situations as a medium to be investigated.

Although scholars typically associate media art with cutting-edge technology–Flash/HTML5 art, iPhone apps, etc.–technology is often incidental to the more difficult questions at stake.  Those questions, for us, had to do with the changing nature of intimacy (what stories the participants chose to share, and how close one could stand next to a stranger and listen), and with the ways in which we reuse and appropriate language, an idea we probed by using found words and phrases in our computer-generated poems.

What digital literacies, if any, do users need to have before engaging with the project?  Have there been “digital divides” evident that preclude users from getting the full benefit of the project?

We set up a phone line for volunteers to record their messages, so having access to a phone was the only requirement.  This made participating a spur-of-the-moment decision: one respondent called moments after crying in a bathroom stall, when she saw one of our posters on the bathroom door.  Of course, in order to hear the poems that resulted, participants had to physically experience the installation, a challenge to anyone with mobility or time restrictions.  A few people suggested that we create a searchable database of poems online, which would allow web users to participate.

Has the project changed since its inception? If so, in what ways?

Originally, we wanted to install our project in two spaces, but the massive amount of work involved in building even one structure led us to quickly change our minds (see the photo below, showing each foam piece individually numbered and sorted for assembly.)



The other change we made was to have volunteers read out the computer-generated poems.  We originally had a computer-synthesized voice, but test subjects thought it was distracting.  (It sounded a little bit of ELIZA, the computer-therapist program developed in the 1960s.)  As a result, we learned how much a poem is conveyed through simple gestures of enunciation, whether a pause, emphasis, or varying the rhythm of the voice.

What have the users of this project taught you?

The computer scientists we worked with were initially shocked by how poorly the voice-recognition software understood the messages.  The text was degraded to the point that perhaps 60-80% of the words were garbled or unrecognizable, and the computer scientists indicated that it would be impossible to produce any form of meaning out of such low-quality data; such sentences just wouldn’t parse.  But speech is very rich; each of our sentences contains multiple ways of conveying meaning.  To our surprise, we discovered that our computer-generated poems often arrived at a very similar narrative–for instance, of pregnancy, or of anxiety over one’s body image–as the original messages, even with only 4 out of 5 words “missing.”  Even with most of the details altered, a few words were enough to hint at a rich emotional story.

Are there practices that you would change if you mounted this exhibit again?  Are there plans to mount an exhibit of the project again?

I’d like to refine the software we used to author the poems.  We had a very primitive implementation of natural language processing; I’d love it if the program could also understand the “prosody” of the message, that is, the way that a speaker stresses or accents certain words.  (One example is iambic pentameter, the meter used in Shakespeare.)  Prosody is a fundamental part of poetry, but it’s viewed as an increasingly arcane art: my students sometimes think of their free-verse poems as completely free of rules, as purely textual creations on the page.  One of the results of our project was to show how much the voice of poetry continues to matter.

At this time, we’re regrouping, but I hope to launch the exhibition elsewhere.  Our project was fairly rooted to the University of Michigan: we designed the installation specifically for our hallway site, and the poems that resulted mirrored the stories of the community.  We had a fair share of stories about classroom anxieties.  It’d be interesting to see what our project reveals about a different community.

Photos by Tung-Hui Hu

Leave a Comment


Previous post:

Next post: