Conversation Design Interface

You open a book of poems, and you read, in the middle of page number three, a long sentence that has been arranged into four lines. After those four lines, five words make a declarative sentence, just one line long. Long sentence, short sentence—an interrogative, an optative—the poem continues to make a pattern through this material, sentences of varying length organized into lines that rhythmically emphasize different parts of speech. You read the poem aloud or, more often, a voice almost like your own recites the words in your mind, almost like being silent. What you hear in your head—and sometimes, what you can’t quite make out—is your performance of the poem: your internal recitation directed by the poem’s unimpeachable arrangement, how it asks to be read. You open a book of poems, and on page three, you notice the first poem of the volume is called “Interior Life.” After thirteen sentences, the poem concedes to itself: “I hear my own voice from a distance: / I don’t want to say the words of what happened.”

To say the words of what happened is to dramatize the event through the process of saying—of what it takes to speak after hearing one’s own voice from a distance; elsewhere, another person performs that dramatization. Ryann Stevenson’s Human Resources, selected by Henri Cole in 2021 for Milkweed Editions’ annual Max Ritvo Poetry Prize, is a book that imagines this admission in different contexts, the event of hearing one’s own voice refuse to elaborate—I don’t want to say the words of what happened. Stevenson’s poems make this admission a new kind of discovery each time it occurs.

If there is a single persona that conveys these poems to us—often called a “speaker,” a twentieth century invention that never seems quite right to me—she is a Silicon Valley employee who specializes in the construction of Artificial Intelligence, a voice that sounds increasingly less human, more like itself. Stevenson herself happens to have worked for Silicon startups designing voices, remarking in an NPR interview that this work prompted her extended meditation on “voice and disembodied speakers calling to an unknown user” and that it “had a direct correlation to poetry, the speaker of a poem, and the readers.” Marianne Moore worked as an editor; T.S. Eliot was a banker and editor; Wallace Stevens, an insurance executive. The work of poetry is often historically distinguished from the kind of work that practically sustains our lives—the work that keeps rents paid, food available, insurance possible—though in Stevenson’s case, the metaphor is, for obvious reasons, unprecedented. “First order of business is to define what intelligence is,” she anticipates in “Work from Home,” “then how to avoid a dystopian eventuality.” So much depends on the impossibility of these tasks strung through the jargon of “dystopian eventuality,” the sterile possibilities that suggest one outcome can only be understood in the context of other outcomes.

So much depends on the fact that the possibility for this eventuality—or the anticipation and eventual resistance to it—takes place in the poem at home, in a private space away from the conference rooms of Silicon Valley, dominated by men. The responsibilities derived from particular places become less categorical, much like the difference between human and artificial intelligence for the person who needs to make sure one seems almost like the other. We find the inverse experience in “Anticipatory Design,” a poem that attends to a sample sale where “the exposed / brick decals will be peeled / off the walls” and to trash this fleeting cover is reminiscent of a “castle / in a children’s pop-up / book, the whole scene / will cease to exist.” The wall is almost like a wall with exposed brick, and this imitation is figuratively generative of a book that one might read at home with a child, or that one remembers reading as a child. Like the scene in the book, the material of a poem, its language, continues to point to our mortality, the fact that we are humans who can make voices—which means that, inevitably, we will not be.

What do we mean when talking about a poem’s voice? Because the word is often used to describe the manner in which a poem transfers itself to us, a way of approaching style, voice becomes a short-hand metaphor for a kind of process more specifically identified in the interplay between syntax, lineation, diction, rhythm: the limited material of language externalizes an event that happens internally—thinking, feeling, reacting—and that externalization creates a drama on the page:

                        When I’ve finished populating

the conversation design interface

with every possible word

a toddler could have for “favorite food,”

I retire to the sofa, sit in the spot

thick with golden light, and imagine

my face as an ancient bust,

warm and ignored in a niche

of a Roman apartment building.

Here, in the beginning of “Decision Tree,” it is not only the suggestion that our choices, even artistic decisions, can be anticipated and programmed, creating what I’ve called a dramatization of a particular experience; what makes that feeling happen in the poem, and what makes us feel like it happens each time we read the poem, is the way Stevenson organizes this sentence. By the time we reach the predicate, we’ve moved from future—a point when the product has been designed and children can discuss with the machine their favorite foods—back to the present—where the worker retires—and, finally, back to the past: the worker’s face not represented as sculptor but as sculpted and ignored.

After asking the bot programmed to know everything you love to then tell you what to love, who is left thinking about the person who designs its voice? In the volume’s title poem, the utility of all this technology after it’s been designed and distributed is called into question: “I spend all day trying to break a female / bot who wants to coach me / to be my best self.” Now the robot not only has the ability to teach its maker, but it can functionally express desire. The phrase “best self” becomes tautological, something a robot would say when they are trying to sound human, like the phrases “thoughts are things” or “trouble area”: “The Face Eraser, I read, yields best results when stored in the freezer, then rolled across trouble areas morning and night,” Stevenson writes in the prose poem “Trouble Areas,” and here is another example of semantically-branded technology with figurative possibility: “I read the copy until I remembered who I am.” Not the product but the language that describes the product helps its user come to a sort of aesthetic conclusion; poems do not have an instructional copy, though by way of a particular formal arrangement, poems ask to be read over and again in a specific manner. The poem says, You are here not only reading language arranged into lines, but engaging in the way syntactical tension transforms that language in time.

Here is Silicon Valley in the twenty-first century; here is America, and American poetry. For those who have studied American poetry—even briefly, leisurely—the word “resources” may be difficult to parse from the sound of “Dream Song 14,” from John Berryman overenunciating the word between its modifier “inner” and the subsequent renewal: “I conclude now I have no / inner resources, because I am heavy bored.” The important distinction between Inner and Human Resources perhaps suggests something about our changing relationship to interiority enacted in a poem. Here is the past without robot screens, and here is the future that we cannot but try to anticipate through them. It is memorable then, while anticipating, that the person who designs AI throughout Human Resources does not always look at her own screens but, more often, through other windows, with the “neighbor’s TV / flashing silently, / as if he were still awake.”