Research

After the Book

The common worry about AI and reading is that machines will turn readers into non-readers. I think that gets the history wrong.

Literacy never transformed how most people think. Deep reading, the slow and sustained kind that builds a mind, has always been practiced by a small group. Most people in literate societies have stayed mostly oral in how they learn and talk and decide. What conversational AI does is offer a more complete and more seductive oral environment than anything before it. That makes the case for becoming a deep reader harder to win.

After the Book takes that argument through ten linked essays. It starts with orality, moves through the bias literate cultures have toward literacy, and ends with what a way out might look like.

It isn’t being made as a conventional book. It’s planned as a multimodal reading app, portrait-first, with video, audio, animation, and interactive figures set next to the prose. Orality is embodied. It lives in faces, gestures, and voices, so a book about it shouldn’t be flat text alone.

The research draws on the orality and literacy tradition (Walter Ong, Marshall McLuhan, Eric Havelock, Jack Goody, Harold Innis), on Plato’s Phaedrus and Derrida’s reading of it, on second-order cybernetics and Gregory Bateson, and on the cognitive neuroscience of reading (Maryanne Wolf, Stanislas Dehaene).

I publish working notes at postliterate.org.

AI interfaces and the unrepresentable

The book grew out of a research statement I wrote in 2024. Its argument: chat interfaces return us to something like Walter Ong’s “secondary orality.” But instead of broadcast media talking to millions at once, there are now millions of separate one-to-one conversations, each adapting to the person on the other end and drawing on a huge body of writing.

The statement also looks at what can’t be captured in training data: tacit and physical knowledge, emotional intelligence, cultural context, and the significance of things that didn’t happen. Systems of notation have always had blind spots. Western musical notation couldn’t write down the blue note, and for years blues was denied serious analysis because of it. I want to know what else falls outside the record, and what it means that AI systems are built only from what’s inside it.

Journalism, systems, and democracy

With Heather Chaplin, I’ve spent years mapping why local news is collapsing and what that does to democracy.

• Crisis of the Free Press: A System Map, from the Wicked Problem Task Force, 2016–24.
• Journalism Ecosystem Model, an agent-based simulation of how news deserts form, built with Francis Tseng in 2020.
• AshLab, a digital archive of John Ashbery’s house, 2012–14.