Happy spring, folks. I'll admit it's been difficult to muster words in the last few months. Last year was full of them for me, and the well was well near dry come January. Add to that the escalating depravity of American and Israeli attacks on a widening number of people, and I've been left at a loss. Silence is at best cowardly and at worst a kind of violence in itself, so here's a small addition to an all too quiet chorus: threatening the extermination of a "whole civilization" is genocidal intent, normalized by the lack of action against a still-ongoing genocide in Gaza (ceasefire, another word that has lost its meaning) and the deliberate starvation of Cuba. (And, talk is cheap.)

Under this pall, though, things still emerge. The Homes for Living paperback came out in February on the heels of its inclusion on The New Yorker's Best Books of 2025 list. For Electric Literature, I put together my own little list, of books about retaking and rebuilding our commonwealth, as ballast for myself (and hopefully others) amid these rough waters. I've been chatting publicly with a number of folks on housing matters: with City Limits on a recent report on the state of Mitchell-Lama housing in New York; with Open New York, a YIMBY group with a refreshingly nuanced platform, on the importance of social housing and a long-term view of supply interventions.

 

In these settings, I'm more often than not the one being asked questions—which is why I so enjoyed getting to speak with my friend Sharon about her excellent new book The Last House on the Block and why I'd like to ask you a few questions about this newsletter.

 

I'm a question asker by nature and training. Posing questions is how we probe the world, and different approaches to how to ask a good one is a through line across my career: training as an oral historian, working as a journalist, undertaking academic research, guiding government innovation processes. This recurrence is, I suspect, not surprising to you, but when I begrudgingly dipped my toes into our present-day AI LLM fever dream for the first time, I was struck that much of the available guidance for how best to use these tools focused on "prompt engineering," or, in less overwrought terms, how to ask a good question. This emphasis had me ruminating: what was a good question in that setting? What was such a question for, and what could the relationship between the resulting answer, "the facts," the asker, and the answerer really tell us?

With the notable exception of the researchers plumbing the "mind" of LLMs and people in intimate relationships with AI companions, most LLM users' questions or prompts are almost exclusively extractive. They hope to ask a good question in order to pull information from their respective interrogees, and the fact that the resultant answer may be factually incorrect is evidence of a flaw—in the honesty or reliability of the entity being asked, or in the question itself.

 

Oral history has a rather different conception of the purpose of questions, and a rather different context. Despite a question's usefulness in tracking facts down, oral history is not solely concerned with accuracy. The exchange between interviewer and interviewee is thought of as a collaboration in which trust and rapport must be built to co-create something of value. Where a “wrong” answer calls into question the whole purpose of an LLM, discrepancies between the historical record and memories shared in an oral history interview creates tension filled with meaning. As one dean of the field, Alessandro Portelli, wrote: “oral sources tell us not just what people did, but what they wanted to do, what they believed they were doing, and what they now think they did.” From certain errors of memory, we can learn certain facts of experience.

Setting aside the meaningful consideration of an LLM intentionally lying, is there meaning in LLM mistakes beyond identifying a need for a model's improvement? I say: not really. Behind the facade of personhood, there's just a jumble of probabilistic composition. These aren't generative mistakes; they're just an output of what we likely want to hear in response to our prompts bent on extraction. Our understanding of the world, of experience, and of each other is no greater from understanding the root of an inaccuracy with an LLM. (We have plenty of other ways of identifying our biases, for one). With people, it often is, and that has much to do with the context of the question and the relationship between asker and answerer.

 

I am personally not much concerned with questions of AI personhood, but I think there's something here that informs that discussion. We develop our individual identities—we become ourselves—in relationships through an ever evolving conversation of questions and answers, call and response. The meaning behind our faults—or ideologies, beliefs, harms—is fundamental to our personhood. If there's nothing more to something being wrong than an attempted prediction, then there's no person there. Could be wrong; ask me about it in a few years.

 

I am, however, concerned with how the prevalence of asking questions of these bots might shift the very idea of what questions are for and how they should be asked. Are questions tools for extracting what we want from another, or paths toward understanding, even when the facts are off? Lose the latter perspective, and we lose a piece of our humanity; being wrong is, after all, a meaningful part of being human.