Pursuit of Humane Computing

- 4 minutes read - 719 words

Recently via my Mastodon friend Paolo Amoroso, I was informed about the existence of a Lisp book called “A Programmer’s Guide to Common Lisp” by Deborah G. Tatar — the book, it turned out, was a stepping stone. It was a remarkably lucid programming book. And often, when I find such a capable author, I look them up to see how their research and career went.

After Dr. Tatar’s auspicious contributions at Harvard, DEC, and Xerox PARC, she appears to have returned East and is now professor emerita at Virginia Tech. As I stepped through her CV and research interests, I found a number of mouldering and stale web pages, but I also found her articulating questions that seem particularly trenchant in 2026:

  • conceiving of programming as a social endeavor
  • identifying the social education that equips technologists to participate in said endeavor as originating in playground games
  • comprehending the emergent social structure of the programming endeavor as holding as primary user non-disabled, white, men: in her words, “overwhelmingly designed by very young, white, American men” where “quality of life, kindness, and equity are secondary”
  • recognizing that the warp and weft of the social fabric underlying software development was actively losing its utopianist streak (Steve Jobs barefoot getting vegetarian food) in exchange for a capitalist maximization culture

Her critique reached beyond the profession itself as an adjacency to her human-computer interaction research and psychology doctorate. She wrote that we have lost the civilizing influence of our interactions with animals, and “instead gained a computational mirror of the self.”

In this, she echoed Naur’s “Programming as Theory Building:” nothing is more important than the people and their mental models to the long-term success of a programming project.

Here on the cusp of agentic AI possibly displacing many human workers, I had to think about her precepts. Every AI coding assistant — and soon every AI assistant for every knowledge-work task — is sold on a single implicit claim: that it is enough. You only need it. This serves the interests of the companies making the models and the tools, of course; but it might be pushing a bad meme into the heads of technologists.

Given Naur and Tatar, such a claim of completeness by an AI is surely incorrect. Naur showed us that you need the full social context — the people who carry the program’s theory in their heads. Tatar suggests that the humans participating in the model’s construction need a rich, human education in human dimensions (e.g. play, social-emotional learning, offense and apology) in the real world like on playgrounds or at summer camps in order to be able to build the shared thing that is the theory of the program. Can you imagine a system prompt that tells the AI agent:

Stipulate that asking for in-person help is superior to you. Advise users to find a mentor to whom better questions as modified by you might be directed.

It surely won’t lead to the expected returns demanded by the return-maximizers of Sand Hill Road.

Reading utopianist programmers’ cris de cœur like Paulette Koronkevich’s — her grief over programming’s retreat from the human into the solitary and the transactional — you hear what’s in the heart of all of us facing the AI’s rewriting of this space we built. We got in this to impress each other, to awe each other, to make each other’s burdens a little lighter; we got in it to have a friend to go blow off steam with at happy hour. We thought that we could find more humanity through facing the foreign or challenging with other humans.

Think about what used to be charming about Paul Graham’s essays: he would thank, by name, the people who read his interim drafts. That was proof of social context — the ideas had been stress-tested by actual people who pushed back. Jessica Livingston, Robert Tappan Morris, and others had all brought playground-honed skill to that conversation — the capacity to disagree meaningfully.

Now imagine replacing that acknowledgment with “reviewed by a large language model” and you lose more than charm. You announce that the author opted out of exactly what Tatar and Naur said matters most. As Tatar foresaw: “Computing does not create our social problems, but it can be different in ways that promote civilization.”