Last year, my university lost track of me. A date-based system decided I had graduated, but a paperwork-based system lacked one administrator’s signature. With my institutional identity fragmented, I ended up without health insurance for almost a month.

​​​I study what happens when rigid social practices fail to account for each other’s existence. This happens all the time. You can’t get a bank account without an address, but you can’t rent without a bank account. Worse, many folks profit off of these institutional mismatches; some duct-tapers’ careers even depend on their existence! So what to do? Can we reassess the fixity of our institutions before we hitting the point of crisis? And how can we amend our complex institutions without introducing even worse institutional collisions down the line?

On how effective altruists can become trapped on the firefighter’s treadmill.

On how political riots can ocularize deep grievances within the status quo.

Most AI researchers don’t think of their work as highly social. But social forces decide who does and doesn’t get the compute and funding to keep working. As a result, social dynamics subtly constrain both the form and direction of AI research.

​

I study the social lives of AI technologies—how social forces shape them from development through deployment. Which papers become classics? Which metrics become standard? What counts as a “good” result, and why are most of these charts so lousy?! The answers depend on social micropractices that help police what counts as “good research practice.” But what researchers value does not always align with what practitioners actually need! Once AI tools are deployed, they encounter completely different social micropractices and priorities.

On the value of creating assignments to teach dialogical writing with LLMs.

On overlooked epistemic limitations when evaluating LLM-as-a-Judge.

Our terms for physician-assisted suicide are constantly shifting. Is it physician-assisted death, medical aid in dying, or medical assistance in dying? And what if pursuing descriptive accuracy has ethically impoverished our language?

​

I study the tradeoffs of such terminological merry-go-rounds. For one thing, patients’ loved ones end up asking why grandma might need help dying. But at a deeper level, our words are shared tools that shape our ethical thinking. I argue that our pursuit of “value-neutral” description at all costs not only overlooks the fact that the new terms quickly acquire connotations in usage. More strikingly, this normative hollowing of our language is actively undermining our ability to publicly engage with deep value questions inherent to matters of life and death.

On the hollowing of ethical well-being into narrow economic self-interest.

On the hollowing of caring for patients into economizing for QALYs.