Many instructors see the use of LLMs like ChatGPT on course assignments as a straightforward case of cheating, and try hard to prevent their students from doing so by including new warnings of consequences on their syllabi, turning to iffy plagiarism detectors, or scheduling exams to occur in-class. And the use of LLMs probably is cheating, given the sorts of assignments we are used to giving and the sorts of skills we take ourselves to be instilling in our students. But (...) despite legitimate ethical and pedagogical concerns, the case that LLMs should never be used in academic contexts is quite difficult to see. Many primary and secondary schools are cutting back their writing instruction in an effort to teach to the test; at the same time, many high-end knowledge workers are already quietly expected to leverage their productivity with LLMs. To prepare students for an ever-changing world, we probably do have to teach them at least a little bit about writing with ChatGPT. (shrink)

Building on work by Peterman, Aas, and Wasserman (2021), we modify their prospective benefit analysis to include only medically-relevant information about patients as persons without reference to their broader lives. Because patients (not their lives) must be treated equally, we argue that patients are the appropriate targets of allocative justification. We go on to challenge some of our current data-collection practices on this basis.