In the midst of grappling with my personal AI Policy
Educator/author Hank Green faced some backlash last month for his scripts sounding AI generated. Regardless of whether those exact videos were or not, Hank admitted to overreliance on LLM’s and did something really positive in the wake of the criticism: He established a personal AI policy.
As AI norms continue to be established, broken, and re-established, one thing we can do is ideate on our own policy to hold ourselves to and allow others in the professional and creative world (whichever our content belongs to) to see, helping instill trust in our work.
I am still working through the development of mine, but as I have been talking to people I come back to three tenets I want to center in it: creativity, connection, and privacy.
Creativity: We know that LLM’s are probabilistic. They gravitate to the average; their outputs are homogenized (Doshi and Hauser, 2024). I (over)simplify it to my students as auto-correct on steroids: guessing, one word(token) at a time, what most likely comes next from all the context before it. And the recent discussion of text watermarking gives another reason to think about this problem. Anthropic dismisses choosing between words like “overcast” and “grey,” saying, “Under most circumstances, it doesn’t matter much to the reader which of these latter two words the model ultimately chooses” (Anthropic, 2026). But every word does matter. Phrasing is music. I do not want an LLM to constrain my thinking by succumbing to cognitive surrender (Shaw and Nave, 2026) and accepting output I know is not wholly my intention. I even worry about letting it revise writing, designs, or code for fear of it making inadvertent changes to traits earnest to my style. It is able to accelerate idea generation; It is able to automate the tedious. But not all friction is tedium. Some friction is critical to creativity. And creativity is critical to the human endeavour. My personal AI policy, when fully grappled with (will it ever be fully grappled with?) will prioritize original first drafts and scrutiny in LLM assisted revisions.