Junk Drawer Logo Junk Drawer

For all those little papers scattered across your desk

I stand for the Constitution, for due process, and for community that takes care of each other. Hope is a discipline.

Twisting facts to suit machines

D. Ben Knoble on 28 Oct 2026 in Blog

A thought from I Don’t Want to Read What You Didn’t Write.

Colin Breck wrote his article in the spirit of many others on the LLMish being shoved into our eyeballs. At the end, he quotes Simon Sarris’s Resist Summary:

If you ask a machine to do something for you, you are at risk of not asking for what you want, but what it can do well. If what it can do well is summarize, and you rely upon the tool, you may tune your questions so that you get better answers. This looks like training a machine to do what you want, but its also training yourself to ask questions about the world in a certain way. What are the questions you are omitting?

I’ve written before about the effects of data and machine-oriented thinking on our conversations. Even if you doubt that LLMs affect our language, it is clear that using tools affects us, and so I think that of course using a machine to answer a question will twist the questions we ask.

That’s but a twist on Goodhart’s Law: “When a measure becomes a target, it ceases to be a good measure.” Or, how we arrive at answers will change the questions we ask. Everything we do changes incentives, and we must be careful lest we incentivize the wrong behavior.

Perniciously, as Simon points out, it can feel like we are bending the machine even as it also bends us. We know from Configuration Work research that this is a normal outcome of integrating LLMs in the workplace.

Remember, the machines have only ever existed to facilitate our human goals!


Tags:

Categories: Blog

Load Comments
Previous Next
Back to posts