Maybe once every 6 - 24 months, S1 tells a lie on my behalf. I'd like a lot more control over this bad habit S1 has.
Some context/first person phenomenology: personally, I experience most of the speech I utter as "system one like". It's fluent and automatic. Sometimes I can even do a certain mental move where "I" am able to watch the speech flow through me and out my mouth. Sometime I speak much more with S2 in the loop, and that's very qualitatively different: when I need more precision over my meanings, it's slower and more resource intensive and "I" am much more in the loop.
The lies that particularly bother me tend to be elicited when I get a query that's unexpected and somehow has some personal shame attached to it, or a sense that the answer might be upsetting to the person I'm talking to. These lies flow out just as a part of S1 producing speech utterances and reaching for the most immediately local way to solve a problem.
I legitimately don't have many things in my life that have a lot of shame attached to them, and the unexpected nature of the query is an integral part of what's happening -- if S2 has a chance to intervene, it's a different kettle of fish.
How does one work on these kinds of extremely rare S1 reflexes?