Link Search Menu Expand Document

Thoughts on Error Messages

Date: 2020-10-20

context: I’m thinking about error feedback in general, and I’m also thinking about the sorts of error msgs programmers get in the regular course of programming.

  • are error messages a good way to organise issues? (e.g. in software dev).
  • they have an important role: they guide ppl who know less (than the developers, or other community members, etc).
  • if error msgs were a bad way to organise issues then there must be a better alternative system. what would such a system be like relatively speaking?
    • it would put more burden on the ppl affected by errors b/c it’s harder to know/learn how to report and solve the errors
    • it would mean responsibility for the quality of error reporting would be shifted towards the shoulders of newbies
    • such an alternative system would treat the relevant (preventable!) errors less seriously
  • why could that be good?
    • it’d mean there was a higher bar for engaging with top tier ppl
    • it filters out ppl who are not able to understand the problem at least enough to figure out how to begin to deal with it
    • if the best ppl don’t know how to prevent relevant errors then isn’t it better for them to focus on solving those problems rather than helping ppl who aren’t as valuable?
  • why could it be bad?
    • higher bar to error correction -> less error correction
    • easy to discourage ppl and end up reinforcing static memes / driving ppl away
    • if the best ppl didn’t know how to prevent the relevant errors then they end up working on the problem anyway; makes sense that there’s an equilibrium here; after all, ppl are voluntarily participating on both sides.

relevant to:

  • helping the best ppl or helping the masses
  • error msgs and ~responsibility of senior members
  • there is no one constant set of behaviours that makes sense WRT helping the best ppl vs masses, what matters is context. is it a good time to help one or the other? if lots of ppl have really bad ideas then it’s probably worth helping the best ppl – so we can find a good soln to that problem. conversely, if we don’t have any great ppl at that time, or are otherwise short of great opportunities, then there’s more utility helping the masses. there needs to be fertiliser for future generations, but also nourishment for current ppl in their prime. those great opportunities can be vicarious, ofc. Man’s first journey to the Moon was a journey shared by a Nation.
  • there’s a big question raised by this: how should we react to learning of a great opportunity?

Finishing up: what happens if someone goes to an effort to make error msgs as good as possible?

  • organisation gets better b/c the error messages are better suited to the associated errors
  • it gets easier for ppl to help with / do error correction b/c the msgs/explanations match the contextually best ideas more closely and are more reliable to reason about.
  • exponential/geometric increase in effectiveness of relevant key ppl – their time can be better allocated, delegation gets easier, etc
  • mutually beneficial for all parties. (note: this relies on the ability to improve error msgs and the right ~economic context to make it the easy choice. OTH I think that’s reasonably common. most non-optimum situations don’t hurt much and can be easily controlled via the 2nd-derivative (~acceleration). if there’s a bit too much work on good error msgs then you can just reduce the hours per week by 10%; it can be gentle without much harm. the harm I mention here is wasted resources in a generic sense.)

clarifying stuff

I didn’t put a huge amount of thought into particular word choices because they felt difficult and I didn’t want to ruin the flow. Here are some clarifications:

  • responsibility as in ~responsibility of senior members: i don’t mean anything like an obligation, but if there was a clear moral decision then it’d line up with that.
  • 2nd-derivative (~acceleration): controlling the rate-of-rate-of-change is useful if you want to control the outcomes of some (simple enough) system, and acceleration is a reasonably common way of talking about that.

You can leave a comment anonymously. No sign up or login is required. Use a junk email if not your own; email is only for notifications—though, FYI, I will be able to see it.

Comments powered by Talkyard.