Danyc0 / Society-voting-bot

A Discord bot for doing society elections
Apache License 2.0
2 stars 4 forks source link

Consider implementing an improved elections style to avoid bias against those near the bottom of the running order #37

Open Danyc0 opened 3 years ago

Danyc0 commented 3 years ago

Support new voting style such that people standing for posts lower down the running order can still stand for other posts if they do not get elected to their desired post.

As things stand, if people stand for a post near the top of the running order, but do not get it, they can stand for a later post from the floor. This is a very good thing, but is not available to those standing for posts near the end of the running order as there are no more posts left for them to stand for from the floor.

One proposal to fix this would be:

  1. Everybody stands for all posts which they even have any sort of desire to run for
  2. All voting happens without revealing the results
  3. Announce which people won more than one post
  4. Let those candidates pick which one they want to keep
  5. Re-run the IRV voting calculations on the ballot papers for the posts that were not chosen, as if that candidate did not stand in the first place (no more voting has to happen, as this can be calculated from the original ballot papers)
  6. Repeat 5 until no more candidates hold more than one role

The issues I can see here are that: A. people in this scenario will often stand from the floor for posts which have 0/few candidates, which is unknown at the point they are standing (maybe this could be bypassed by candidates registering in advance, publically, what their preferences are for the post they want to hold, but this might amplify issue C. B. This could result in each role having a huge number of candidates C. This could bias people's voting as they may not want to vote for a candidate for more than one post (for example: "well I voted for person X for the president, so I won't vote for them for secretary"). This could maybe be mitigated by making this possibility very clear and educating people on not doing it. D. The order in which the IRV is re-run could impact the results, as those who get their IRV re-run first are more likely to get their first choice of post. Maybe this order could re randomised to avoid this?