Another suggestion if we're just brainstorming. This one is for online elections like say, Debian or Wikipedia. Votes are cumulative style, either l_1 or l_2. At the start of the primary, every voter's ballot is "locked". Randomly and over time, the voters' ballots are unlocked, and they can then tentatively vote for candidates. The voters can also update their votes at any time until the ending phase. The current total/score is updated continuously and the voters may use that total to inform how they update their votes. The threshold is prominently shown (e.g. the Hare or Droop quota). Near the end of the election, voters' votes are randomly locked. The time until a ballot is locked is also shown on the voter's display, and known to the voter from the start. The gradual unlocking of ballots incentivizes the initial voters to more thoroughly look into the candidates' positions because the initial scores will inform the feedback process; and the gradual locking near the end is meant to discourage sniping attacks where a coordinated minority pretends to vote for someone else, then shifts their votes to their true favorite at the last second. Once every ballot has been locked, the first k by score proceed to a more traditional general (probably ranked voting). The core idea is that the voters would be able to see which candidates are safe and thus reallocate their votes to either spend on other candidates or to look into which of the above-threshold candidates they like the most. This is definitely not perfect. The path dependence means that it's most likely not monotone, and the continuous nature means there's room for a lot of sophisticated strategy involving making your candidate look more hopeful than he is. It can't be done for large-scale political elections either because online election security is near impossible. But it might be useful in some contexts - possibly! (If you want to be real fancy, you could probably use control theory ideas to stabilize the system further. E.g. when ballots are changed, the scores are not instantly updated, but there's a rolling average instead. But I'll leave that to people who know more about control theory than I do.) -km

In control theory for stability you want critical damping of oscillations, which means barely enough friction to prevent predicted overshoot. This control friction is proportional to the opposite of the sampled rate of change of the state variable being controlled ... in this case the cumulative vote vector. ... something like that :-) On Sat, Aug 26, 2023, 3:11 AM Kristofer Munsterhjelm <km_elmet@t-online.de> wrote: > Another suggestion if we're just brainstorming. This one is for online > elections like say, Debian or Wikipedia. > > Votes are cumulative style, either l_1 or l_2. At the start of the > primary, every voter's ballot is "locked". Randomly and over time, the > voters' ballots are unlocked, and they can then tentatively vote for > candidates. The voters can also update their votes at any time until the > ending phase. > > The current total/score is updated continuously and the voters may use > that total to inform how they update their votes. The threshold is > prominently shown (e.g. the Hare or Droop quota). > > Near the end of the election, voters' votes are randomly locked. The > time until a ballot is locked is also shown on the voter's display, and > known to the voter from the start. > > The gradual unlocking of ballots incentivizes the initial voters to more > thoroughly look into the candidates' positions because the initial > scores will inform the feedback process; and the gradual locking near > the end is meant to discourage sniping attacks where a coordinated > minority pretends to vote for someone else, then shifts their votes to > their true favorite at the last second. > > Once every ballot has been locked, the first k by score proceed to a > more traditional general (probably ranked voting). > > The core idea is that the voters would be able to see which candidates > are safe and thus reallocate their votes to either spend on other > candidates or to look into which of the above-threshold candidates they > like the most. > > This is definitely not perfect. The path dependence means that it's most > likely not monotone, and the continuous nature means there's room for a > lot of sophisticated strategy involving making your candidate look more > hopeful than he is. It can't be done for large-scale political elections > either because online election security is near impossible. But it might > be useful in some contexts - possibly! > > (If you want to be real fancy, you could probably use control theory > ideas to stabilize the system further. E.g. when ballots are changed, > the scores are not instantly updated, but there's a rolling average > instead. But I'll leave that to people who know more about control > theory than I do.) > > -km > ---- > Election-Methods mailing list - see https://electorama.com/em for list > info >