I'm in a 12-person committee where we have to rank 5 candidates whom we have interviewed for a position. Some members of the committee have a conflict of interest, so they had to leave the room when a particular candidate was interviewed and they may not vote for that candidate (but they will rank the other candidates). What would be a fair voting system?
Ranked voting by a committee when some members cannot vote for a particular candidate because of a conflict of interest
1Do you object to calling the 12-person "jury" an "interview committee" to avoid confusion with a trial jury in a court of law (which votes unanimously only)?– ohwilleke ♦Sep 6, 2022 at 20:23
3This does not appear to be about politics as defined by this site but it might be a fit on workplace.se. Also it sounds like it would cause problems because they might not be able to vote for the most qualified candidates seems to me they are either able to vote for all of them or none of them.– Joe WSep 6, 2022 at 20:31
10@JoeW I would say that voting systems are on topic here, although the relevance is limited to just that. Its applicability to the situation and any office politics involved would be off topic. This is basically asking for a form of ranked choice voting where candidates are neither penalized nor benefited from not being chosen by a given voter.– BobsonSep 6, 2022 at 20:40
1@Bobson be careful with the term “ranked choice”, which is sometimes used for a specific member of the range of ranked/preferential voting system.– origimboSep 6, 2022 at 21:54
1I would suggest any system which prohibits someone from getting a vote from someone is unfair by the very nature. It means that people could be forced to vote for people they don't want to.– Joe WSep 6, 2022 at 22:32
they may not vote for that candidate (but they will rank the other candidates)
This is a highly problematic stipulation. A voter could be accused of tactically voting for the "weakest" candidate, so that the "strongest" competitor of his/her candidate of interest has less chance of succeeding.
A system that tries to tackle that allegation, used for example in German academia, is single elimination in an exhaustive ballot (which ultimately also produces a ranked list if you exclude instant winning by overall majority). As long as the candidate of interest is still in the race, the person that declared a conflict of interest may not take part in a voting round. Only after his/her elimination, the member may return to the table.
All in all, having a committee member with a conflict of interest is always a perilous situation. Currently, there is a prominent case involving the district mayor of Berlin Mitte. For a senior position in the district administration, the mayor took part in the selection committee, despite one of the candidates being a close affiliate, and the mayor himself declaring a conflict of interest. The mayor did not issue any opinion about his friend, but otherwise took part in the proceedings. In the end, said friend was selected for the position. Promptly, a competitor sued the district, accusing the mayor to have worked against him.
(The press articles originally detailing the case are unfortunately behind a pay wall.)
The whole case came to prominence because of what happened afterwards, and which is not strictly related: The mayor tried to placate the competitor with a compensation payment in exchange for withdrawing the legal complaint. While this is a common occurrence in German labor law when private companies are involved, a public administration might not be able to act this way (at least the district lawyers ruled it out in this case, citing budgetary restrictions). Then allegedly the mayor offered to pay the compensation from his private money (he denies this, but accepts that a message could be misunderstood). He lost the political support by most parties in the district council, including his own, and was voted out of office.
I have some friends who are members of the mayor's party district organisation. Asked about the affair, what annoyed them most was that, after declaring the conflict, he did not make it his top priority to avoid any impression of nepotism. They would have expected him to withdraw completely from the committee.
From a public point of view, a formally "fair" process might not be enough. As long as the accusation of favoritism can be made, true or not, proceedings remain vulnerable.
10That is the key point, if someone has a conflict of interest, they should not be involved in the decision at all.– quaragueSep 7, 2022 at 6:59
2+1 for mentioning that fairness is not enough, and any appearance of potential unfairness should be avoided. Complete withdrawal is the key. Sep 7, 2022 at 13:15
The problem here, of course, is that having everyone with a conflict of interest recuse themselves fully may not be practical if conflicts of interest are common enough. For an extreme example, consider the (not altogether uncommon) case where every voter is also a candidate. Sep 9, 2022 at 11:51
@IlmariKaronen I think you are geting a bit off-topic here. The question is about hiring someone to do payed work, not filling an elected office. I actually found myself once in the position of the first case. I was a member of a committee (preparing a jubilee) that posted some job offers. I was naive enough to apply, and was told in no uncertain terms that for the application to stand, I had to resign from the committee. And they were right, of course. But had the committee statutes provided to select one of its members to act as a manager, it would have been perfectly acceptable.– ccprogSep 9, 2022 at 14:22
Although there are a number of methods which are complicated by your situation here, and it's impossible for a voting system to be unambiguously fair, there are still quite a number of ranked/preferential systems which will "work" (in the sense of generating a well-ordered ranking, which might include ties).
One of the simpler options, both to explain and to implement is Score voting (also known as Range voting), providing that the scores are aggregating via averaging, rather than summation. To run this, the committee members give a numerical score for each candidate that they are eligible to judge. When processing the votes you would then take an average (typically the mean), then rank the candidates in order.
A disadvantage, and potential unfairness is that without a mark scheme, commitee members scores are likely to be arbitrary, though probably self consistent. Specifying marks in some way, even to the level of "a typical candidate should get 50, a very strong candidate 80" can mitigate this to a great extent. An advantage is that you can then trivially set a cut-off mark for acceptable candidates, useful if it's not guaranteed that the best ranked candidate will accept the position.
If time and complexity aren't a problem, then other methods can be modified appropriately, for example Copeland's method, where the committee members with a conflict don't vote on comparisons involving that candidate.
4@ohwilleke yes, but the average is the key: it gives the same opportunities to win to all candidates independently of the conflict of interest (COI). Without the average, the candidate with COI would have one vote less than the others.– effediciSep 7, 2022 at 11:30
3I wish I could select multiple answers as "the best", but since this answer gives me the practical solution that I need for this real-world problem, I have selected it as the "accepted answer". Sep 7, 2022 at 12:00
4I'm missing where you state what score the conflict-of-interest candidate should receive. 0 by default? Shouldn't they receive the average score? Also, wouldn't the "committee members scores are likely to be arbitrary" issue be solved by having a scoring ballot normalized to 0? Say -5 to 5. Sep 7, 2022 at 23:15
8I don't see how this removes the conflicted interviewers' ability to impact the outcome favourably for their preferred candidate. If they gave low scores to everyone that they are able to vote for, wouldn't that favour their preferred candidate even if they weren't able to vote for them?– AubrealSep 8, 2022 at 14:48
5I would argue that if a voter cannot score a candidate due to a conflict of interest, that candidate should receive a score corresponding to the average score given by that voter. This would mitigate the problem described by @AlexandreAubrey: If you give bad scores to all candidates, then your favoured candidate also receives a bad score. Note that this system can still be gamed though, e.g., if the conflicted voter gives a high score to a hopeless candidate to boost the voter’s average. Sep 9, 2022 at 7:04
The fact that one member cannot vote for a particular candidate due to conflict of interest introduces several problems:
It incentivises the committee to enagge in tactical voting.
It incentivises the candidates to engage in strategic persuasion during interviews.
Bar the candidate in question from the get go to avoid conflict of interest.
Require the committee member in question to not participate in any voting and ask their alternate / deputy to vote in their place. This is also a way to remove conflict of interest.
Either way, the current workaround is deeply undesirable in my opinion.
Any Condorcet method should work, provided you normalize the result of each pairing in the matrix according to the number of voters that caould express a preference.
The ballot is an ordered list of all candidates from best to worst, omitting those that you have a conflict of interest with. Then you make a table where for each pair of candidates you tally which of the two is ahead in most ballots.
If some candidate "wins" all the pairings they appear in, they're the winner. If not, there are several methods to choose from to break the "loop". I'm partial to the Shulze method to resolve such situations. There are many others.
2+1 for mentioning Condorcet. This can be used in conjunction with scoring-style ballots, too. A scoring ballot normalized to 0 easily allows a candidate to have a default score: 0. So conflict of interest is easily handled. Sep 7, 2022 at 23:11
@LittleEndian unfortunately a default average score would fail the Independence from Irrelevant Alternatives criterion badly. When top candidates score high, an average mark will put you out of the race in practice.– Rad80Sep 8, 2022 at 6:24
Not sure I was clear about the scoring part. So, ignore scoring and I'll reframe the "normalization" bit in terms of Shulze: The Shulze method fails IIA as well. In fact, it "strictly prefers all ranked to all unranked candidates". So, the conflict-of-interest becomes, in effect, the least preferred candidate. A better way to handle this is to augment a Shulze-like ballot to allow the voter to specify negative marks. In that way the voter can say "I'm indifferent about those candidates (unranked), but I really don't want these candidates (negative marks)". Sep 8, 2022 at 15:51
I guess if enough comittee members have (different) conflicts of interest, there might be pairs of candidates which won't pair of against each other at all, so you might have multiple "Concordet-winners". Sep 9, 2022 at 21:42
I'm not an expert on voting, but assuming you don't actually care about the final rankings but only want to have a single candidate who has the best ranking, here's one pretty solution:
The 11 jurors rank the all 5 candidates and the one juror that only ranks the 4.
First you look at the ranking from the 11 jurors, if the particular candidate is 1st then he's the "winner" if he isn't, he eliminated and is removed from the rankings from the jurors and you get the ranking out of all 12 jurors for the 4 candidates and get a "winner"
1There are situations where a ranked list is actually needed. For example, the committee recommends a list to the board, which has the final say and can override the committee and select someone not ranked first.– ccprogSep 8, 2022 at 13:33