[EM] Rank Codes

[Juho Laatu]

I guess this idea overlapping marks interpretes as equal rankings would work with any kind of marks (dashes, diagonals, ...). The number of candidates a voter can rank would depend on how thin marks (probably vertical lines) the voter can draw (and how thin lines the computer can read). When the computer notes x% overlap, those rankings would be read as equal.

The computer program should probably be smart in the sense that if there are two separate markings on one row, maybe the first one would count, and the second one would be ignored.

Also, if the centre of the mark is considerably darker or thicker than the edges to the left and right, only the centre part could be interpreted as the intended mark.

The required overlap (x) could be different for different size markings (more overlap required for large markings). This would make it easier to mark ties when the marks are already getting thin. And it would be ok to make relatively large (slightly overlapping) marks without accidentally marking candidates equal.

If there is a set of candidates that are all connected to each others based on the x% overlap rule, but the leftmost mark and the rightmost mark are already far apart, it would make sense to break that chain of equals at some point. That point could be e.g. where the overlap is weakest. That splitting process would continue until the leftmost and rightmost mark would be close enough to each other, so that they can be considered "meant to be equal".

If there is a mixture of wide and narrow markings, even some more rules may be needed. The final algorithm could become quite complex and heuristic by nature, but that is not really a problem, since all the rules are anyway meant to understand the intended preferences of the voter better. It is not a problem if some of the candidates will be interpreted as equal although they were not intended to be equal (but just almost equal), and the other way around. I guess this is the main point of this mail - use of heuristics is quite ok here.

The pens that are used in making the marks should leave a clear mark, i.e. they should maybe not be hard pencils whose marks could sometimes be ignored by the computer (if the marks are weak enough). It is better to err in tie vs. not a tie than in ranked vs. not ranked at all.

A recount, made by the same computer and algorithm that made the first count, could give somewhat different results at the second time since the mechanics in the optical reading device could give slightly different images each time. This is not a big problem. One just has to be able to agree which reading process of the ballots is the final one.

(Also an electronic voting machine could have a similar user interface, maybe with ability to push the already marked candidates to the left or to the right. But that's already another story, with other challenges. That machine could print paper ballots with (more compact and exact) number based rankings.)

BR, Juho


> On 15. Feb 2021, at 18.21, Hahn, Paul <manynote at wustl.edu> wrote:
> 
> I was about to suggest pretty much this very thing, though I might have used diagonal marks, and ½-2/3 overlap instead of 80%.
>  
> --pH
>  
> From: Election-Methods <election-methods-bounces at lists.electorama.com> On Behalf Of Forest Simmons
> Sent: Sunday, February 14, 2021 7:00 PM
> To: EM <election-methods at lists.electorama.com>
> Subject: Re: [EM] Rank Codes
>  
> IMHO Juho's idea of a kind of slider sketch is the best ever on this topic ... truly worthy of a Toby award!
>  
> If horizontal dashes are allowed, then the midpoint determines the rating, and two dashes with more than 80 percent overlap should count as equal ... the shorter the width of a dash, the more obvious lack of intended overlap, etc.
> ----
> Election-Methods mailing list - see https://electorama.com/em for list info

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20210216/e4552759/attachment.html>