# MinorityProgrammers: ElectionHackathon $1000 prize Design a platform for your own democratic voting system

Not sure if they want to simulate elections or they want a good systemâ€¦ I mean lets just give them the list

RRV

SSS

SMV

SPAV

They want this, in 17 days:

Develop a software solution that has its own democratic system for voting on issues and / or candidates.

The problem is that in the â€śIssues to Considerâ€ť section they ask for a lot of things, extremely complicated to do properly in just 17 days.

Even if you do, $1000 + â€śprizeâ€ť is nothing for such a project.

However, on a theoretical level I propose this:

Distributed Voting

A project that would be entered in this system would seem to concern several issues that are not completely contingent. For example, one system could have a great UI but use a terrible method for winner selection. (Or vice versa). Whatever wins, it may well be improvable by borrowing elements from some of the losers. They actually have a long list of concerns, but any given project could probably not address them all. IMO it would be worthwhile to recognize winners in different subcategories.

Their examples of proportional systems are party list and STV. Approval-based proportional systems should be considered *at least* as proportional as party list. You could brand them as â€śpick your own party listâ€ť if you wished. They do recognize approval as a single winner system, so I wonder if they are even aware of proportional approval systems.

They want software that can be used to improve the democratic system somehow. I think you could submit an implementation of the shortest splitline algorithm that takes as input raster data on population and it would qualify. But just the mathematical definition of a procedure would not.

I think we could work together and come up with a very good system using elements of the systems we know are fairly effective. I have a suggestion that is a utilitarian modification of STAR (with discrete integer scores from 0 to N). Rather than taking the sum of score values for a candidate, use a different measure on the score distribution as follows:

Choose some positive real number B in the interval [0,1]. The assign each candidate the metric

SUM[k=0 to N] p(k)*B^k

where p(k) is the fraction of voters who scored that candidate as a k, and then choose the two candidates with the two *lowest* metrics to compete in the runoff.

I think N=5 and B~0.6 is a reasonable choice for the parameters. @Essenzia and I were discussing systems with similar properties here:

If you do the sum in the first round, there is always the problem of min-maxing.

Min-maxing is still a potential issue, but the exponential payoffs of lowering the score of a candidate in terms of preventing them from overtaking a more preferred candidate may entice voters to indicate their preferences between candidates rather than to min-max. I consider how I would vote in this system, and I do think that I would rather indicate my preferences than do a full min-max, I think it would probably be more likely to pay off well for me. Or, at the very least, itâ€™s much less clear to me what is a better strategy, especially with the runoff. Not to get off topic here though.

In Score Voting the min-maxing strategy consists precisely in making oneâ€™s interests â€śexponentialâ€ť which honestly would be linear.

An honest vote like this, with linear values:

A [5] B [4] C [3] D [2] E [1] F [0]

strategically it would become like this:

A [5] B [3] C [2] D [1] E [0] F [0]

to the extreme, it would become like this:

A [5] B [5] C [0] D [0] E [0] F [0]

If the idea is â€śI give the voter ratings that are already exponential, to make the vote more honestâ€ť however, a vote like this in your system:

A [5] B [4] C [3] D [2] E [1] F [0]

it wouldnâ€™t necessarily be honest.

My reasoning is different:

I want to give the voter an exponential scale so that he can better represent his interests.

If the ratings are:

{1,4,9,15,16} or {1,2,4,8,16} instead of {1,2,3,4,5} the voter in the first case would have a range [1 , 16] even if the ratings are only 5 (that is, a simple vote to write like that [1,5] but more precise).

In general, the min-maxing problem can be reduced but as long as it remains as a possibly effective tactic, voters will end up using it anyway.

The only real optimal solution is to make sure that using min-maxing will almost always produce worse results.

(post withdrawn by author, will be automatically deleted in 24 hours unless flagged)