Imagine you are Gus Gorman, and you live in the United States – a country which uses the Electoral College for their national election process – while each state or province may award electors based on the popular vote (either an all-or-nothing approach, or proportionally), a pure “most votes” count does not necessarily indicate a national winner.
Oh – and you are a systems programmer for the company that builds and distributes all of the electronic voting machines for the country. These machines print a Scantron-like paper slip with the voter’s choices marked as “filled-in”, and candidates they did not vote for left “empty”.
How might you be able to “put your finger on the scale” to increase the likelihood of your preferred national candidate to win an election?
41 years ago, you successfully captured all of the truncated fractions of a cent “lost” on dollars-and-cents transactions at a previous employer and were able to buy yourself a fancy sports car (which, unfortunately, is what alerted your employer to your fraud).
Now, though, you have decided to give one last hoorah to your career – one which will never be caught, yet you will know you did – you will not make the mistake Stanley Motss did.
How would you go about perpetrating the greatest fraud in history?
Might I suggest you take a listen to this for a moment or two – https://www.youtube.com/watch?v=Pe4tUzxCmWY?
Welcome back, Gus. What have you come up with?
Allow me to give you a suggestion – one I am sure someone as clever as Gus could greatly improve upon.
The problem with systematized fraud like this, is that it needs to be subtle – which translates into needing to be extremely spread-out. It also needs to be “unverifiable” – sans some kind of exhaustive, public (and published) source code review, discovering the fraud needs to statistically “impossible”. So here is how I would do it – if I were both so inclined and in a position to “pull it off” (which, thankfully for all of you, I am not).
It all comes down to semi-randomized, weighted vote-flipping. There are two scenarios for which vote-flipping by the software has to be completely avoided (and/or reverted): first, if someone votes “straight ticket”, the machine is not allowed to flip the vote for the top-of-the-ticket candidate. And second, if, when reviewing their ballot choice on the printed slip, a voter notices the flip happened and they need to redo their vote, the process to invalidate the paper slip (prior to final submission) will also tell the voting machine to not flip the choice this time around.
With those two “do not flip” scenarios out of the way … here is one way you could increase the likelihood of your preferred candidate winning: set a [small] percentage random-number generator (ideally, use a CSPRNG (there is already one available (or one that is extremely close) on most computers)) to change votes from your ‘undesired’ candidate to your ‘desired’ candidate on a minimal basis AND a smaller percentage chance of your ‘desired’ candidate to flip to your ‘undesired’ one – say … 1/500 times a vote for your undesired candidate will flip, whereas 1/2000 a vote for your desired candidate will change to your undesired one.
Oh! and one last thing … in states wherein ‘your’ candidate is a “clear winner”, take more votes from them and give them to ‘not your’ candidate (and, of course, the inverse – in a state where ‘not your’ candidate is dramatically ahead, give ‘your’ candidate a higher percentage of flipped votes.
How does this work mathematically? Take the “clear winner” scenario first – say there is a state where your candidate is ahead (both in current pre-election polling, and historically by party) by a very wide margin – 2:1, 3:1, 7:4, etc. It will not “hurt” your candidate to “lose” a larger percentage of votes there for the opponent (for example, out of 1,000,000 voters, you ‘know’ about 750,000 will vote for your person, and only 250,000 for the other person – it will not ‘hurt’ the total outcome to drop that by a few thousand (or even a couple tens of thousands) for your candidate and give them to their opponent – 75:25 vs 72:27 will have no effect on the Electoral outcome for the state).
By ensuring you vote-flip some votes for all candidates, albeit in a weighted manner, you can keep the overall numbers very very similar – 30,000 votes in a ‘blow-out’ state can be spread around in ‘close’ states a lot more smoothly than just magicking 30,000 spare votes with no change anywhere else.
But, in the US, it is precisely the “close” states that end up swinging the election one way or the other. In a state with 1,000,000 voters who are split as close to 50:50 as you can imagine, a couple-few thousand votes in one direction or the other can have a dramatic effect on the outcome: flipping 1 of every 500 votes for “the other candidate” to “your candidate” is 1000 more votes for “your person” (out of 500,000 voters), while flipping only 1 of every 2500 votes in the other direction is just 200 extra votes for “the other person” – that may be too close to avoid a recount, but those values can be adjusted in those tighter states easily enough – change it to 1:100 votes away from “the wrong” candidate, and you give “your candidate” 5000 spare votes in this example, well within the statistical margin of error in a tight race lie the one I outlined.
But because you are also giving “spare” votes to “the other candidate” elsewhere, the overall numbers will barely change – but where they count the most, they could be swung quite dramatically.
Of course, we all know this could never happen, right?
Right?