CGP Grey
3 HN mentions · mentioned in 3 months, 2012–2017
3:38 · 692K views · Jul 12, 2011
A short CGP Grey bonus video explaining the shortest-splitline algorithm, which draws electoral districts by repeatedly splitting a state along its shortest dividing line while balancing population. It is offered as a neutral, algorithmic fix for gerrymandering.
Cited as the standard example of algorithmic redistricting when gerrymandering and electoral reform come up.
3 comments from Hacker News
PR mitigates Gerrymandering because it reduces the amount of "discardable" surplus votes. Imagine you have 100 single-winner districts that are each roughly 50/50 Dem/GOP, with the slight edge breaking 50/50, so that your legislature has 50 reps from each party. Now we redistrict to pack a bunch of Dems from the GOP-majority districts into one super-Dem district, shaving down their margins in numerous districts. We could conceivable get 99 GOP reps even though the population is 50/50 Dem/GOP. That super-Dem district "discards" or "wastes" 50% of the votes of the Dems inside it, because they're surplus votes in excess of what they need to win the district. Now imagine you switch to 20 5-winner districts elected with PR. The Dems could win all five seats with conceivably 83.3% of the vote (5 candidates getting 16.6% each means they get all the votes). So the most you can possibly wast is 16.7% of the votes instead of 50%. > It only replaces districts with parties No. Most PR systems still have districts. Exceptions like Israel, where the whole country is one big district, tend to be the poster children of PR gone wrong. Plus there are PR systems where you vote for candidates, not parties. Single Transferable Vote in Australia's Senate is an example. You just rank the candidates (although many voters blindly follow the guidance of their preferred party). Another cure to Gerrymandering is algorithmic redistricting, which I'm supportive of. this video As for Score Voting (aka Range V
I always think that the best way to combat the unwillingness of the commoner to actively engage in politics is to bring them to the commoner. People behave very differently when comparing communities to even just the town or city they live in. They behave to their neighbor much differently than someone a 15 minute drive from, and absolutely differently from someone they never see. Given that (in the US at least) approximately 1.5~ in 10 people are public sector workers (I forgot where I got that statistic, but anything > 1 in 50 would suffice), you can easily go full republic and have small communities of 1,000 people (psychological analysis studies seem to lean towards something in the 700 - 1300 range being the maximum number of people any one individual can actively relate to at a human level) who elect a singular representative amongst themselves every year (with no re-elections, and at such a small scale, the politics of campaign finance break down completely - everyone should know each other.) You simply select, emotionally, the person amongst yourselves you feel best represents you individually, and those that want to take the local seat can debate one another in forum with whoever is interested. The rest can be expected to take sides, but the beauty of this idea is that as long as people remain mobile, you can expect common political interests to attract one another, and as long as you can preserve individual mobility (which is much easier in the modern world) you can
I'd prefer the algorithm if it was public and neutral (or at least consistent). See for example the Shortest Splitline Algorithm: this video