The article doesn’t seem to be describing a system or “algorithm” that can be universally used to show some “best” drawing of election districts. I am quite certain that is an impossible and hopeless quest.
There is a better solution to gerrymandering that makes drawing and redrawing of districts almost irrelevant. It involves, for example in the case of U.S. Congressional elections, the combining of districts and then using ranked choice voting to automatically choose
multiple candidates that fairly and
proportionally represent the genuine will of all the voters.
Many Americans are now becoming aware of ordinary Ranked Choice Voting (RCV) and its superiority as a system for choosing a single winner of an election with three or more candidates. The same RCV system can be adapted to choose
multiple winners of combined contiguous election districts and results in fairer and more democratic outcomes.
This is a “Gold Standard” reform that — probably after RCV becomes better understood and more widely instituted — can solve the party-partisan gerimandering problem, and others as well. It should be noted that while this system in practice will probably lead to a fairer representation of minority and women voters, it is completely neutral, color blind and gender blind, and is “proportional” not in terms of these categories, or even of parties, but in terms of votes actually cast.
It is used already in Europe in several countries. It takes some study to understand its inherent advantages, though for voters themselves it should not prove much more complex or difficult than an RCV vote for a single vacant position.
Here are details:
fairvote.org