COMPUTING SCIENCE
Machine Politics
Brian Hayes
Reapportionment and redistricting are vexing problems of
meta-politics. They are "meta" issues because they concern
the rules of the political process itself—the way the teams
are chosen. In the American political system, a decision about
reapportionment and redistricting helps determine who will make all
other decisions for the next 10 years. It's no surprise, then, that
disputes over these matters have often been rancorous. The first
Presidential veto in American history (handed down by Washington in
1792) rejected a Congressional reapportionment plan. By the 1920s
the reapportionment issue had become so contentious that the decade
ended before Congress could agree on a new formula. More recently,
hundreds of redistricting plans have been challenged in court; two
Supreme Court decisions last summer invalidated Congressional
districts in North Carolina and Texas.
This history of bitter conflict prompts speculation on the
meta-meta-political question of how best to resolve meta- political
questions. In particular: Would it be feasible to take the process
out of politics—indeed to take it out of human hands
altogether? The answer is surely yes. Computer programs could
readily draw legislative districts. Drawing good districts,
however, is a more challenging assignment. And harder still would be
persuading the legal and political establishment to give up control
of the process and accept an algorithmic solution.
Whether or not computerized redistricting would make for good
government, it offers some interesting exercises in mathematics and
computer science. Algorithms for redistricting exploit techniques
from computational geometry, graph theory, combinatorics and
optimization methods. Even if such algorithms are never embodied in
law, perhaps they can suggest some ideas that would be useful in a
more conventional approach to redistricting.
» Post Comment