BARD: Better Automated Redistricting
Main Article Content
Abstract
BARD is the first (and at time of writing, only) open source software package for general redistricting and redistricting analysis. BARD provides methods to create, display, compare, edit, automatically refine, evaluate, and profile political districting plans. BARD aims to provide a framework for scientific analysis of redistricting plans and to facilitate wider public participation in the creation of new plans.
BARD facilitates map creation and refinement through command-line, graphical user interface, and automatic methods. Since redistricting is a computationally complex partitioning problem not amenable to an exact optimization solution, BARD implements a variety of selectable metaheuristics that can be used to refine existing or randomly-generated redistricting plans based on user-determined criteria.
Furthermore, BARD supports automated generation of redistricting plans and profiling of plans by assigning different weights to various criteria, such as district compactness or equality of population. This functionality permits exploration of trade-offs among criteria. The intent of a redistricting authority may be explored by examining these trade-offs and inferring which reasonably observable plans were not adopted.
Redistricting is a computationally-intensive problem for even modest-sized states. Performance is thus an important consideration in BARD's design and implementation. The program implements performance enhancements such as evaluation caching, explicit memory management, and distributed computing across snow clusters.
BARD facilitates map creation and refinement through command-line, graphical user interface, and automatic methods. Since redistricting is a computationally complex partitioning problem not amenable to an exact optimization solution, BARD implements a variety of selectable metaheuristics that can be used to refine existing or randomly-generated redistricting plans based on user-determined criteria.
Furthermore, BARD supports automated generation of redistricting plans and profiling of plans by assigning different weights to various criteria, such as district compactness or equality of population. This functionality permits exploration of trade-offs among criteria. The intent of a redistricting authority may be explored by examining these trade-offs and inferring which reasonably observable plans were not adopted.
Redistricting is a computationally-intensive problem for even modest-sized states. Performance is thus an important consideration in BARD's design and implementation. The program implements performance enhancements such as evaluation caching, explicit memory management, and distributed computing across snow clusters.