The main purpose of this software is to test and analyse the results of various social-choice functions.

Ballots must be complete rank-order ballots of candidates numbered 0-n as a list of lists, tuples of tuples, numpy array, or something similarly compatible.

This software can also be used to run some hard-to-find implementations of important social-choice functions such as:

The random ballot:

from lottery_scfs import random_ballot, multi_lottery
from ballot_generators import gen_ranked_preferences_zipf

pref_ballots = gen_ranked_preferences_zipf(n_candidates=5, n_voters=5000, zipf_param=1.5)
winner = random_ballot(pref_ballots)

Various variations of multi-lottery style elections:

winner = multi_lottery(pref_ballots, points_to_win=5, method='borda_decay')
winner = multi_lottery(pref_ballots, points_to_win=7, method='borda')
winner = multi_lottery(pref_ballots, points_to_win=20, method='plurality')
winner = multi_lottery(pref_ballots, points_to_win=5, method='iterated_borda')

Variants on Instant runoff such as Benham-Hare, Smith-Hare, Tideman-Hare, and Woodall Hare:

import irv_variants as irv

hare_obj = irv.IRV_Variants(pref_ballots)
winner = hare_obj.benham_hare()
winner = hare_obj.smith_hare()
winner = hare_obj.tideman_hare()
winner = hare_obj.woodall_hare()

Although the methods above may be useful, this software is made more for use to simulate several prominent social-choice functions with numerous randomly-generated samples of different types of voting populations to learn which vote-tallying techniques statistically tend to have the best desired results for democratic processes. This software includes some of my ideas for new and hopefully better social-choice functions.

The overall goal of this project is to find a social-choice function that better addresses constraints of election manipulability according to Gibbard's Theorem, and constraints from Arrow's Theorem. Another important consideration is the design of election systems that would maintain the stability and feasibility of governments, and maintain democracy in societies with highly polarized politics. Selecting a social-welfare function that optimizes social welfare relative to such constraints is the overall goal.

This software is a work in progress, and some parts may still need refining.

This code runs with Python 3 only. It is not compatible with Python 2.*

All simulations assume results based upon completely sincere voting which is never true in theory except for the random ballot. Most social-choice functions are easier than others to manipulate: https://www.rangevoting.org/mani-focs.pdf

Although you can use the tools above, the main purpose of this software is simulation to compare social-choice functions with much better resistance to manipulation to more popular social-choice functions. This is done under ideal circumstances to consider how much potential loss comes from decreasing manipulability.

You can run this simulation by navigating to the folder with the simulated_elections.py, and executing:

python3 -O simulated_elections.py

If you want to save the text output from the simulation, run:

python3 -O simulated_elections.py &>> summary.txt

The box and whisker plots from the simulations plots the social happiness results from numerous simulations of election results using a variety of popular social-choice functions, and variations of multi-lottery social-choice functions with different points-to-win parameters. Each box and whisker plot corresponds to a given number of candidates, and a given simulated type of voting populace with the associated parameters. This information is in the folder name and file name of each plot.

The bar plots are for analysing multi-lottery methods only. They provide happiness frequency results from numerous simulations on the same population, and winner frequencies on the same population. This is repeated for a variety of population types, and number of candidates, each with their corresponding plots.

The license for using this software is GPL v3 as written in the included file License.txt.