class svvamp.RuleMajorityJudgment(**kwargs)[source]#

Majority Judgment.

Parameters:

  • min_grade (number) – Minimal grade allowed.

  • max_grade (number) – Maximal grade allowed.

  • step_grade (number) –

    Interval between two consecutive allowed grades.

    • If step_grade = 0, all grades in the interval [min_grade, max_grade] are allowed (‘continuous’ set of grades).

    • If step_grade > 0, authorized grades are the multiples of step_grade lying in the interval [min_grade, max_grade]. In addition, the grades min_grade and max_grade are always authorized, even if they are not multiples of step_grade.

  • rescale_grades (bool) –

    Whether sincere voters rescale their utilities to produce grades.

    • If rescale_grades = True, then each sincere voter v applies an affine transformation to send her utilities into the interval [min_grade, max_grade].

    • If rescale_grades = False, then each sincere voter v clips her utilities into the interval [min_grade, max_grade].

    See ballots for more details.

  • Options

  • ------- – 

    >>> RuleMajorityJudgment.print_options_parameters()
cm_option: ['exact']. Default: 'exact'.
icm_option: ['exact']. Default: 'exact'.
iia_subset_maximum_size: is_number. Default: 2.
im_option: ['exact']. Default: 'exact'.
max_grade: isfinite. Default: 1.
min_grade: isfinite. Default: 0.
precheck_heuristic: [False]. Default: False.
rescale_grades: is_bool. Default: True.
step_grade: isfinite. Default: 0.
tm_option: ['exact']. Default: 'exact'.
um_option: ['exact']. Default: 'exact'.

Notes

Each voter attributes a grade to each candidate. By default, authorized grades are all numbers in the interval [min_grade, max_grade]. To use a discrete set of notes, modify attribute step_grade.

Note

Majority Judgement, as promoted by its authors, uses a discrete set of non-numerical grades. For our purposes, using a discrete set of numerical grades (step_grade > 0) is isomorphic to this voting system. In contrast, using a continuous set of grades (step_grade = 0) is a variant of this voting system, which has the advantage of being canonical, in the sense that there is no need to choose the number of authorized grades more or less arbitrarily.

The candidate with highest median grade wins. For the tie-breaking rule, see scores.

Default behavior of sincere voters: voter v applies an affine transformation to her utilities preferences_ut[v, :] to get her grades, such that her least-liked candidate receives min_grade and her most-liked candidate receives max_grade. To modify this behavior, use attribute rescale_grades. For more details about the behavior of sincere voters, see ballots.

  • not_iia():

    • If rescale_grades = False, then Majority Judgment always meets IIA.

    • If rescale_grades = True, then non-polynomial or non-exact algorithms from superclass Rule are used.

  • is_cm_(), is_icm_(), is_im_(), is_tm_(), is_um_(): Exact in polynomial time.

References

Majority Judgment : Measuring, Ranking, and Electing. Michel Balinski and Rida Laraki, 2010.

Examples

>>> profile = Profile(preferences_ut=[
...     [ 0. , -0.5, -1. ],
...     [ 1. , -1. ,  0.5],
...     [ 0.5,  0.5, -0.5],
...     [ 0.5,  0. ,  1. ],
...     [-1. , -1. ,  1. ],
... ], preferences_rk=[
...     [0, 1, 2],
...     [0, 2, 1],
...     [1, 0, 2],
...     [2, 0, 1],
...     [2, 1, 0],
... ])
>>> rule = RuleMajorityJudgment()(profile)
>>> rule.demo_results_(log_depth=0)  

************************
*                      *
*   Election Results   *
*                      *
************************

***************
*   Results   *
***************
profile_.preferences_ut (reminder) =
[[ 0.  -0.5 -1. ]
 [ 1.  -1.   0.5]
 [ 0.5  0.5 -0.5]
 [ 0.5  0.   1. ]
 [-1.  -1.   1. ]]
profile_.preferences_rk (reminder) =
[[0 1 2]
 [0 2 1]
 [1 0 2]
 [2 0 1]
 [2 1 0]]
ballots =
[[1.   0.5  0.  ]
 [1.   0.   0.75]
 [1.   1.   0.  ]
 [0.5  0.   1.  ]
 [0.   0.   1.  ]]
scores =
[[ 1.    0.    0.75]
 [-2.    2.   -2.  ]]
candidates_by_scores_best_to_worst
[0 2 1]
scores_best_to_worst
[[ 1.    0.75  0.  ]
 [-2.   -2.    2.  ]]
w = 0
score_w = [ 1. -2.]
total_utility_w = 1.0

*********************************
*   Condorcet efficiency (rk)   *
*********************************
w (reminder) = 0

condorcet_winner_rk_ctb = 0
w_is_condorcet_winner_rk_ctb = True
w_is_not_condorcet_winner_rk_ctb = False
w_missed_condorcet_winner_rk_ctb = False

condorcet_winner_rk = 0
w_is_condorcet_winner_rk = True
w_is_not_condorcet_winner_rk = False
w_missed_condorcet_winner_rk = False

***************************************
*   Condorcet efficiency (relative)   *
***************************************
w (reminder) = 0

condorcet_winner_ut_rel_ctb = 0
w_is_condorcet_winner_ut_rel_ctb = True
w_is_not_condorcet_winner_ut_rel_ctb = False
w_missed_condorcet_winner_ut_rel_ctb = False

condorcet_winner_ut_rel = 0
w_is_condorcet_winner_ut_rel = True
w_is_not_condorcet_winner_ut_rel = False
w_missed_condorcet_winner_ut_rel = False

***************************************
*   Condorcet efficiency (absolute)   *
***************************************
w (reminder) = 0

condorcet_admissible_candidates =
[ True False False]
w_is_condorcet_admissible = True
w_is_not_condorcet_admissible = False
w_missed_condorcet_admissible = False

weak_condorcet_winners =
[ True False False]
w_is_weak_condorcet_winner = True
w_is_not_weak_condorcet_winner = False
w_missed_weak_condorcet_winner = False

condorcet_winner_ut_abs_ctb = 0
w_is_condorcet_winner_ut_abs_ctb = True
w_is_not_condorcet_winner_ut_abs_ctb = False
w_missed_condorcet_winner_ut_abs_ctb = False

condorcet_winner_ut_abs = 0
w_is_condorcet_winner_ut_abs = True
w_is_not_condorcet_winner_ut_abs = False
w_missed_condorcet_winner_ut_abs = False

resistant_condorcet_winner = nan
w_is_resistant_condorcet_winner = False
w_is_not_resistant_condorcet_winner = True
w_missed_resistant_condorcet_winner = False
>>> rule.demo_manipulation_(log_depth=0)  

*****************************
*                           *
*   Election Manipulation   *
*                           *
*****************************

*********************************************
*   Basic properties of the voting system   *
*********************************************
with_two_candidates_reduces_to_plurality =  False
is_based_on_rk =  False
is_based_on_ut_minus1_1 =  True
meets_iia =  False

****************************************************
*   Manipulation properties of the voting system   *
****************************************************
Condorcet_c_ut_rel_ctb (False)     ==>     Condorcet_c_ut_rel (False)
 ||                                                               ||
 ||     Condorcet_c_rk_ctb (False) ==> Condorcet_c_rk (False)     ||
 ||           ||               ||       ||             ||         ||
 V            V                ||       ||             V          V
Condorcet_c_ut_abs_ctb (False)     ==>     Condorcet_ut_abs_c (False)
 ||                            ||       ||                        ||
 ||                            V        V                         ||
 ||       maj_fav_c_rk_ctb (False) ==> maj_fav_c_rk (False)       ||
 ||           ||                                       ||         ||
 V            V                                        V          V
majority_favorite_c_ut_ctb (False) ==> majority_favorite_c_ut (False)
 ||                                                               ||
 V                                                                V
IgnMC_c_ctb (True)                 ==>                IgnMC_c (True)
 ||                                                               ||
 V                                                                V
InfMC_c_ctb (True)                 ==>                InfMC_c (True)

*****************************************************
*   Independence of Irrelevant Alternatives (IIA)   *
*****************************************************
w (reminder) = 0
is_iia = True
log_iia: iia_subset_maximum_size = 2.0
example_winner_iia = nan
example_subset_iia = nan

**********************
*   c-Manipulators   *
**********************
w (reminder) = 0
preferences_ut (reminder) =
[[ 0.  -0.5 -1. ]
 [ 1.  -1.   0.5]
 [ 0.5  0.5 -0.5]
 [ 0.5  0.   1. ]
 [-1.  -1.   1. ]]
v_wants_to_help_c =
[[False False False]
 [False False False]
 [False False False]
 [False False  True]
 [False False  True]]

************************************
*   Individual Manipulation (IM)   *
************************************
is_im = False
log_im: im_option = exact
candidates_im =
[0. 0. 0.]

*********************************
*   Trivial Manipulation (TM)   *
*********************************
is_tm = False
log_tm: tm_option = exact
candidates_tm =
[0. 0. 0.]

********************************
*   Unison Manipulation (UM)   *
********************************
is_um = False
log_um: um_option = exact
candidates_um =
[0. 0. 0.]

*********************************************
*   Ignorant-Coalition Manipulation (ICM)   *
*********************************************
is_icm = False
log_icm: icm_option = exact
candidates_icm =
[0. 0. 0.]
necessary_coalition_size_icm =
[0. 6. 4.]
sufficient_coalition_size_icm =
[0. 6. 4.]

***********************************
*   Coalition Manipulation (CM)   *
***********************************
is_cm = False
log_cm: cm_option = exact
candidates_cm =
[0. 0. 0.]
necessary_coalition_size_cm =
[0. 3. 3.]
sufficient_coalition_size_cm =
[0. 3. 3.]
property ballots_#

2d array of integers. ballots[v, c] is the grade attributed by voter v to candidate c (when voting sincerely). The following process is used.

  1. Convert utilities into grades in the interval [min_grade, max_grade].

    • If rescale_grades = True, then each voter v applies an affine transformation to preferences_ut[v, :] such that her least-liked candidate receives min_grade and her most-liked candidate receives max_grade. Exception: if she is indifferent between all candidates, then she attributes (min_grade + max_grade) / 2 to all of them.

    • If rescale_grades = False, then each voter v clips her utilities into the interval [min_grade, max_grade]: each utility greater than max_grade (resp. lower than min_grade) becomes max_grade (resp. min_grade).

  2. If step_grades > 0 (discrete set of grades), round each grade to the closest authorized grade.

Examples

>>> profile = Profile(preferences_ut=[
...     [ 1. ,  0.8, -0.6],
...     [ 0.2, -0.4,  1. ],
...     [ 1. ,  0. ,  0.2],
...     [ 0.2,  0.8, -0.2],
... ])
>>> rule = RuleMajorityJudgment(min_grade=0, max_grade=5, step_grade=1)(profile)
>>> rule.ballots_
array([[5, 4, 0],
       [2, 0, 5],
       [5, 0, 1],
       [2, 5, 0]])
property scores_#

2d array of integers.

scores[0, c] is the median grade of candidate c.

Let us note p (resp. q) the number of voter who attribute to c a grade higher (resp. lower) than the median. If p > q, then scores[1, c] = p. Otherwise, scores[1, c] = -q.

property w_#

Integer (winning candidate).

Candidates are sorted lexicographically by their median (scores[0, c]) then their p or -q (scores[1, c]). If there is still a tie, the tied candidate with lower index is declared the winner.