2017-2018 InfoVis project

Defense

Expectations

You are expected to present your visualizations, and to explain how they help answer the questions you choose to investigate. You are expected to justify the design choices for the visualizations, i.e., why is the visual mapping you choose pertinent given the data at hand and the insights you are looking to find. You are expected to send by email at blanch@imag.fr the material used for the defense.

Schedule

Friday 19 jan., room H105.

The UIS student will defend their project later in January.

Data sets

The data sets were produced by an experiment about voting systems conducted during the first round of the 2017 french presidential election. In the experiment, the voters were asked at the polling station to vote twice: first using the official rules, and then with different alternative rules.

Voting rules

The first voting rule studied is called approval voting: the candidates are all listed, and the voter can choose to approve any number of candidate, instead of only choosing one candidate. The second voting rule studied is a variant of evaluative voting: the candidates are all listed and the voter is asked to score each of them on a scale. In our case, the scale was continuous, ranging from "against" at one end of the scale to "in favor" at the other end, with the center of the scale labelled as "indifferent". At the back of the voting paper, some demographic questions were asked, and also who was the candidate (if any) they voted for in the official ballot.

Results

The experiment was conducted at the "Vieux temple" vote place in Grenoble, where three districts ("VT1", "VT2", "VT3") vote. Each file lists for each voter that participated in the experiment their answers.

The following information is given by the columns:

• #I: a unique number (per district)
• AV_NDA, AV_MLP, AV_EM, AV_BH, AV_NA, AV_PP, AV_JC, AV_JL, AV_JLM, AV_FA, AV_FF: 11 0/1 values giving wether the voter approved (1) or not (0) the corresponding candidate
• EV_NDA, EV_MLP, EV_EM, EV_BH, EV_NA, EV_PP, EV_JC, EV_JL, EV_JLM, EV_FA, EV_FF: 11 [0.0-1.0] values giving the evaluation of the 11 candidates. See below for precisions about possible values.
• VOTE: real vote using the initials, B for blank vote, NSPP ("ne se pronnonce pas") when not given
• AGE: 18/30/40/50/60/70/NSPP giving an age range: 18-29, 30-39, etc.
• SEXE: F/M/NSPP
• ETUDE: 1/2/S/NSPP, studies (primary level, secondary level, graduate)
• PROF: occupation
• AV, EV, OF: 1-5/NSPP, how they rated the approval/evaluation/official voting procedure

The continuous scale is mapped from 0.0 fo "against" to 1.0 for "in favor" and .5 is "indifferent". Voters could choose to not score some candidates (documented as being equivalent as putting the lowest score, i.e. 0.0). In this case the value is "None" in the table. As the scale was on paper, some voters choose to put marks beyond the actuel ends of the range, so there are scores below 0.0 and above 1.0.

The results are available as csv and tsv file. Please do not diffuse those files, there is no licence attached to them yet, so they are covered by copyright.

Metadata

The candidates are abbreviated using their initials, and in the table below, you will also find the hex encoding of the color traditionally used for those people, and the number of votes they got in the overall 1st roud of the election, and at the 3 vote place studied in Grenoble (both at the official vote, and in our experiment):

The candidates are traditionally ordered from left to rigth in this order (note that some "small" candidates are hard to locate on such a political scale): {NA, PP}, JLM, BH, EM, FF, NDA, MLP.

NA and PP are politically indiscernible for most people, but NA is a woman, and PP made several "coup" during the campaign (like refusing to wear a suit, refusing to check EM hand and to pose for a photo with the other candidates during the first TV debate with all the candidates). JL has been elected as member of parliament as a member of MoDEM which is at the center right, and allied to EM, but his personal positions lean more towards the right (~FF or even more right). JC and FA are mostly unknown, the first wants to massively cut taxes and FA advocated for FREXIT (France should leave EU); they can be put at the right end of the political scale.

Tasks

Here are some lists of possible questions, but you can also suggest your own. You should produce visualizations that help answer at least 6 questions, from at least 3 categories listed below.

Getting global insight:

• what is the distribution of the score used by the voters?
• how people score candidates for which they did (not) vote?
• does the range of score used by voters depend of the candidate they support?
• which candidates are approved together?
• does the number of approved candidates depend on the official vote?
• etc.

Comparing evaluation and approbation:

• are approval and evaluation consistent for each voter?
• how is the minimal note of approved candidate distributed?
• how is the maximal note of non-approved candidate distributed?
• etc.

Focusing on candidates:

• who would have won the election?
• is there different kind of score distributions allowing a classification of the candidates?
• can we order the candidate along a 1D axis (left-right)?
• etc.

Focusing on voters:

• are there different kind of voters?
• is the behaviour of voters determined by their official vote?
• etc.