Skip to main content

What will France do with 'National Debate' data?

Caroline Blumberg, POOL/AFP | French President Emmanuel Macron (2nd R) meets with single mothers during a visit to a professional integration center as part of the 'Great National Debate', in Bordeaux, western France, on February 28, 2019.

In an attempt to calm Yellow Vest protests and inform future policy, French President Macron launched a “Great National Debate”, a combination of in-person exchanges and online surveys. But what will happen to the mountain of data it has produced?

Advertising

It all started in mid-January, with a website and advice for anyone wanting to organize local sessions. Le "grand débat", the massive, nationwide debate launched by President Macron in the wake of the Yellow Vest protests, was supposed to touch on four main themes: ecological transition, the economy and public spending, democracy and citizenship, and public services.

“It was a bet by the government that had no guarantee of succeeding. And it has. Over a million people have turned out or have written and participated directly in this debate,” said Guy Saez, political scientist and research director emeritus at the French National Center for Scientific Research, interviewed by FRANCE 24.

But without any follow-up from the government, the participants risk feeling even more marginalized. “If the people who have participated don’t see anything concrete come out of this, they will feel duped and tricked. The government must come out of this debate with concrete proposals", Saez said.

Since the beginning of the National Debate, some reforms have been put on hold and President Macron is expected to announce new initiatives, including potential citizen’s referendums, in mid-April at the earliest.

The scale of the debate poses a challenge

But the grand débat’s scale poses a challenge in terms of processing the mountain of information collected. So far 6,500 of the 9,250 planned debates have taken place, including a recent one livestreamed on Twitch to try to reach a younger public. They will continue until March 18.

Transcripts for each of the meetings are available on the grand débat website, where citizens also have the option of answering multiple-choice surveys on each of the four main themes selected by the government. So far, over 300,000 people have answered the open-ended questions. Roughly 10,000 mayors' offices have also made physical “grievance” books available to citizens to inscribe their concerns.

Prime Minister Edouard Philippe has recognized “the difficulty of utilising this quantity of data”. As it would be too much for anyone to read, the French government has decided to rely instead on a combination of human and artificial intelligence.

Material analyzed by polling institute OpinionWay

La Biblothèque nationale de France (French National Library) and a consortium of companies have been charged with referencing and digitizing the “grievance” books, while the analysis of the website materials has been outsourced to OpinionWay, a big data polling institute, and Qwam, a company that specializes in text analysis.

The roughly 300,000 responses to open-ended questions will be fed through Qwam Text Analytics software, an automated text analysis program its makers say will identify and sort issues brought up by respondents. The objective, according to OpinionWay’s director of opinion and politics Frédéric Micheau, is to sort each comment into categories like “fiscal policy” and sub-categories like “lowering or raising taxes” in order to come up with a quantitative reports of the number of people concerned with each issue.

“There will be systematic human intervention to verify the coherence of the results and to make sure that the participants' contributions were well understood [by the algorithm],” promised Micheau. But the proposal has been met with considerable scepticism and objections, both from a practical and philosophical point of view. Can an algorithm process complicated ideas? What about irony or sarcasm, for example?

Can a computer understand nuance?

Nicolas Cori, co-founder of the news website Les Jours, wrote a scathing takedown of the artificial intelligence methods proposed for analysing all the data and documents. Would a computer understand that a response mentioning "dough", "tomatoes" and "mozzarella" was about pizza? he asks.

One of the survey questions is, “For which policies or initiatives would you be ready to pay higher taxes?” One answer offered: “For the ecological transition, if I felt the money would be spent in an equitable and just manner.” Or another: “There is already a lot of money handed over to the government that could be better managed and distributed to civil servants on the ground, like police officers, teachers, and hospital employees who are indispensable but often ignored or poorly paid compared to administrators.”

In light of these answers, one might wonder in which category they would fit: environment? Government spending? Social justice? The labour market?

“Some responses that make sense could be passed over by a machine if they don’t have the anticipated key words, whereas a human would understand what they are referring to,” reads a statement by the group called CodeForFrance, which runs a website that tags responses to the open-ended questions by hand.

Critics complain about the National Debate's methology

The government appears aware of the limits of AI. The minister of higher education and research, Frédérique Vidal, called on the national research agency (Agence Nationale de la Recherche) to dig in to the open-source data.

But the criticis, including some who are involved in running the grand débat, have also targeted the methodology of the debate itself. Former President of the National Commission for Public Debate (CNDP), Chantal Jouanno, who had to resign from piloting the grand débat after her salary was made public, criticized the limits on the debate imposed by the government.

“The principle of a public debate is not to ask French people questions. It’s the French people who should be asking the questions, expressing themselves, and saying what is important to them,” she said on LCI. The government’s initiative is more of a public consultation, she believes, with questions that are too pointed.

One of the four designated “guarantors” of the debate, Pascal Perrineau, held a similar view.

“From the beginning, we said that these closed questions were not a good way to stimulate debate,” he said at a press conference on February 14. The respondents are not representative of the French population, he added, which weakens the scope of the data gathered from the multiple-choice questions.

Qualatitative vs quantitative analysis of the National Debate

But quantitative analysis isn’t the point, said Saez, who attended a number of the local debates in person. Even if he thinks it was a mistake for the government to impose 34 questions, he said people ended up talking about anything and everything.

“It’s an unprecedented exercise. It’s very new and innovative. We’ve done experiments in participative democracy on a smaller scale, but this is exceptional. I think it’s important to mention that it’s not just citizens who are eagerly waiting to see what comes of this. It’s the politicians, too. It will fundamentally change the role of political parties. They won’t be able to continue as if [the National Debate and the Yellow Vest] movement hadn’t happened.”

Some Yellow Vest activists have criticized the National Debate as a cynical tactical move on the part of the government to try to appease them. A recent Ipsos poll found that only 38 percent of the French population was confident that the president would take the comments and feedback into account in his policy decisions.

Still, said Saez, “there will be a "before" and an "after" the National Debate. It will change the role of parties, of citizens. It will be a total regeneration of the democracy.”

Page not found

The content you requested does not exist or is not available anymore.