How can digital technology change civic debate?

Thousands of people march in France every year, "yellow vests" or not, to demand that they finally be heard a little more in public debate, and above all taken into account. What if their voice could be enhanced and multiplied tenfold by digital tools?

Jean Sallantin, University of MontpellierAntsa Nasandratra Nirina Avo, University of FianarantsoaMathieu Lafourcade, University of Montpellier and Véronique Pinet, University of Montpellier

This article is published as part of the upcoming Fête de la science (October 5-13, 2019 in mainland France and November 9-17 in overseas and international locations), of which The Conversation France is a partner. The theme of this year's event is "Telling the science of tomorrow, imagining the future". Find all the debates and events in your region on the Fetedelascience.fr website.

Our experiments with digital debate tools have shown that formalization and the use of artificial intelligence (AI) give debates a rationality that enhances their effectiveness.

Digital debate will thus be a factor of intervention in democratic life if it enables a greater number of citizens to participate directly in public debate, and if it avoids the biases currently encountered in face-to-face debate and discussion sites.

The pitfall of traditional debate

As it is difficult to follow a debate containing the contributions of a very large number of participants, it would then be desirable to develop an artificial intelligence that assists everyone in studying and pursuing the debates.

Rationalization, formalization and the use of AI in the digital debate must show that it is an effective and controlled auxiliary. We are going to introduce this approach and its tools by illustrating it on the debates of the Terre de convergence eco-meeting, which took place from August 13 to 18 in Attuech (30).

The art of calculation

Argumentation is linked to logic, the "art of thinking correctly", to rhetoric, the "art of speaking well", and to dialectics, the "art of dialogue", but is not yet "an art of calculating well".

However, two computational operations - indexing and classification - are used to structure the arguments of a debate. Indexing an argument means determining which of all the terms in the French language are associated with it. The classification of arguments groups them into sub-classes characterized by the common terms indexing them. The artificial intelligence methods used here treat all arguments equally. They reveal the incompleteness of an argument by grouping arguments that appear impertinent to the participants, encouraging them to complete the argument by adding new arguments or by indexing more precisely the arguments already present; there is no possibility here of deleting arguments.

A classic scientific approach is used here for debate arguments. Scientists correct their arguments by reporting on the work of their peers and guiding their doctoral students, but they cannot systematically correct the arguments of a science/society debate if they have no way of identifying those that specifically concern their knowledge.

The mathematical formalization of digital debates focuses in particular on these processes of argument indexing and classification, as it must show why the techniques used are capable of effectively guiding participants to specify and complete the set of arguments.

The mathematical formalization of digital debate uses topos as the basis for artificial intelligence methods used to structure debates.

The terre de convergence debates

The Terre de convergence eco-meeting is a project supported by the French Ministry of the Ecological Transition. Some 2,000 people and 200 associations took part. The aim of these meetings and the use of collaborative digital tools is to initiate networking between local transition players.

The terre de convergence debates focus on issues concerning energy, the environment, waste, food, solidarity, etc.). They take into account the dimensions of personal development and collective life, the local and the global, and the political, from real citizen participation to the permanent questioning of representatives. The digital debates have prepared and will continue the in-person debates by broadening the audience.

Debates mark the willingness of citizens to take part in decision-making on territorial transition, a moratorium on 5G,collective housing, and Energy transition: context and challenges.

Participants also looked at more remote themes, such as the situation of the so-called "first peoples", a theme made even more topical by the fires in Amazonia. And last but not least, collaborative digital technologies.

To understand the intervention of digital technology and AI in the digital debate, we invite you to a debate on a text by Edgar Morin.

Survivalism as an example

To illustrate the intervention of AI, let's consider an argument made in this debate concerning survivalism.

The IDEFIX indexing software is based on a web-based data acquisition program, JeuxDeMots. IDEFIX will be used to index this argument from all the French language terms associated with it in the context of the "Terre de convergence" debates, enabling the participant to see how his or her argument relates to the other arguments in all the debates indexed by one of the associated words.

Let's show the sequence of operations that will be used to index the argument as belonging to the theme of survivalism.

Argument text = "Survivalism is the end of collective living and a return to small groups that exclude life in society".

Terms of argument = * survivalism * end of communal living * back to * small * group * groups * small group * excluders * excluding * communal living

Context in the eco-meeting (max=30) = * end of life in society | 3.257 * group of people | 2.585 * natural disaster | 2.242 * health crisis | 2.155 * collapse of industrial civilization | 2.044 * survivalism | 1.867 * economic crisis | 1.702 - th=0.5 - min=10 - max=20 - nb ideas=1326

"Survivalism" was not initially a term present in the context of the argument. So a reinforcement learning technique is used to move it up the ranking along with other terms that will move up along with it.

Terms to index the argument: group - end of life in society - group of people - natural disaster - health crisis - collapse of industrial civilization - survivalism - economic crisis - life - survivalist -

Eliminating erroneous coincidences

To further determine the completeness of the argument, WEBRA, a machine-learning software package registered by the CNRS, performs a formal concept analysis, signalling coincidences between terms indexing the arguments of the current debate. When a debate has few arguments, it produces many erroneous coincidences, such as the following: petit_groupes_finis+ and survivaliste+. We then know that we need to enrich the set of arguments in order to eliminate this erroneous coincidence. In this way, they provide effective entry points for a rapid deepening of the debate.

Many public debates concern issues such as the territorial transition due to climate change, they are widely open to the public and they involve linking the arguments of citizens, associations, politicians and scientists.

Such debates require artificial intelligence to compensate for the human impossibility of reading, indexing and classifying everything. However, this AI's methods must verify mathematical principles making arguments equal in their treatment. What's more, this AI must prove its ability to learn from arguments, so that it can appropriately solicit contributions from debate participants (i.e., it doesn't intervene in the debate).

Jean Sallantin, Emeritus Research Director at the CNRS, University of Montpellier; Antsa Nasandratra Nirina Avo, Research-teacher Mathematical modelling of the argumentation, University of FianarantsoaMathieu Lafourcade, artificial intelligence, automatic natural language processing, automatic reasoning, inference, machine learning, University of Montpellier and Véronique Pinet, Professor of Philosophy, University of Montpellier

This article is republished from The Conversation under a Creative Commons license. Read theoriginal article.