Powering up E-DELIBeration: towards AI-supported moderation

Independent research group funded by the German Ministry for Education and Research (BMBF)

Hiring! Have a look at the job announcement for 2 PhD ad 1 Postdoc position

What is E-deliberation? 

E-deliberation enables citizens to participate in public decisions through large-scale online consultations about topics of immediate relevance for a community. Such consultations take the form of discussions followed by a voting, as it is typical of a direct democracy setting. A typical direct democracy scenario is participated budgeting: a city has a budget of 100k EUR, and a consultation is held among citizens, concerning how this amount should be used. Citizens are then asked to point out problems, propose a number of solutions, and engage in debates to converge to a decision (the deliberation on the budget).

E-deliberation can be seen as a “digitally augmented” instance of deliberation in a direct democracy setting, where Internet brings in the possibility of reaching out to a vast public, thus favoring emergent collaborative phenomena. Unsurprisingly, in the recent years, several dedicated online platforms have been designed by specialists to support the different steps in a deliberation process, and at the same time existing exchange platforms (e.g., wiki, forums) have been employed for debating purposes, in the context of deliberation processes.

Powering up E-deliberation with NLP:  how?

Discussion targeted at decision making is challenging. We all experience this when we need to take light-hearted decisions affecting small groups of people (which restaurant should we go to tonight?). Let’s just imagine how this scales up to decisions that are relevant for a large group of citizens (the city has a budget of 100k EUR: how should we invest it?). Moreover, having a fruitful discussion on the Internet is even more of a challenge, for many reasons we are aware of as social media users. This is where moderation comes into play: Can we help discourse participants by pointing out commonalities in their positions? What can we do to ensure that all discourse participants get to be heard (also those that are less “vocal”)? 

From perfect NLP debaters to reasonable NLP moderators 

The aim of E-DELIB is to employ the NLP artillery in a strong synergy with Political Science and Decision Making to empower e-deliberation with NLP-supported moderation.  The end product of the project will be the prototype of a NLP-moderator to be integrated into existing e-deliberation platforms to keep argument exchange optimal as required by the best-practices in digital deliberation. 

A project like E-DELIB will have to face multiple challenges, and accordingly the extent of its contribution to the respective fields will be manifold. We will deal with argument mining on an underexplored textual genre such as deliberative data and we will cope with the scarcity of annotated data for our target phenomena. We will integrate argument mining solutions into discourse network analysis, and use network optimization methods to detect the need for an intervention fo the moderation. Finally, we will redefine the notion of success in argumentation away from the game of the best debater and towards a more realistic and societally grounded definition: a successful argument is one which contributes productively to a collective decision process.

In the spirit of the High-tech strategy 2025, we aim at producing technological innovations or suggesting ways to exploit the existing ones, with the goal of strengthening social cohesion by triggering dialog and discussions among the citizens. More efficient tools to support such large-scale discussions will help citizens overcome frustration which often accompanies “real- life” online discussions and come up with better decisions (better, because shared, and because they originate from a cooperative decision making process) which will make them more and more confident in their decision making capabilities, and power.




To the top of the page