Joonsuk Park of the Cornell University Department of Computer Science, Dr. Joan-Josep Vallbé of the Legal Information Institute at Cornell University, Professor Dr. Claire Cardie of the Cornell University Department of Computer Science, and colleagues, will present a paper entitled Facilitative Moderation for Online Participation in eRulemaking, at dg.o 2012: The International Conference on Digital Government Research, to be held 4-7 June 2012, at the University of Maryland, College Park, in College Park, Maryland, USA.
Here is the abstract:
This paper describes the use of facilitative moderation strategies in an online rulemaking public participation system. Rulemaking is one of the U.S. government’s most important policymaking methods. Although broad transparency and participation rights are part of its legal structure, significant barriers prevent effective engagement by many groups of interested citizens. Regulation Room, an experimental open-government partnership between academic researchers and government agencies, is a socio-technical participation system that uses multiple methods to lower potential barriers to broader participation. To encourage effective individual comments and productive group discussion in Regulation Room, we adapt strategies for facilitative human moderation originating from social science research in deliberative democracy and alternative dispute resolution for use in the demanding online participation setting of eRulemaking. We develop a moderation protocol, deploy it in “live” [U.S.] Department of Transportation (DOT) rulemakings, and provide an initial analysis of its use through a manual coding of all moderator interventions with respect to the protocol. We then investigate the feasibility of automating the moderation protocol: we employ annotated data from the coding project to train machine learning-based classifiers to identify places in the online discussion where human moderator intervention is required. Though the trained classifiers only marginally outperform the baseline, the improvement is statistically significant in spite of limited data and a very basic feature set, which is a promising result.
For the full text of the paper, click here.
Thanks to Dr. Vallbé for granting permission to post the abstract.