A Semi-Automated Content Moderation Workflow for Humanitarian Situation Assessments

Link D, Ling J, Hoffjann J, Hellingrath B

Abstract
Although online social media has been recognized as a source of information that is potentiallyrelevant for humanitarian organizations, it remains to demonstrate positive impact. The authors arguethat relevant information isn't yet incorporated effectively into decision-making because the keyrole of humanitarian situation assessment experts and their methodologies hasn't been sufficientlyrecognized and incorporated into information systems design. In particular, the authors focus onthe content moderation process (i.e. on examining, correcting and enriching data and controlling itsdissemination) and argue that existing systems, which often follow a human-is-the-loop approach,either lack automation support or flexibility. In contrast, they present an interactive, semi-automatedcontent moderation workflow and an instantiating prototype that follows the human-is-the-loopapproach and centers on assessment experts. The evaluation of the new system practitioner interviewsand serious games suggests that it offers good compatibility with experts' work practices, moderationquality and flexibility.Keywords

Keywords
Disaster management; situation awareness; machine learning; information systems