Social media-driven disaster risk management

Social media have been described as a form of distributed cognition, a mechanism for understanding a situation using information spread across many minds. The interactions among people in social media are a form of collective intelligence, as they allow people to make sense of a developing event collectively. Social media users can contribute to creating a "sensor" for citizen-generated data that modelling or monitoring systems can assimilate during a crisis.

We think that social media constitute a growing data source to help improve response in the early hours and days of a disaster, when gaining situational awareness is critical and time-sensitive. However, social media platforms may not provide the functionality of summarising the information that is useful for crisis responders.

Our work aims to research how to streamline the processing of text and images extracted from social networks in near real-time during crisis. We use a combination of artificial intelligence and DevOps (Development + Operations) techniques developed in-house at JRC as well as through collaboration with top researchers and practitioners around the world.

The mission of SMDRM Task Force is to identify, understand, and address the challenges for improving adoption of non-traditional social media data for disaster management by taking a collaborative approach between researchers and practitioners in disaster management to co-design solutions.

In particular, practitioners widely recognise the potential of accessing timely information from citizen-generated data on social media, however, integrating it as supplemental information within traditional information systems of emergency management services is a complex challenge. The task force members have been actively engaged in organizing workshops to increase collaborations between researchers and practitioners to identify various challenges and discover solutions, which includes the first inaugural workshop [Lorini et al., 2021] in December 2020.

Presentations can be found here:
https://drive.google.com/drive/folders/1wuAZFnFpY7H3oY7mXKP5zZ9lPlgZMFCR?usp=share_link

You can find the workshop recordings here:
07.06: https://webcast.ec.europa.eu/social-media-for-disaster-risk-management-researchers-meet-practitioners-day-1
08.06: https://webcast.ec.europa.eu/social-media-for-disaster-risk-management-researchers-meet-practitioners-day-2

Agenda
Download the Agenda PDF (708 kB) v.05-06-2023 - New updated version

Report
Link to the Report

During our 1st workshop, "Social Media for Disaster Risk Management: Researchers Meet Practitioners", held in November 2020, practitioners, although widely recognizing the potential value of social media for accessing timely information, outlined some critical challenges for improving its adoption during crises. These challenges include validating and integrating near real-time information generated on social media with authoritative information and more traditional information systems; and preventing negative impacts from misinformation and disinformation. All participants expressed their desire to continue the discussion towards identifying new directions for research and development of systems that can better serve the information needs of emergency managers. 

In the 2nd workshop we engaged more practitioners and learnt how they used user-generated data during crisis. We understood their concerns and their doubts. However the discussions helped  closing more and more the gap between their two worlds with common data frame and procedures. We saw an emerging figure among practitioners, pushing for the use of non-authoritative data, caring for their validation and integration into crisis datasets.

From our 3rd workshop panels and presentations, it emerged how disaster response practitioners are increasingly aware of OSINT data harnessed from social media. Practitioners from emergency management communities presented several cases for which social media and AI have been deployed successfully for impact assessment. Our Taskforce showcased preliminary results from our 2022 Survey and related focus group research. As in many other ICT fields, several researchers showed their test of Large Language Models (LLMs) for Disaster Management, especially for summarization and text classification.
Two key points:
Generative AI will play a significant role in our research (text and image analysis, text production, image production, data augmentation, validation, etc.) In 2021 we pushed for having more research in practice and more practitioners in research. Therefore we were delighted to see that several initiatives successfully supported real-time disaster response.

• Hemant Purohit, George Mason University, USA (Co-chair)
• Steve Peterson (Montgomery County Community Emergency Response Teams, USA (Co-chair)
• Valerio Lorini, European Parliament, Luxembourg (Co-chair)
• Cody Buntain, University of Maryland College Park, USA
• Carlos Castillo, University of Pompeu Fabra, Spain
• Jorge Gomes, Virtual Operations Support Team, Portugal
• Amanda Hughes, Brigham Young University, USA
• Alexandre Penha, National Civil Protection Authority, Portugal
• Nelles Scholiers, European Commission, DG ECHO, Belgium

SMDRM Task Force is a consortium of researchers from universities and practitioners from disaster management organizations around the globe, facilitated by researchers from the Copernicus Emergency Services at the European Commission's Joint Research Centre.

You can reach out to any of the following co-chairs: Dr. Hemant Purohit (hpurohit@gmu.edu), Steve Peterson (stevepeterson2@gmail.com), Valerio Lorini (valerio.lorini@europarl.europa.eu)