Lessons Learned: Social Media Monitoring during Humanitarian Crises

In support of the Nepal Earthquake Assessment Unit, this research is intended to add knowledge to the "Communication with Communities" (CwC) project by assessing the monitoring of social media following the April 25 2015 earthquake in Nepal. The Assessment Capacities Project (ACAPS) document summarises the experiences of the pilot social media monitoring project and draws out key lessons and recommendations.

ACAPS carried out this social media assessment in English and Nepali from June 1 to August 27 2015 through fortnightly monitoring reports and then "a lessons learned workshop in Nepal as well as interviews and email exchanges with members of the project and external recipients of project’s reports....ACAPS is a non-profit initiative of a consortium of three NGOs [non-governmental organisations)(Action Contre la Faim - ACF, Norwegian Refugee Council and Save the Children International) created in December 2009, with the aim of supporting the humanitarian community with needs assessments."

Key findings from the report:

• "Monitoring of social media conversations... was found mainly to be useful in two ways:
• Analysing public reactions to media reports...to see which issues were widely discussed, and whether these conversations led to sustained discussion or merely short-term spikes.
• Seeing the relative prevalence of topics and identifying changes...a qualitative analysis was able to identify which sub-topics gained importance. For example, a shift in conversation from response-related topics towards reconstruction.
• Social media monitoring was not useful in breaking down needs geographically. The digital divide between rural and urban populations, as well as between different socio-economic groups, led to a bias in the data.
• The social acceptability of topics plays an important role in the scope of possible analysis: while queries related to issues such as shelter or food returned results of consistently high quality, some WASH [water, sanitation, and hygiene], protection and health issues could not be easily monitored as they were not discussed publicly.
• Social media monitoring in a rapid-onset emergency should start as soon as possible... since the volume of social media updates is largest in the first days of the emergency.
• It is vital to have qualified, computer-literate national staff who are familiar with social media, the local media landscape, the local geography, and basic information management techniques. A social media expert should be deployed on-site during the first phase to set up and customise the systems, help train staff, and increase awareness of the possibilities and limitations of analysis.
• Social media monitoring could provide significant value to decision makers in contexts where humanitarian access is poor, the information landscape is fragmented, and social media is widely used...
• Like all other forms of assessment, social media monitoring alone cannot provide a comprehensive overview of needs or opinions...."

The document gives a list of specific lessons learned on setting up and running the monitoring of social media conversations to help: improve situational awareness; identify needs, as well as opinions and attitudes; and assist with resource allocation, among other responses. It gives examples of the uses of monitoring, such as finding out about the reactions of people to news on hearing about the international and local response, such as a spike in interest on food-related issues. Monitoring can show interest shifts as time passes, such as the number of posts on destroyed homes, the posts on tents, and, over time, the shift to posts on new houses and rebuilding houses.

Examples of specific lessons include:

• Use a local team to respond in real time to immediate needs of affected people and households.
• For an early monitoring response: in disaster-prone regions, develop agreements and systems with monitoring or CwC organisations for a timeframe of several years; provide them training - including through videos customised to the event; and review and renew agreements when timeframes end. Find an expert to do remote training when a disaster occurs.
• Identify languages and evaluate platforms that support local languages, as well as tools that might be visual or map-based and function regardless of language.
• Refine search tools "beyond Boolean searches and include NLP [natural language processing] or AI [artificial intelligence] mechanisms to reduce noise."
• Understand where bandwidth is available to carry social media (not in all rural areas).
• "Be clear on the issues that are suitable for social media monitoring and analysis, according to what is acceptable for public discussion."