How hope changes the information ecosystem – understanding rumors

The first time I thought about creating the rumors tracking methodology I was in Liberia. I had a huge problem: I needed to understand the information ecosystem of the country, work with local media and provide information to the local population and the humanitarian community. In a month. And fight Ebola. Yeah..we have all (more or less) been there.

The rumors tracking came out because everyone I was talking to had a theory about Ebola, and I had listen to so many theories in only one month, that I decided to take that as a sample of my population. Journalists, politicians, colleagues, the drivers, humanitarian organizations staff..everyone had it’s own theory about everyone else.

When we started the work with UNICEF, which offered us to use the short code that they were already using for their U-Report project, and to access the data via their Rapid Pro platform, I met the Head of Section for the UNICEF C4D Unit in Liberia, Rania Elessawi. One beer (or more) later, the project was already happening. She had been struggling to help her teams on the ground, the one risking their lives going door to door to sensitize communities about Ebola, because their reach, for as wide as possible, was never enough. I had radios and content. She had the field network and the expertise.

This is how we ended up creating a database of 2,500 people in between religious leaders; women leaders; journalists; community health workers; social mobilization officers; ETU staff; red cross volunteers and others. This network of people listening and talking to communities on the ground was now able to report rumors and receive answers over SMS.

On the ground, the information given to them was also fed to the local media, so that all the major information sources and trusted sources would deliver the same message and reinforce each other. The all point of tracking rumors was the we had the network and the ability to stop them on the ground. The rumors tracking was a necessary passage to achieve change in perceptions, but also change in services provided – in this case making sure that communities were respected and involved in the decision making processes about themselves.

As the first rumors tracking mechanism, in that situation we were experimenting really and it worked. We tried, and assessed and adapted ending up with a self sufficient loop where religious leaders, community leaders, women leaders, youth leaders, local authorities and health authorities were able to work with “their” local media to create content but most of all to engage in a conversation with the local community.

As the emergencies go by, Internews replicated the system in Nepal, in Haiti and in Greece. This last one was a very interesting project and probably the most sophisticated we have ever done. In Greece the rumors we were tracking had a totally different dimension to it, due to the fact that the population was indeed moving, and the rumors were moving with them.

As we started our initial assessment, we realized that one of the most important component of the information ecosystem, trust, was not necessarily obvious. Who would these people trust now that they were “on the move”? As we did in Liberia, we were looking for the big numbers: how can we build a big network of trusted sources? Especially when the trusted sources are moving away? Well, we couldn’t.

So, we created a smaller network, made of a self selected group of migrants and refugees in almost all of the locations of camps, arrivals or other form of settlements. The network was also somehow changing, depending on the day in which our staff would visit that specific location.

What we discovered than was something that changed the way we are looking at rumors: while it was possible to start identifying rumors by their spread (or their statistical representation in terms of numbers), this was not necessarily the best way to really address rumors. Because when a rumors is spread or representative, it is already too late.

So, while it will always be important to stop the rumors that are already spreading, the one that are not out there yet, those are the one that we have to look at. For example:

If 2,000 people have been locked in a island in country X, and they have no idea what is gonna happen to them; why they are there; what’s next; etc, and ONE single person says that he thinks there is a boat down at the beach ready to bring them to country X much time do you think it will take for that rumor to spread? and most of all, what could be the possible consequences of that rumor for the people believing it?

The issue here is that to be seriously addressing rumors we need to stop thinking about “data validation”, “verification” and “statistical representation”. Those concepts do not really apply to something that you need to prevent from happening. The only way to look at those rumors is not by their value as of today, but by thee potential negative impact on the affected population combined with the likelihood of that rumors to be believed.

And here is the second paradigm shift: when people have hope, they will believe anything.  As very well explained here:

In Greece we realized that rumors do not necessarily spread following the normal information ecosystem matrix, where trust is one of the strongest variable in defining information sources. Those with stronger trust normally, are the most important influencers in the conversation. But when people have false hopes, the level of trust does not matter anymore so much – people can be strongly influenced by a non-influencer, or a temporary one. For example one single refugee saying that the border with Bulgaria is open and if people run there they will be able to leave Greece it’s enough to get hundreds of people to run to the border.

In this case, what we know about the information ecosystem, and what we know about countering rumors, needs to be mediated by a different variable: hope.

This is why when we look at rumors, we do not look at it as any other piece of information, like perception surveys (70% of people liked the water pump, therefore the water pump is good) – we need to look at it in its singularity: one rumor, shared by one person at the right time, and with the right people, has far more potential to negatively impact the local population or the humanitarian community than a rumor that 50% of the population agrees on.

This is what makes a rumors tracking project successful: the ability to understand and recognize the potential impact of a rumor, it’s likelihood to be believed, and the ability to understand what piece of information is needed to proactively correct what was not understood without spreading the rumor (it goes without saying that something like “someone said XX but this is not true” is not a good way to address rumors).

This is also the secret of really using an Information Ecosystem Assessment in a meaningful way: understand how the system adjust itself when variables are introduced and how that changes the way communities are using, understanding and acting on information received.

And, as the last point: tracking rumors makes no sense whatsoever if you are not able to act on them, and if you do not understand or have the power to deliver the right information, at the right time, to the right people. To be able to do so, you will have to abandon the all idea of statistical representation and verification, and you have to focus on the potential and likelihood of a rumor to have a negative impact. To do so, you have to engage communities and make sure they are involved in actually analyzing the rumors. Yeah, I know. You actually do have to work with the communities themselves to make sure you know what the potential is.

Also, Rumors are NOT counteracted with MESSAGES. They are counteracted by DIALOGUE and the reason is that rumors changes continuously and are strictly related to very volatile preconditions: emotions and information flows.

For more information about rumors, CDAC has just published their Rumors’ Guide here.

And if you are seriously thinking about doing Rumors Tracking, feel free to ping me anytime 🙂


Hate Speech on Facebook: the causation effect

I have to say, I have never been a huge fan of Facebook, but as everyone else, I use it. I came to terms with the fact that if I do not want my data to be owned by them, I should not use it. I also came to terms that there are TONS of very good policies with respect to data privacy and ownership that we really SHOULD enforce and apply, not just to Facebook but also to a lot of other companies (starting with mobile and internet providers, if you ask me).

But, this post is not really about Facebook. This post is about how mixing correlation and causation, when it comes to analyzing social media and hate speech on social media, can have very damaging effects and why we REALLY should try to avoid it, even if it makes for a pretty catchy headline.







In January this year this article was published and circulated on Twitter and elsewhere , followed by some discussion about how social media is being used for hate speech and incitement to violence. The information in the article is not news to anyone that has been looking at social media and hate speech in conflict settings before: from the amazing work done by iHub Research with Umati; to the work done by Holocaust Museum Fellow Rachel Brown to produce Defusing Hate: A Strategic Guide to Counteract Dangerous Speech and of course the amazing work one by the PeaceTech Lab on creating a taxonomy of terms used in hate speech in South Sudan.

The issue of Hate Speech and online content is really a tricky one and while it has been flagged quite some time ago – a big event that triggered a lot of conversations about this was, in the era of predominance of non-digital tools, the events of the 2008/2009 post election violence in Kenya, where mobile phones and radio were used to spread rumors; incite to violence and to perpetrate hate speech.

One of the great work that was done following the Kenya events was the research implemented by iHub Research under the Umati project. The Umati project had as objective to figure out what type of hate speech on social media were happening, and if they were in fact triggering actions on the ground.

The questions I am trying to highlight here is very simple: is social media hate speech content the cause OR the effects of actual violence happening on the ground? The article mentioned before makes the point that social media are being used in South Sudan to trigger and actually fuel violence on the ground. What the article does not say, but implies, it is a concept that has many more repercussions on the way we see social media hate speech and violence: that is that if we stop hate speech online we indeed will be able to lower the level of violence in the country.

In fact, this is the biggest problem I have with this rationale: considering hate speech on social media as the cause of a problem (violence on the ground, and in the South Sudan case, a genocide), as opposed to the symptom of the widespread violence already happening in the country.

But is this the case? What data do we have to say that there is a causation effect in between hate speech on FB and violence on the ground? And this is the problem: the data we have proves the contrary.

Let’s start again with the Umati project. Two main findings in fact prove the contrary of that.

  1. Hate Speech online does NOT mirror the violence on the ground: “Despite this clear increase in the volume and severity of hate speech incidences online during and after elections, there was little reported violence on the ground as compared to the 2007/8 post-election period. Notably, had the trend in Umati data been used in isolation to predict the possibility of post election violence, the prediction would have been that there would be violence, given the highly vitriolic data Umati came across in February and March (examples provided in Appendix II). The peaceful election outcome suggests that there are overriding factors that can strongly contribute to the serenity of an election period.” (from here)
  2. When there is a relationship in between hate speech online and events on the ground, it is indeed a reverse relationship: ” We have observed that most dangerous speech occurs as a response to events that happen on the ground. These events come to the attention of the commenters mainly through traditional and online media reporting. The media therefore, plays a vital role in highlighting what topics are discussed online and suggests that responsible reporting by media houses, whether in newspapers, online or on radio, may shape the type of conversations that form around those reported topics.” (from here)

When I spoke with the Umati research team, back in 2013, they explained to me that they could not prove that hate speeches on line did in fact translated into specific actions on the ground. And, if you know a little about social media and how they works this makes totally sense: social media are a mirror of what people and communities are talking about. They are indeed a barometer of the temperature of that community, meaning that they are the arena where conversations happen (freely and in between peers) and in this way they are a way to look into a certain community and its dynamics. In fact, if we assume that Social Media like Facebook have an automatic way to influence real actions on the ground we would have never came out with the term slacktivism.

But there is more to this: while I do agree that social media hate speech can have an influence in the conversations that people have, and therefore increase the level of acceptance of violence, and therefore increase the overall outcome (meaning people are more likely to commit violent acts), I think we need to make sure that we do not end up curing the symptom and disregarding the real disease.

If we focus all of our attention on hate speech on social media, and specifically on Facebook, we run into the risk of not understanding why and how violence on the ground is happening, and with this, not really focusing on the fact that real violence, committed by real people, is happening, and that the role of media (see above from the Umati reports) is here key in making sure that events and facts are indeed reported accurately and as such cannot be used as a source of hate speech.

This is all to say that we need to ficus on the right issues and try to solve the right problems:

  • If hate speech is happening on social media, we have to look at the reality on the ground to see what is triggering it before we assume that they are the cause. The fact that two things happens in the same time or follow the same trajectory does not mean that they are positively correlated, as shown below (real data):

  • Hate speech is based on facts: facts are used to trigger a certain vision of the other person/group (they have done this and therefore they are “less of a human/person/group). If we want to fight against hate speech, facts needs to be set straight (and with this I mean REAL facts, not alternative facts);
  • Demonizing social media platforms because they are a place where hate speech happens is not helping, neither us, nor social media platform, to learn how to address the problem. Furthermore, it may actually translate into repressive or censorship-like laws under the umbrella of anti-hate speech regulations, something that we have seen before can be easily used to repress freedom of speech (a very good reading on the subject is here)
  • When hate speech on social media are rising. there is something MUCH more effective that we can do to stop it, which is to fight is here it is. Back in 2011 in Egypt, social media activists were targeting social media pages of the regime by blasting them with real facts and counter arguments (peacefully) until the owners of those pages were basically giving up the pages themselves.

So, all in all: engagement is the way you counter hate speech. And you can use Facebbok for hate speech as much as you can use it for engagement, you just need to look at the full picture. Go where people are, and use their means of communication, rather than demonizing them.



Combating Rumors About Ebola using SMSs

[This article was first posted on Medium on March 27, 2015. This article is an account of the project I managed as a Country Director in Liberia during the Ebola crisis]

When misinformation is a case of life or death, aid workers and communities need an ear to the ground.

The rumor began with a fever. On March 4th, 2015 in a public school in rural northern Liberia on the border with Guinea, a child was diagnosed with a high fever. Following protocol in a region where Ebola cases remained a concern, the school called an ambulance to take the child for treatment. But when the ambulance arrived, students began to panic and flee the campus. Parents began to frantically connect by phone and a rumor quickly spread to the larger town of Sanniquellie, 35 km to the south.

The rumor?

“People are vaccinating children in schools and the vaccinated children are then taken by ambulance and hospitalized.”

Within the hour, parents throughout the region were rushing to local schools and removing their children.

Luckily, in this case the rumor was short-lived. Information about the panic was quickly communicated to local radio stations who set the record straight, calming the anxious population. The broadcasts were followed up by local health teams who made school visits. The rumor was stopped before it reached the highly populated Ganta school system, another 35 km south of Sanniquellie.

Replicating this system to refute future rumors is not so simple. Rumors spread quickly and generally through word of mouth, SMS and social media — channels that are hard to track and monitor. Rumors that start far from the capital, where most aid organizations are based, can grow out of control very quickly.

Rumors can kill

What is now clear to healthcare organizations working on the ground in West Africa is that the Ebola epidemic has been driven as much by misinformation and rumors as by weaknesses in the health system. It is common sense that information is a critical element in combatting disease, particularly when contagion from common social practices, such as bathing the corpses of the deceased, were central to so much of the early spread of the disease. But in the context of a massive disease outbreak, when hundreds of international organizations and billions of dollars flood into a region whose fragile infrastructure has been damaged by years of civil war, information dissemination becomes a powerful challenge.

In the Ebola outbreak, the international community quickly created a series of wide-scale social behavior change communication campaigns, a typical approach in humanitarian aid. The result was that local populations were bombarded with massive but poorly-coordinated blasts of messaging on billboards, in print, on radio and TV, through health outreach workers and community organizations, via SMS and call-in hotlines.

One thing that was routinely missed in this chaotic information environment was that one size does not fit all.

Local beliefs and attitudes were in some cases a serious impediment to people acting on the messages. Some communities believed, for example, that the bleach sprayed by health workers to sanitize the environment was the government spraying the virus and spreading disease. In other cases, people believed the disease was the result of black magic. In this context, misinformation could spread quickly and along with it, infection and death.

Even for those communities where generic messaging was absorbed, or where there was hope that Ebola was contained, significant challenges remain. People needed ongoing information about Ebola prevention and preparedness for future epidemics. They were hungry to hear about situations on the borders and the resumption of trade and commerce. Many parents were fearful of sending their children to schools that were once used as Ebola holding centers and were resisting regular vaccination schedules for measles or polio because they feared Ebola vaccinations.

”DeySay” SMS

In this complex environment, aid workers were grappling with how to get a better handle on rumors, and how to refute them quickly. In partnership with the Liberian National Red Cross Society, UNICEF and Project Concern International, Internews developed a simple but critical new tool. DeySay SMS (“Dey Say” refers to how people speak about rumors in Liberian English), to detect and manage rumors in as close to real-time as possible.

DeySay began with an SMS short code, provided by UNICEF free of charge to hundreds of health workers, NGOs and volunteers on the ground throughout Liberia. When anyone connected to the system becomes aware of a rumor, they texted it via the short code to a central coordination hub in Monrovia.

DeySay System Architecture. Credit: Internews

The information was then collected, analyzed for trends, and disseminated to local media partners in the field with details about the rumor so they could stop its spread. Once the system was fully functional, aid workers and social mobilizers in the relevant regions were put on alert so they could go door-to-door to calm anxieties and correct misinformation.

In conjunction with the rapid response system, DeySay also produced a weekly newsletter for local media throughout the country and partners on the ground. The newsletter highlighted trends in rumors and their geographic locations, and helped identify the most critical rumors at any given time. The newsletter also offered insights for local media into information gaps and challenges around Ebola and health reporting.

Information Saves Lives Newsletter. Credit: Internews

DeySay SMS offered both rapid response to rumors and, over time, collect and housed valuable data that could be analyzed and used to train media and health workers so they could be more prepared the next time the region experiences a crisis. Knowing which areas were prone to rumors, where the pockets of resistance are and how to truly communicate in ways that people could understand was critical not only for combating epidemics but for creating a healthy recovery.

The DeySay project was an Internews project in collaboration with UNICEF, Liberian National Red Cross Society and Project Concern International. The project is funded by USAID under the Health Communication Capacity Collaborative (HC3) project.

The Humanitarian Information Dashboard 7 months later

[This blog post is an abstract of the “Next Steps Report” produced by Internews and Aptivate in August 2015]

Seven months later, I am really excited to talk about what we have done with the Humanitarian Information Dashboard (HID). The system was designed to track and analyze community views and humanitarian related data during emergencies. The platform aggregates feedback from affected communities to give humanitarian responders unprecedented insight into community information needs and make informed decisions that save lives.

The HID supports Internews’ belief that information is an essential part of humanitarian relief and helps Internews respond to issues of information overload during a time of crisis. The main goal of the dashboard is to create:

  • greater trust and security around information;
  • an improved efficiency of response efforts during emergencies;
  • identified knowledge gaps within communities affected by a humanitarian disaster;
  • more targeted information delivered to communities affected by a humanitarian disaster.

Review of developments to date

A great deal of useful, valuable and complex work has been completed during the first phase of the HID development, done between February and August 2015 with Aptivate, an established NGO that believes in the power of knowledge and communication to alleviate poverty, suffering and conflict, and in the right of every individual to inform and be informed. Aptivate is dedicated to developing ICT services that facilitate communication for unconnected communities, empowering ordinary people across the developing world to improve their lives and has been selected by Internews to work on the this project. Associated with the developments achieved to date, there has been some important learning that has strengthened Internews and Aptivate with regards to their understanding of what a successful HID comprises and what opportunities and challenges exist for future development.

Screen Shot 2015-10-10 at 2.20.06 PM

What has been delivered

  • A working prototype of the HID for use by Internews in Liberia, designed for low bandwidth and with a responsive design interface (key requirements of the HID user environment).
  • Supporting documentation and deployment support in Liberia.
  • Architectural design and technical implementation of the REST API – a significant foundation for future developments of the HID based on robust scale and connectivity infrastructure.
  • Inception report and a more refined scope for the HID based on iterative discovery, development and testing with users.
  • A comprehensive evaluation of existing practice, data streams and software tools used in gathering information in global humanitarian response efforts.
  • Refined vision and implementation scenario modelling from field work, personas and user stories, user consultation and participatory discovery processes.
  • Huge insights into field office reality and field office working within which any HID solution needs to operate.
  • A more detailed understanding and documentation of future development options in the form of a features and requirements backlog and associated future planning.
  • Field user testing and a more defined understanding of digital literacy, functions and tasks of the target users.
  • Enhanced capacity of Internews staff based on training and support on agile product owner role and user testing.
  • Scope of the project refined to focus on Liberia office needs – that has been an important development – this is a big delivery in itself. Embedding in the office has been crucial.
  • An effective project management and implementation team working environment between Internews and Aptivate staff. With improved knowledge and decision making around the project as a result.

What is unique

  • The REST API is a unique contribution to this project and the wider information for development community.
  • The demonstration prototype of the HID is a unique and promising application. It is built using Open Source technology that has the potential for adoption and adaptation by others working in this area.
  • HID end user functionality that allow tasks and actions to be completed that it was previously not possible to do.
  • User testing and field usage of the HID prototype that has allowed previously abstract ideas to be field tested and trailed in a field deployment scenario.
  • Strengthening and enhanced dialogue around “communication and information as aid”
  • An understanding of how the HID can facilitate reciprocal and active engagement between humanitarian response efforts and affected communities, and consequently encourage responders to hear and respond to the most urgent questions from the community and that the community responds and provides feedback to relief agencies to ensure its efforts are relevant and actionable.
HID REST API Development

Significant effort was used in developing the HID REST API. A REST API is a software architecture style for building scalable web services. REST gives a coordinated set of constraints to the design of components in a distributed hypermedia system that can lead to a higher performing and more maintainable architecture. This was identified as important during the inception and discovery phase of the project. The REST API has the following features;

●      It provides a common architectural framework for future technical development.

●      It enables potential future changes to information inputs or feeds, their analysis and visualisation, without the need for re-engineering what has been done before.

●      It provides a scalable architecture that will support information feeds and updates in real-time with the potential for significant data volume handling in response to incoming messages and information (Event Driven or Real time).

●      It provides a framework to support integration of different web technologies, meaning that the system is programming language and technology agnostic. External modules could be linked or development more readily as a result.

●      It provides a separation of the UI from the back-end, meaning the HID could have multiple interfaces, using different appearances, even created using different technologies.   This also allows for replacing component parts of the system (e.g. changing database choice) when a need is identified without needing to replace what has been done before.

●      It provides a caching framework so that the system could potentially be refined in future to support true ‘offline’ use, with re-synchronisation when re-connected to the Internet.

●      It enables future extension so that other systems can integrate with the HID, either by sending out or passing in data, using a standard format.

●      It enables scaling to support potential analysis of very large quantities of data from social media and other sources.

●      It supports improvement measures (such as load balancing and queueing) to handle and maintain performance under high loads of incoming data or data processing.

●      It can help maintain performance when storing very large amounts of data.

What has been learned

  • A much clearer understanding of the role and field use reality of the HID has resulted from developments and learning to date – in terms of user requirements, user digital literacy, field teams tasks and operational reality and technical requirements. However, more still needs to be learned from a wider range of field deployments of prototype HIDs.
  • A clearer vision about what the priority implementation environment, context and deployment model is required to prioritize the next stages of development.
  • A good understanding of HID potential user and requirements (personas) has been completed. These were developed as generic personas but they have been most closely tested and validated in the Liberia country office context. Additional effort will be required to extend their testing and validation to other country offices during any future phases of development.
  • The initial plan of reviewing existing tools and taking those and integrating them into the HID was misplaced in what it could achieve. Aptivate was over ambitious in what they could achieve and Internews had too high expectations about what this would result in.
  • Moving from piecing existing tools together to custom development meant the project ballooned in complexity.
  • Investment in robust scale and connectivity infrastructure takes time and effort and, as a result, has significantly slowed technical development during this first phase of work.
  • The design decisions around scale and connectivity infrastructure may be useful for continued development but they will continue to make rapid end user functionality more difficult and expensive. A decision should therefore be made about continuing development within the initial design decisions framework.
    • Important scale and connectivity infrastructure back end, architecture and data modelling has been completed. This will allow future work on processing of data / links (which can be a highly complex undertaking in terms of processing data and volume data storage issues) to be addressed, if that is still a priority.
  • Real time visualization within the HID will be both a challenge and significant opportunity in the next stage of development.
  • Team structure and working has been refined and improved. An effective agile development relationship has been established between Aptivate and Internews. This bodes well for future developments.

Screen Shot 2015-10-10 at 2.32.01 PM

Delivery of value discussion

The prototype HID that has been developed, has been tested and deployed in a single limited environment (for the Liberia office). This small country office had a 3-person team dedicated to collecting (and helping journalists respond to) rumors from communities affected by the Ebola crisis, linking journalists with key decision makers, government bodies and NGOs, conducting SMS polls and essentially acting as knowledge intermediaries to local journalists. This information is sourced from a mix of data streams, in particular; RapidPro, GeoPoll, XLXS files.

The HID in this first phase was envisioned to streamline this information gathering process as it operated in Liberia. It allows users to tag and categorize content from these different sources (RapidPro, GeoPoll, XLXS) through uploading data via spreadsheet. This information can now be handled in a single location rather than collected and processed manually and coordinated via emailing spreadsheets within the office.

Processing data has been supported through batch tagging and deletion capabilities, as well as administrator-configurable bar charts showing data by category against selected dates.

In terms of delivery of value to Internews in Liberia, there has been real value delivered in the following areas:

  1. Operational efficiency of Internews Liberia field team
  • Reduced manual effort or day to day tasks around information flows and analysis
    • Visualizations (bar charts) can be created for sharing on the HID showing categorization of information needs (Questions) from the community. These are generated automatically for tagged data and can be set to each display the same information for different time periods enabling comparison. More can be done to increase the ease by which different time periods can be compared – such as being able to visualize two different time periods on one chart.
  • Improved team coordination
    • Team members can now coordinate tagging and data cleaning for both Questions and Rumors through a single location, removing the need to pass spreadsheets by email for coordination. Data can be added to the HID through spreadsheet upload (specific format for Questions and specific format for Rumors) or by manual entry.
  • Improved usability and ease of system use contributing to increased productivity
    • The HID enables batch processing – deletion of items and categorization of Question types – making the work of initial data handling easier.
    • Data can easily be sorted by imported date, creation date, or by category where applicable.
    • The most recent additions to the system can easily be viewed. The user can choose to see the full set of information held about an item on a per item basis.

2. Increased effectiveness and improved impact of field interventions

  • Increased analysis capability by field teams
    • Questions and Rumors can now have tags applied on a per item basis. This will form the foundation of some powerful data exploration and analysis capabilities once search and filtering have been added to the system. In the interim this data can still be captured for future use.
  • Improved data or information validity through increased data handling, machine processing and improved accessibility to information
    • Categorization capabilities, such as the implemented example of question categorization, enable consistent application of terms to items.
  • Improved information visibility within teams
    • Shared access and storage of information within the HID has moved information out of individually owned spreadsheets (with associated access, security and backup risks) into a single place, visible and accessible to all HID users.
    • Information handling visibility and progress tracking e.g. staff can observe progress and quality of information handling and tagging.
  • Improved opportunities for communication of complex information issues tailored to core field partners and user groups
    • Partners (invited users) can now visit the dashboard to see current information as provided by the team through the use of configured tables, bar charts and textual updates.
  • Improved information reliability and associated trust
    • Data handling process is now shared by users of the HID (in the country office), providing a central canonical storage location for information, increasing data security and hence reliability.
    • More open and shared data handling processes build confidence and trust in data and processes among HID core and invited users.

In addition, there has been considerable value delivered in terms of organizational learning within Internews and Aptivate, that can help guide future HID development.

Regardless of next steps, some core value and important developments have been made;

  • Core architecture has been designed and developed.
  • Prototype delivered for use in the field in Liberia.
  • Experience of all involved available to guide future development.
  • A much clearer set of functions and features and implementation options for Internews to consider to guide further development of the HID. These include a much more nuanced set of effort and complexity costing estimates.
  • Learning, ideas and data to guide the next stages of HID development and deployment.

Now, we are happy to share the code developed so far on GitHub and we invite all interested organizations, developers and agencies to download it, contribute to it or let us know what do they think.

The next phase of development will start once Internews has secured more funding for it, so stay tuned!

Launching the Humanitarian Information Dashboard

Blog post originally published on

Current methods of collating and monitoring information in emergencies are time-intensive and do not allow us to track trends, gaps, or insights over time. Internews’ experience in working with local media in emergency situations is grounded in the work of the Internews Humanitarian Information Programs Unit, which has been responding to emergencies in more than 20 countries in the past 10 years.

The standard Humanitarian Information Service product, which combines different components and communication channels such as SMS, social media, radio programs, TV, and other sources of information, meets the variable and dynamic information needs of any given community. Most recently implemented in Gaza, the Internews’ integrated information systems approach relies on different technological tools specifically chosen on the bases of information needs, targeted audiences, and technology access. In Gaza, this system encompassed radio, TV, SMS, Twitter, Facebook, YouTube, smart phone app Zello, and bulletins for the humanitarian community. In order to analyze feedback from affected communities, local media partners typically track multiple incoming information flows manually. Journalists must monitor, collate, and categorize information from various sources, than structure the feedback in a way that is usable for the design of information programs that can answer expressed needs. This is incredibly resource-intensive, and simply managing and maintaining this process typically precludes any systematic learning of trends, gaps, or insights that reveal themselves throughout the program life cycle.

The Humanitarian Information Service in Gaza

When I was working with the Standby Task Force we had the same problem, but we could overcome it using hundreds of volunteers to monitor different streams of data coming from different sources. This is why UNOCHA and other agencies activated the SBTF in emergencies: the number could overcome the difficulty of monitoring and aggregating data coming form so many sources.

The ability to be able to use and monitor different tools, like mobile phones, mobile apps and social media is becoming more and more important as the ability to collect information on the ground becomes more difficult due to insecurity, structural issues, geography and so on.

To respond to this, Internews is extremely proud and happy to announce the development of an innovative and integrated solution through the creation of a “Humanitarian Information Dashboard”, thanks to funding coming from Global Giving, Google and the Rubin Foundation.

The Dashboard will work like a “Tweet Deck” that monitors, tracks and reports out on information flows across all platforms and where all aid agencies, governments, local media groups can participate and benefit from. Our project in Liberia will be the testing ground for the development of the platform, and we hope that once the platform is ready this will become a standard tool that humanitarian organizations, digital volunteers, and local media will be able to use in a variety of situations. If humanitarians today are “Information DJ’s”, working on “multiple tracks of information and make the best mix they can to augment situational awareness,” to use the words of Patrick Meier, Internews is now building the DJ mixer equipment.

The “Humanitarian Information Dashboard” will simplify and automate the management of information flows from various sources within a standard Internews Humanitarian Information Service. The Dashboard will enable the viewing of information by source (i.e. not only Twitter, but Facebook, YouTube, SMS, voice messages, etc), as well as by topic (i.e. food, water, shelter, security, etc). The Dashboard is envisaged to be a “smart” system that becomes more sophisticated through user interaction with it. For example, a user could tag and associate words with existing taxonomies or categories (i.e. anything with “bread” should be categorized under “food”), enriching the system’s database of word associations along the line of systems like the Health Map gazetteer.

To create the platform Internews will start by looking at already existing platform and codes that can be used for the creation of the dashboard.

The Dashboard will be built using a modular approach to meet the needs of both local media and international responders:

  1. MODULE 1: SMS – the dashboard will allow for several different SMS systems to be fed into it – RapidSMS, FrontlineSMS, Clickatel, a simple modem with a SIM card. This module will be built so that the system used to send and receive SMS can be then chosen on the ground depending on the technical capacity of the organization using the platform or the technical infrastructure. This module will allow for SMS to come into the platform and being categorized, and for SMS to be sent back, either as bulk SMS or as individual one.
  2. MODULE 2: IVR – the dashboard will also allow for Interactive Voice Response Systems to feed information into the platform. Similarly to the SMS module, this system will allow for the several types of IVR system to be used: Verboice, Freedom Fone, Souktel, etc, with the idea that the choice of the specific tool be used can only be done once on the ground and depending on the situation. This module will allow for the administrator to listen to the voice messages recorded, tag them by topic and category and send voice messages back as a call to the sender. This system could be used for any audio clip, including radio audio clips.
  3. MODULE 3: Social Media – this module will be created to add information coming from social media to the dashboard. In this first instance, only two social media will be considered, Twitter and Facebook. The information coming from a specific hashtag or Twitter user, and from a specific Facebook page or hashtag will feed into the platform allowing the administrator to tag them, delete the irrelevant information and respond back if relevant. This specific module will be specifically important for the understanding of the information needs of the diaspora, which plays a very important role in the information ecosystems of the affected countries.
  4. MODULE 4: RSS Feeds – This module is specifically designed to aggregate information coming from specific websites that have RSS feeds capability. This module will allow for the administrators to see in real time information related to the Ebola response as they are published by relevant websites. As in all the other modules, in this too it will be possible to tag and categorize information according to topic and location.
  5. MODULE 5: CSV, XLS and HDX uploads – as described before, numerous organizations are already creating their own information systems, some of which already have a two way capability. With this module those organizations will be able to upload their data, and specifically the records of the questions they receive from the local population, in any format they prefer (other formats will be considered depending on the results of the first preliminary assessment highlighted the following section). Once the information is on the dashboard, the admin will again be able to categorized and tag the information according to the location and the topic.
  6. MODULE 6: FORMS – if organizations using and feeding into the platform do not have access to any of the previous information channels, they will simply be able to insert their data directly by using a form, very much similar to the Google Form system.  Many organizations we work with, including the GSMA, are already generating this content and sending it in, and with this module we would be able to capture this information, tag, categorize, and share it.

Each of those modules can be deployed and used separately so that the dashboard can be customized and adapted every time it is deployed. It can be also used as an incremental product, starting with one module and adding others as required. The Dashboard itself will then aggregate all the data received from the different channels used and allow for several tasks to be performed with the data:

  1. TASK 1: AGGREGATION – as specified before the data in the platform will be aggregated according to the channel though which it is coming from and depending on the channel that administrator will be able to see the sender, the date and time in which the information was uploaded or sent in, and the location, when available.
  2. TASK 2: TAGGING, CATEGORIZATION and FILTERING – as the information is fed into the platform, in the administrators will be able to quickly scroll through the data and delete the irrelevant information; tag the relevant ones with the specific location (which may be the location the information is referring to or the location of the sender of the information); and categorize the information with the appropriate topic of the question (i.e. burial procedures, hand washing, means of contagion, etc.). For each tag and category, the dashboard will allow the administrator of the platform to filter the information categorized under that specific category or tagged with that specific tag. Internews will also add a component that serves as a learning algorithm that will “learn” from these human interactions and start predicting tag and categories. Over time, the algorithm will make this procedure much quicker and fast.
  3. TASK 3: ANALYSIS and REPORTS – All the information coming into the platform, once categorized and tagged, will then be automatically processed to produce analysis and reports. The dashboard will allow the administrator to look at trends, like the change in the topics of information needs, and also at sudden changes in needs and location of needs over time. This analysis will be then be visualized as charts and maps and downloadable as PDF or Word document. Once created those reports will be sent to local media, international organization and all relevant organizations that can use this information for the design of more targeted information system.

Internews will start the design and testing of this platform in our current project “Information Saves Lives” in Liberia. The platform will be designed and tested module by module, using user centered design workshops to design each module, and then working with a test user – most likely a radio station – as a “live user”.  The development and testing phase should last around 6 months, after which we hope to use the platform in a real case before it is finalized for the release. The platform will of course be released an free and open source software that everyone will be able to download.

Are you interested ion this project? We are recruiting the Project manager to be deployed in Monrovia immediately so if you think you want to be part of this, ping me at

We are also looking for partners, technology companies, NGOs, coders, that are interested in working with us on this exciting project, so let us know.

Learning from Kenya: using SMSs for a rapid response mechanism in Central African Republic

One of the biggest problem right now in Central African Republic is the perception in the two different communities, Christian and Muslims, that the two fighting movements, the Seleka and the Anti-Balaka, are the same thing than the communities themselves. This means that when the Seleka attack or kill people, the Christian community attack the Muslim communities living in their areas as a rapresail, and so it happens when the Anti-Balaka attack Muslims communities.

One month ago I had the pleasure to spend two full days with Rachel Brown, the person behind the creation of a peace building organization called SiSi Ni Amani (SNAK) in Kenya. Internews Global Initiative program asked Rachel to share with us her impressive knowledge and her lessons learned from her 3 years project in Kenya.

One of the most impressive and important part of her project for us was the way Rachel understood and used information dynamics to study the decision making processes that lead to violence in the context of the Kenyan society.

In this regard Rachel has done an incredible work based on the fact that violence behaviour is almost entirely based on information ecosystems dynamics, and how the way information is delivered, used, manipulated and spread has an impressive trickled down effect that affects entirely the outcome of a violent action or not.

During my stay in CAR I am having an additional confirmation of this kind of dynamic and I truly believe that the more we are able to adapt Rachel methodology to this context the more we can actually start designing some interesting way to mitigate violence in between the two communities.

Rachel created and carried out two very interested methodologies to understand violence and its development. She first was able to understand and realize that the violent behavior is not the result an immediate and sudden decision but it is the result of a process where information flows in a community and depending on the point in time, format, and the content itself, the informations triggers decision making processes that then results in specific action and behaviors. This dynamic is therefore not immediate but consequential, and therefore an escalating process that can be eventually predicted or guessed if the steps and the triggers are indentified in advance. Rachel called this the Football Match model.

The football match process basically looks at the ball as information. The ball is first kicked by player one in the field, and it is then taken by player two – player two can be of the same team or the opposite one. Once the ball in on player two he will also kick it and the all match continue until there is a score. If we look at this as an information system what happened is that information travelled Ina community from one person to  another and in the process people use that information to make decisons. This decisions and behaviors than trigger other players actions and behaviors. When it comes to violence this is normally translated into the escalation process.

If we look at the CAR dynamics of violence right now this is exactly what it is happening: Seleka fighters attack Christian civilians, and this triggers Anti-Balaka to attack Muslim civilians. This also triggers Christian civilians to attacks Muslim civilians and Muslim civilians to attacks Christian civilians. This all dynamic altogether also triggers civilians of both religions to flee from the areas wp they live in to escape violence, resulting in IDPS camps that become also source of tensions and targets for more violence to happen.

DO_NOT_USE_MARCUSC_2826353b                                              (photo from

Using the football match model Rachel did an analisys of the micro-local behavior that would lead to the final decision of actually engaging in a violent act. She did this by using a second analytical tool, whihc she called the Trigger analisys. The trigger analisys basically looked at whihc type of information, in which point in time would trigger a specific decision to be taken. Since the foot match model allowed her to identify the steps through which the information flows in the community, the trigger analisys was able to give her insights about which specific stage of the process was the one that would make people chose to spread an information or use it as a decision making base.

The third analitical tool Rachel used was what she called the hotspot analisys. The hot spot analisys basically used the insights from the two previous tools to indentify recognizable external evidence that a violent action was about to take place. This kind of analysis was only possible to be done because Rachel worked with members of the communities that had a deep knowledge of their own communities behaviors. It was also possible because Rachel staff had experienced already a violet outbreak, during the elections in 2007/2008 and those events provided very good insights on the development of violence in different communities in the country.

With the use if those three analitical tools SiSi Ni Amani was able to design very specific mitigation programs to use the already existing dynamic in their favor. How did that work?

Basically SiSi Ni Amani had local “monitors”, all members of the communities, that would simply observe their own community over time to identify when a behavior was associated with a hotspot. Once that behavior was indentified, the SiSi Ni Amani staff would use mobile phones to provide information that could potentially change the decision making process to avoid the decision to engage in violence.

Let’s see an example. One one the behavior that was identified by SNAK as an evidence of a possible decision to engage in violence was the fact that young people in certain areas were suddenly gathering in the street and talking in an animated way. This would normally follow the circulation of a rumor or information in the community about a recent violent event against the local community. Once the information was reported to SNAK the team would used already existing pe designed messages and send them as SMS to the local community youth.

Those messages had been designed by other youth in that same community in order to appeal to their peers. Basically the idea behind the design of those messages was that different people in the same community may be subscetible to different appeals. An example would be that young people are very susceptible to the idea of belonging to a group, and therefore if an action is seen as required to belong a certain group, they would most likely engage in the activity. A message that would break that dynamic would be to show them that the may be different other groups they could chose from, and that a certain action may not be the only choice to belong to a certain group.

The SiSi Ni Amani model proved to be effective – in the preliminary results of their extensive survey after the use of this system for the previous Kenyan elections, it looks like more than 45% of the people changed their mind after receiving a message from SiSi ni Amani. It also looks like that more than 60% of people forwarded the message or talk to other people about the issue.

What I am very interested about right now is on how we could potentially adapt this system to Central African Republic. We know that this trickling down dynamic is happening, and we know that rumors as well as information about attacks are at the base of people engaging in violent behaviors. Could we actually replicate this system in CAR to design a rapid response mechanism to prevent outbreaks of violence, at least the one committed by members of the communities against each others? (of course for this to happen the government would have to allow SMS again in the country).

Bluetooth is the Facebook of Central African Republic

[Cross posted from the Internews Centre for Innovation & Learning website]

It has been 2 years since I have been in Bangui. It is exactly how I remembered it: dusty, noisy, chaotic and fascinating.  On the way from the airport to the home/office I am surprised by how the recent events seem not to have changed the look of this city. Sylvain, our country director, reads my mind “Today is a good day, it looks like it is quite. But not all days are like this”.

I look for sights of the recent clashes in between the Seleka and the anti-Balaka..I can’t find any. I only notice that the place where I was used to go get roasted meet back in 2012 is now closed. Later on in the day Jonathan, our resident journalism trainer, will explain to me “There is a food crisis coming – beef meat is not available anymore – the Muslims have historically been the one breeding animals but now they have been pushed out of the country or are too scared to leave their houses, so there is no beef meet available anymore. There are already shortages of meet in some cities in the north and more is about to come. Since beef is not available anymore now, also the other meet is becoming more pricy and the local population have been over hunting bush meat, so that one is getting shorter and shorter.”

It is the end of my first day in Bangui and in the dark of my room I have a lot of things to think about. I loved this place since the first day I came here in 2011. And I thought there was nothing worst than CAR. Now, I can’t stop thinking that this place got worst than I thought it would have been possible, while I hear gun shoots in the background and a helicopter is flying over the house.

I could use this space here to explain what lead to this, but I am not sure I have the answer to be honest. I know it is not about religion, this is for sure, but I am not sure I know what this is all about. I will leave analysts and experts to explain what this conflict it is all about. I will focus on what I am here for: supporting local journalists to give people the information they need to make really informed decisions, and possibly to lead to a reconciliation process that is based dialogue and not imposed by a far distant “international law” that means nothing to the people living in CAR.

As it is always on my first day of mission, today was the day of questions. I believe that there are no experts when it comes to information systems, because by the time you become an expert the situation on the ground is changed already and your expertise is of no use. So the real expertise becomes asking questions and challenge all your assumptions. Only if you do that you can understand what it is gong on.

One of the things that I always tell people that ask me about technology and ICT4D in Central African Republic is that CAR has a mobile coverage of 30% and an Internet penetration of 0. 1%. This is why Internews in the country is working mainly with radios – radio is without any doubts the most widespread mean of communication in the country. Even now, when more than 50% of the radio stations have been looted or destroyed. So what’s about ICT4D?

On the way from the airport Sylvain tells me “There is something interesting happening – people are taking videos of the massacres and the killings with their phones and they share it”. My skeptical me start thinking “How do they share it? And how many people can afford a phone here?”

When I get to the office I immediately start chatting with T. the manager of our local partner organization and a very good local journalist. I ask him what phone he has – and he takes out of his pocket a new Samsung Galaxy 5. I ask him where he bought that. He smiles and says – “We may have no meat, but we have phones.”

Then he disappears only to re-appear 5 minutes later with 4 different phones:


The first one is a fake blackberry and it costs 15,000 CF, almost 32$. It also allows you to watch TV – all without a data plan. The rest of the phones are all around 12,000 CF, 24$. They all have 2 things in common: they have a camera to take pictures and to make video, and they all have Bluetooth.

I spend my lunch chatting with Jonathan about the meaning of this. Basically people take videos and pictures of the massacres and the killing happening in their areas – and share them in between each other using Bluetooth, or sometimes using memory cards. This is a completely closed and untapped information system that we still need to figure out. You get in the circle of information only if you are part of it already.

This got me thinking about the concept of homophily as it was explained by Ethan Zuckerman in the PeaceTech conference in Boston in 2014. In his closing remarks Zuckerman talked about how social networks are becoming vehicle for less dialogue and mutual understanding rather than the contrary, because of their complete reliance on the concept of homophily.

This paper summarize it well the homophily principle [translates into the fact that] people’s personal networks are homogeneous with regard to many sociodemographic, behavioral, and intrapersonal characteristics. Homophily limits people’s social worlds in a way that has powerful implications for the information they receive, the attitudes they form, and the interactions they experience. Homophily in race and ethnicity creates the strongest divides in our personal environments, with age, religion, education, occupation, and gender following in roughly that order. Geographic propinquity, families, organizations, and isomorphic positions in social systems all create contexts in which homophilous relations form. Ties between non-similar individuals also dissolve at a higher rate, which sets the stage for the formation of niches (localized positions) within social space.”

Basically Bluetooth is the Facebook of CAR right now, where closed information systems are created, systems that can function without Internet and still diffuse information that appeal to people sharing the same values – or in this case, sharing the same fears and maybe also hatred. But the real question is “What can we do?” How can we enter those closed and polarized information systems, if we are not even part of it? How can we create systems that prevent or fight against this new form of hate speech – because this is what it is, in the end – while we are not even sure that people really know what they are seeing – it is the Seleka killing the anti-Balaka or the contrary? Or it is just mob violence? It is used to fuel more violence, to show victory or to increase fear? How can we use the same system to fight against it?

Well as I said, this is the day of questions – not of answers, so you have to be patient with me – I really do not have the answer to this. But this all system reminded me of the use of What’s App during the Westgate attack in Kenya, something I blogged about here. Basically those closed systems offer a lot of advantages for people that want to use them to spread rumors or false information:

1) they are closed systems and rely on peer to peer trust – I trust you and therefore I trust what you are giving me – which allows for the primary source to become completely irrelevant to the reliability of the information, because the trust is transferred to others;

2) it allows for the information to spread fast because it is free and relies on homiphily;

3) it does prevent any sorts of cross-verification to happen: only people that are inclined to trust the information will receive it and they only share it with others that have their same values, so the likelihood of someone within the system to doubt the information declines considerably.

What this all system is making me realize is that technology is not only democratizing information but it is also “ghettoing” it, confining it into small areas that we cannot reach it that easily anymore, therefore enabling the creation of closer systems, rather than open one.  I am not advocating at all for a pessimistic “alarmist” approach, but I am wondering, what can we learn from it? How can we investigate and understand those new systems and enter into it, so that our approach to information is truly “global”?

Why the emergency will not be twitted

I have meant to write this blog post since very long time. But I never did since I realized that I am not sure I know what exactly I want to say on the matter. So, take this blog post for what it is: a reflection on what I have been thinking about the issue of using social media to communicate during humanitarian emergencies, and specifically on their usage to communicate with affected communities (I personally dislike this term, but I will use it for lack of a better one).

I’ve always been a big fan of using social media to communicate during disasters. If we look at Haiti for example, we can see the incredible use that not only “official” organization have done of the tool, but also and foremost the local population. The example of the NYC police department during Sandy or of Carel Pedre in Haiti are notable examples of the way different actors have been using social media as a powerful tool during emergencies.

On the other side, if we look at the events coming out from the Arab Spring, we see other very interesting examples of the use of social media by local communities, examples that we have just barely really analyzed, due to the fact that the majority of the interactions were happening in Arabic, while most of the analysis of those phenomenon are being done by English speakers academics.

Notable examples of people using social media during natural disasters can be found in the Philippines, like this map generated by OCHA using social media, or in New Zealand during the earthquake in 2011. 

Now, the question I have been trying to answer to in the past 2 years on the matter is: has any of those initiatives demonstrated without doubts that social media generated content can and is used to make decisions by responders (both local population and humanitarian responders), in a way that allow us to say that those tools are indeed necessary for emergency response?

Unfortunately so far the answer that I have given to this question is no. This does not necessarily mean that indeed social media cannot be used to inform decisions, but that we have not yet been able to demonstrate it.

For several reasons:

1. The technology gap is still there. We may see the future as a world where everyone will have a Twitter account, and where the entire world will have cheap and easy access to internet, but so far this is not yet happening. In the refugee camps in Dadaab, in Mali, in Niger, in Central African Republic people do not have access to those means, internet is still unreliable and where present, expensive. Their emergency is not twitted, facebooked it or youtubed it. (Photo: Meridith Kohut / Internews)

Screen Shot 2013-12-15 at 1.13.02 PM

2. Responders still use other means to gather information. So far there is not prove that any emergency responder agency has been using social media data for targeted and specific responses, apart from single and isolated cases. They still need to figure out the verification issues, the filtering of the enormous amount of data, the processing of the data, and the analysis of it. Once they will have this figured out, then they will be able to see if the intelligence gathered from that data is usable at all. On the community side, according to Internews Information Needs Assessments, people still rely on Radio, or word of mouth to gather the information they need to make better informed decisions.

3. The use of more “traditional” tools, like phones, has proven to be more efficient and more reliable. 911 is still the first number people in the US call in case of emergency, as well as in other countries (where there are other similar there are other numbers). Not only phones are much more widespread than internet, but also allow for more structured (and private) interactions in between people in need and responders – without the noise of the social media.

4. The human contact: we often tend to forget that in emergencies people do need a human contact. Talking to someone looking at him/her in the eye has a huge psychological value that a piece of information, for how useful it could be, does not have. Human interactions are still the basics of what people in distress look for, and are still a fundamental part of the recovery path. Sometimes we should ask ourselves not if a toll is useful or not, but if THAT tool is the best way to convey that MESSAGE.

5. Recents studies, like this one, and this one, and this other one, highlight that social media so far do not seem to carry a specific value when it comes to providing quality information as opposed to other sources. It also looks like it much less trusted than traditional media. Even in the social media sphere it looks like traditional media are still the one that people turn to, to understand what is going on, especially in times of incertitude. In this sense, it looks like traditional media is doing a good job at highlighting relevant information and guiding people through the noise of the social media chatter. This study form Farida Vis on the London Riots shows how the top 200 accounts during the riots where in fact Mainstream Media and Journalists:

Screen Shot 2013-12-15 at 12.50.13 PM


Those are some of the potentials and opportunities that I see in using social media during emergencies:

A. Social Media seems so far to be a very good ADDITIONAL tool to be used by official sources to spread information during an emergency. The tricky point is: how prepared official bodies/agencies are in using this tool, and how much of their strategy is being build in advance so that during the emergency they are actually ready to do it?

B. Social Media still need to prove its value when it comes to be a tool to gather emergency information from affected communities. So far, as I said before, there is little data on how social media information has been used by organization to make decisions. This does not means that the value is not there, but that we are still in the process of understanding how this data can be used, by whom and why. Anecdotal evidence is not enough, and we are still struggling to find tools that would allow us to monitor and understand what is the value of the data collected via social media during an emergency. A huge opportunity can be identified here in terms of working more with humanitarian organization in general to understand their decision making processes, so that social media data can be analyzed and provided in a format that make sense to them, and can be integrated with other data.

Screen Shot 2013-12-15 at 1.15.48 PM

On the other side, we also still need to prove there there is anything valuable in the data exchanged on Social Media during emergencies. We still need to demonstrate that Social Media is a tool that communities use during emergencies AND that carries valuable information that are not already known by responders through other means. Do communities really communicate important information during disasters using SM? And do we define what is “important”? Also, do communities on Social Media make decisions out of Social Media content during emergencies?

The third factor to consider is the comparative advantage: does Social Media provide any advantage when it comes to relevant and timely information to be gathered in order to make decisions?  In other words, is SM economically sustainable? Is the time, and people, and technology used to filter, analyze and manage social media during emergencies really well spend?

C. The Social Media network is as valuable as the real social network. If we work too much on building one and do not work enough the build the later one, we are setting ourselves up for failure. On the other side, huge opportunities lies in the study of social networks and their modeling in disasters. If we look at trusted sources and important nodes in those networks, we can observe how they function during disasters and try to predict where people will turn to in case of need. This will give us a lot of insights about which interactions to observe and monitor during emergencies.

All in all, there is a lot that we do not know, and unfortunately there is also a lot of misleading information out there. The way we carry out analysis of Social Media content from affected communities and the intelligence we get from it needs to be based on very scientific and precise methodologies.

In addition to that there is also very much a lack of fundings in this kind of research, as it seems that as much as donors are interested in funding new cool technologies and big data projects, they are not that interested in funding research to understand how that data can become really usable and relevant.

Social Media for Emergencies 101

2 weeks ago I was invited to lead a seminar for the Communication with Disaster Affected Communities Network (CDAC) in collaboration with Greg Barrow from the World Food Program (WFP). The seminar was aimed at helping humanitarian organizations to understand how to best take advantage of Social Media (SM) during emergencies, with a particular focus on the communication with communities aspect.

The Prezi I used for that training is available here, so feel free to download, it, re-use it, or even update it is you feel there is something missing.

Screen Shot 2013-12-07 at 6.44.45 PM

To be honest this is the first time I delivered a seminar like this one, as I normally teach Social Media strategies for Local Media, not for humanitarian organizations. The interesting thing about this seminar is that I realized how little we know about this subject and how much of what it being produced out there in this regard is very much a tentative to figure it out. For this reason, also my presentation can be described as the best of my tentative to understand what is that we can do with Social Media, and how.

To Begin: before you even decide to have a Social Media Strategy

Social Media as we intend it today refers to interactions among people in which they create, share, and/or exchange information and ideas in virtual communities and networks. Andreas Kaplan and Michael Haenlein define social media as “a group of Internet-based applications that build on the ideological and technological foundations of Web 2.0, and that allow the creation and exchange of user-generated content.” As social media is base on the use (and availability) of Internet or mobile technology, the first step to be done before deciding if social media is the right tool in emergency, is to figure out if the telecommunication landscape will allow people within the emergency affected area, to actually use it. It looks like an obvious statement, but believe me, it’s not!

The second thing to look at is the actual user habits and cultural and social customs when it some sot SM in a specific area. What I mean with that is that, for example, if we look at the use of Twitter and Facebook during the storm Sandy, it is pretty clear that the information provided by the users during that even, was adding no value at all to emergency responders. The reason for that, was that while a lot of social media content was produced during the storm, people that really needed support were using 911 or other emergency related tools to communicate with authorities. The availability of targeted and ad hoc tools to report problems and information was what made Social Media mainly a second choice tool.

The third step to be done is the one related to what we call Proxies. During an emergency, and also in normal times, people ten to listen and talk to people they trust. This is true also in our offline life, we talk and listen especially to people we trust. This is why, if a humanitarian organizations is entering the SM sphere during an emergency, it should do so by first looking at who are the most trusted sources in those spaces, and use them as a vehicle for their information or at least rely on their credibility to reach more audience. Of course, this is not always possible and can also be sometimes tricky, especially if the emergency is happening in a very politicized environment (see Syria for example). Of course, looking for the most commonly used Hashtags and Pages is also advised, especially if there are some that are created by Government Institutions or agencies.

The last step is to make sure that you have the capacity to actually manage your social media presence in an emergency context. This is particularly true as sometimes we forget that social media may be free when it comes to create an account, but it is definitely not free when it comes to actually use it: technical capacity and time commitments are really important variables for an affective SM strategy.

A very good example of this. is the graph here, which really shows all the necessary steps to create and set up an effective SM strategy (designed for businesses, but really applicable to any type of organization):


So, Why Social media for emergencies?
This something that I feel a lot of people never really takes the time to think about. In today’s landscape a lot of organizations are increasingly creating a social media presence ore because they “have to” then because they actually have a goal in mind. This is also due to the hype that certain people are creating around SM, hype that it very rarely supported by any real data (or it is supported by data that is being manipulated to prove a certain point). So, in my experience those are different goals that can be achieved by an effective SM strategy:

1. Provide lives saving information to the affected population

If the humanitarian organizations’ audience is online, and it is using social media, then this is where they should be too. Life saving information, like where to find food and water, shelters, education, services, are all information the HAVE to be delivered in as many ways as possible, and social media can be on of those. In this sense, social media can also be an extremely good way for humanitarian organizations to act against rumors, misinformation and to provide psychological support  (the American Red Cross does that in a very effective way).

2. Gather information useful for the response

Here I will have to make a distinction: social media monitoring for general awareness in the first hours of an event has been used before as seemed to have been proven as very effective (UNOCHA engaged several times with VTCs to do that, like in Libya and the Philippines, and according to their statements, it looks that this piece of information, together with other information gathered from other sources, provided a very good contribution to the situational awareness).

The second case is the use of SM for immediate response actions by humanitarian organizations. This second situation is particularly tricky as there are no reliable and solid studies on the general use of SM for this kind of situations. All events reported are very anecdotal, like the Japan case of a woman being saved after she twitted her status. In addition to that, the problem with proving that there is any value in using SM for response is that, while information on SM could be “actionable” in theory, 99% is not really actionable in reality, as humanitarian organizations do not have the capacity to go and deliver food to every single family that has no food in the first 24 hours of an emergency.

The “actionable” data here could be actionable only once aggregated and cross referenced with other information, like the vulnerability of the recipient, their proximity to other people and so on. A second issue that makes the “actionability” of SM content for humanitarian organization tricky is the fact that the verification and the filtering of that data during emergencies may actually be more time consuming than the use of a hot line or an SMS system to ask people to report their situation. Now, the good news is that machine learning and computational technology are really developing in a way that will make all those processes easier and faster, but the real question still remains: is the information provided on SM any faster, accurate or important than the information that humanitarian organizations already have and can act upon?

Screen Shot 2013-12-07 at 6.55.32 PM

3. Engage communities affected

The engagement process is always a very interesting one, as a lot of organizations still believe that engagement is measured by how many people follow their account. Unfortunately that’s not the case! The engagement issue is strictly related to TRUST. Engagement in this sense is a conversation, and to have a conversation you have to talk, but also to listen. SM is not different in this sense, as it required organizations to engage in multiple conversations with their beneficiaries knowing that they cannot control that conversation. So how should humanitarian organizations engage?
A. By being relevant
B. By being timely
C. By being responsive
D. By being ready to be criticized

4. Manage expectations

This is a very important one. One of the situations I observed in a lot of emergencies in the past 3 years is that frustration and incertitude are directly proportionated to the expectations that local communities have regarding the response. Let me be clear here: I am not saying that people are expecting something that should not be expecting – most of the time they are expecting something that they HAVE THE RIGHT to expect – but most of the times the humanitarian community is not able to provide what has promised – and again, to be clear, I do know that there are more valid and less valid reasons behind this. In this sense SM is  a good tool to manage expectations and lower the level of frustration and incertitude due to the lack of aid or the lack of information about when aid will be received.

The Challenges of using SM during emergencies
1. Strategy: to be effective you need to have a strategy, and better if it is a good one. Just suddenly using Social Media during a disaster with the idea of setting up a strategy later on is not a smart approach. Since there are times of calm in between emergencies, those are the right times to set up a strategy, test assumptions (and tools) and prepare for the big one to hit.

2. Resources: as I said before, using SM effectively requires skills and people dedicated to that. In this sense an effective SM strategy is an investment in the long term, but this investment needs to be done in the first place, and it is not a small one. Several tools, paid or not, can be used to manage SM in a more efficient way, but still, the human component to it, is fundamental.

3. Verification: verifying social media content is not easy, but the good news is that there are several best practices and protocols that different organizations have created and are using these days. One of those is being designed right now, and I will be writing a case study for it. In general verification is entirely depending on the urgency of the information processed and can be outsourced, like for example using Social Media to gather intelligence about an event and to ask the crowd to verify it. In the Prezi you can find more information about it.

Using SM for emergencies is becoming increasingly important and complicated, but it is indeed an important development of the Communication with Communities (CwC) issue. Learning how to do it takes time and practice but it is not impossible. The important thing is to understand that using SM for CwC is not the same thing than using SM for advertising an organization work, and that once you engage in a conversation you cannot really withdraw from it, or try to control it or manipulate it.

The conundrum of digital humanitarianism: when the crowd does harm

We saw this coming. It’s like when you start sniffing and coughing and you know the real flu is coming. And you know you cannot do anything about it, because even if you take a very strong aspirin, it is too late and you will get sick for at least a couple of days.

We saw this coming: the truth is that the consequences of Typhoon Yolanda are not only the one that we see on TV every day, but they are virtual, and the worst of those virtual consequences is coming.

The Digital Humanitarians (or the Volunteer Technical Communities, as you want to call them) came to be out of an amazing change in the world we live: the internet and mobile phones, with which now you can help, and talk, and share with anyone in the world. And you don’t need to have money or to be a professional to provide support to affected communities. You can give your contribution and you help just by sitting behind your desk in your comfy sofa.  But is that true? Does “the crowd” knows what to use and how to use it in a humanitarian crisis, and does it know what are the consequences of its actions? The answer is that no, most of the people in the crowd do not know what they are doing. But they are doing it anyway.

People want to help, and they can help, but they are not necessarily professionals, they have not done this before. Most of the times they do not know what the “do not harm” principle is and means in a practical way, and most importantly, they do not see the consequences of what they do.

This is not a new thing though: we have seen this in Haiti, in Pakistan, in Chile, Libya and we are starting to see this in the Philippines. And because Internet has not been restored entirely in the country yet, the magnitude of what it is gonna happen in the Philippines is still unknown.

You don’t trust me? Want some examples?

Here you are:

Facebook 1

This messages was posted in Facebook (I added the black lines to cover the information) and then reposted by several people – I counted 64,  but only the re-posts from one account. Which means that I don’t know how many others have reposted it from their accounts. To some of you this may look like heartbreaking (and it is in fact) and you may think that it is very nice for people to share this story so that maybe someone will help this family. The problem that I have with this is that whoever re-posted this did not knew that he was basically saying to whoever has bad intention: “Hey, here there is a young woman with two little kids, and two wounded people with her. Here is her location, and she is scared and alone, with no means to get for help”. This is more or less like to tweet that you have a million dollars in your house and that you are gonna go out for a drink and leave the door open.

This second one thought, is even scarier:

FB2 copyThis post, also on Facebook, has already 4000 shares and 5000 likes. The black boxes to cover the children faces were added by myself – here. The post has also 820 comments, a lot of those asking the organization that posted this to delete it, denying what they claim (that they will take all the children under 18 in the Philippines to the US for adoption). But there are also a LOT of comments asking how people can apply to adopt those children – some people even choosing which one they want from the picture. If any of the readers of this post have been to Haiti, they have seen this kind of things before.

But this is not all that is happening, there is more to come. Since the Typhoon, I have received around 10 different emails like this one:

Screen Shot 2013-11-15 at 11.58.10 PM

Now, this company is not doing anything illegal, and they are just offering a very good service to send blast SMS to multiple people in little time – a very good and interesting solution. The problem is, is that a coincidence that I received this just when the Typhoon happened? and that other 10 different companies sent me an email with very similar products or offers?

On Twitter and on several online volunteers Skype chats, I have also seen people already talking about how to set up SMS system to send messages to the people in the Philippines, how to set up a Text2Tweet account, how to harvest requests for help and geo-locate them on line, etc.

Now, don’t get me wrong: information sharing, and the use of technology to provide people with the information they need, when they need it, it is absolutely fundamental! This is why I work for an organization like Internews. Because information is KEY and can save lives. But it can also destroy lives if it is not treated properly and handled with cautions.

In Haiti we saw a proliferation of SMS services, requests for help spread on the internet with sensitive data in it, and multiple people organizing their own “shipment” of goods to the country – not knowing if anyone really needed that. A lot of the people behind those projects REALLY MENT WELL and learned their lessons the hard way. Humanitarian Organizations and NGOs were not immune to this “multiplication” of digital information systems. The real victims though, Haitians people, were the once left to deal with it: misinformation, frustration, and sometimes much more serious consequences.

Let’s be clear here: I am not writing this post to ‘demonize” online VTC or DH communities and individuals. This is not the case. I am myself the co-founder of an organization that does that, the Standby Task Force. This is in of the reason why 3 years ago with some colleagues we created this organizations: to be ready, but most of all to educate people, and to be able to support humanitarian organizations during a crisis, while making sure we can help people not to harm others. This is why we have codes of conduct and CDAC trainings for our volunteers.

But the truth is that the beauty of the internet, in humanitarian crisis, is also its curse: everyone can do everything and does not need to be “trained” or to be a “professional”, or to be part of a formal organization.

So here there are some suggestions to EVERYONE that is today looking at the internet to help and to support people in the Philippines:

1. Do not re-post/share or re-tweet ANYTHING that has sensitive information in it. Sensitive informations are: location of people that need help, names of people, GPS coordinated of their location, picture of children or minors. If you see a message like this, you should do 3 things:

A. Tell the person that posted it to take it off, and explain why

B. Contact the relative NGO working on the ground in the Philippines – use this resource here to find out who does what.

C. If you think that the message is really urgent, and that someone CAN do something about it, look for a hot line number – the government in the Philippines is setting those up for the municipalities locally. See here.

2. If you want to help in any way, contact an NGO, or a humanitarian organization and ask them. If you don’t find anyone, email me(I am at or @anahi_ayala). I swear, I will make time to answer to you. DO NOT DO ANYTHING ON YOUR OWN INITIATIVE IF YOU HAVE NEVER DONE THIS BEFORE. You can also join one of the many online communities that have been doing this for some time, see here a list.

3. Do not try to send information to the local communities online unless you are SURE that the information you are giving is correct. If you see a request for help, use the internet wisely: search for a governmental agency, local or international NGOs, UN agencies that are working in that area, see if they have a contact information.

4. Use this resource to find out what is being done and how.

5. Think before you do. I know, we are emotionally touched by what happened and is happening, but the fact that you are online does not means that the people in the Philippines are virtual. Everything you say or do has consequences in their real life.

6. Do not create new software, technologies an other techie stuff to help the humanitarian community without talking to them first. Here is why and how you can do that.

7. If you hear something that sound crazy, probably it is crazy. Do not  report/share or re-tweet it automatically, but try to figure out from other sources. Do your homework!

8. And for the love of god, DO NOT SET UP YOUR FREAKING SMS SYSTEM!!!!!! Unless you are the PH Government, or a coordinated effort from all the RESPONDENTS on the ground. Really, just don’t do it.

The Internet is a powerful tool, use it wisely and think twice. You are not in an emergency, so you can take the time to do that. Keep willing to help, contributing to the community and make your a contribution to a better world, just make sure you don’t screw up.