DEEP — Using Artificial Intelligence to improve global coordination between humanitarian actors

Mitigating data collection challenges in humanitarian contexts during COVID-19 pandemic through United Nations (UN) partnerships and other actors

The WHO Director General has characterized the coronavirus outbreak as a global pandemic since March 2020, creating harm to peoples and economies in all countries.

Access to quality structured information in humanitarian contexts will be elusive for the foreseeable future. Most humanitarian responses will be made on the basis of remote techniques, secondary data, social media, expert judgement or journalism. The least developed countries — which are most vulnerable to the pandemic — are the most impacted by this lack access to direct data, and organizing remote data collection was a necessity.

To address this challenge, Data Friendly Space is partnering with the Danish Refugee Council, GIMAC (a UN lead consortium) and iMMAP through USAID OFDA funding to provide support for data collection and analysis to countries impacted by COVID, by using DEEP (Data Entry Exploration Platform). DEEP was specifically developed by and for humanitarian actors to improve data processes and enable more effective and accurate humanitarian responses to humanitarian crises. Through this work, DFS is helping to inform the humanitarian community, at local, national and global levels, to enable more effective and accurate humanitarian responses to crises impacted by COVID.

Improving the efficiency of data processes allows organizations to focus on what matters the most — saving lives

The humanitarian-response ecosystem is transitioning from information scarcity to information overload. Voluminous quantitative data, secondary data, and media-based sources of qualitative data are generating information on a scale that humanitarian organizations cannot cope with. They are overwhelmed and lack the expertise to utilize the available data to make the right decision during a crisis.

To address this challenge, DEEP was established in 2016 — an artificial intelligence (AI)-driven platform specifically designed to improve data processes and information management to humanitarian crises at national and global levels.

DEEP allows humanitarian stakeholders to compile and collate crucial information from various sources and systematically organize that information through customizable and user-friendly tagging structures so that the information can be easily monitored and analysed to inform decision making.

DEEP is governed by a board which consists of some of the world’s largest humanitarian organizations and includes: OHCHR, iMMAP, UNHCR, UNICEF, UN OCHA, JIPS, IDMC and Okular-Analytics.

Organizations on DEEP’s board, among other INGOs, IGOs, Local & National NGOs, governments and Clusters are also users of the platform for projects all along the Humanitarian Programme Cycle, such as:

• Needs assessment and analysis: Connect DEEP to any online data source and carry out data collection and needs assessment studies through its AI-assisted secondary data review, and the — coming soon — Kobo integration for primary data collection.
• Strategic response planning: After carrying out needs assessment and analysis, you will be able to access actionable geolocalized data and prepare your response plan. Define country strategy — narrative, strategic objectives and indicators — and cluster plans consisting of objectives, activities and accompanying projects, which detail implementation and costing of the strategy.
• Resource mobilization: Map projects & activities, resources, budgets and beneficiaries reached into a user-friendly interface, and fasten collaboration with your team members. Export data, infographics and maps to present it to stakeholders, funders and donors.
• Implementation and monitoring: Carry out multidimensional analysis to gain real time insights on your response plan, from project and activity mapping to monitoring. Assess targeted and reached beneficiaries. DEEP could be used for activities such as 3W : Who does What & Where, 4W : & When, 5W : & for Whom.
• Operational peer review and Evaluation: Support the enhancement of the collective response and identify good practice & learning to share with other operations: leadership arrangements; key obstacles affecting operational, coordination mechanisms, accountability to affected people, and implementation of the humanitarian programme cycle.
• Preparedness: Anticipatory Actions, using historical data on the impact of crises enable the humanitarian community to better estimate expected needs in case of a crisis event of similar magnitude happening in the future.

And much more:

• Research Centers & Academics: Organize research paper production within DEEP, by defining custom analysis framework, simplifying your workflow, automating your tasks and centralizing your information within DEEP.
• Information Management (IM), Coordination between organizations and Clusters, such as GIMAC or Cluster Coordination Teams working with Local & National NGOs, INGOs and Governments.

Towards a global humanitarian data platform for insights based action

DEEP aims to become a leading global consolidator and reference for humanitarian data. It has been piloted in more than 1,200 projects supporting humanitarian responses across the globe since 2016 across all humanitarian sectors. DEEP has been used to support humanitarian responses whose combined targeted recipients number more than 98 million people, with 66 million earning less than 5 USD per day. DEEP was crucial to the UNHCR in responding to the Venezuela migrations crisis and to the ACAPS in responding to the Rohingya crisis. In 2019, IFRC showcased its use of DEEP to support emergency needs assessment in the Venezuala migration crisis.

DFS would be glad to collaborate with organizations that are interested in making data in DEEP more accessible. Some of the COVID related data currently available in the platform includes:

• Hundreds of monthly analysis points at sub-national levels (here is a sample for Syria), thematic reports analyzing the evolution of ongoing situations, e.g. quarantine effects, protection concerns, access to basic goods and services, market prices, etc.
• Hundreds of weekly snapshots of available information at sub-national levels summarized visually in dashboards for each country.

Interested individuals can be given access to these projects upon request. The DEEP platform can be accessed for free here.

A dashboard in DEEP showing assessment data for the Venezuala migration crisis

Improving global coordination with the help of Artificial Intelligence (AI)

DFS and the DEEP project aim to significantly cut data processing time for humanitarian organizations, saving critical time in emergency responses. Today, DEEP, with more than 85,000 annotated humanitarian response documents hosted on its platform, is in a unique position to create Natural Language Processing (NLP) models that serve the humanitarian community at large. With the on-going development of new NLP models, DFS is aiming to assist and eventually replace the need for secondary data reviews done by humanitarian analysts which in turn will allow for much faster response to humanitarian crises.

DEEP aims to use AI to leverage existing information to better anticipate humanitarian conditions. Historical data on the impact of crises can enable the humanitarian community to better estimate expected needs in case of a crisis event of similar magnitude happening in the future. This will be especially useful for large events where telecommunication systems are impacted, and direct data collection on the ground is not possible. For example, in the case of tropical storms in Bangladesh, key stakeholders could anticipate the impact of future storms on the population and clearly identify priority needs, geographical areas or affected groups. This will allow a faster and more accurate response for such situations where the first hours are critical.

Our long term vision is to provide a global solution for inter-organizational resource optimization with the help of AI. Both DFS and DEEP partners are convinced that the technology has a growing role to play in improving humanitarian coordination by contributing to faster and more accurate humanitarian responses. By sharing resources at the global level between humanitarian actors, the time required for humanitarian responses could be greatly decreased, and lives saved.

Data Friendly Space and DEEP Now

Through a partnership with the Danish Refugee Council, USAID recently provided DFS and partners 2 years of funding for development and technical management of DEEP. DFS is seeking additional partner organizations to continue to foster the growth of DEEP and expand its critical role in the international humanitarian community.

More specifically, DFS is seeking to:

• Share COVID-19 related data. Partner organizations could for example access geolocalized data on the number of COVID-19 cases with underlying factors, people in need for humanitarian response on the ground, population displacement information, legal measures adopted by governments, effects of confinements on local economies, schools, health systems etc.
• Partner with academic and private sector organizations to improve the platform’s AI and NLP functionalities. DFS and DEEP are willing to share NLP datasets and models created with standard humanitarian taxonomies. DEEP provides NLP services to help humanitarian organizations automate data collection and classification, which could be opened to new actors or connected to existing systems on demand.
• Secure support for both additional improvements to the DEEP platform and foster adoption of DEEP amongst additional humanitarian organizations, research centers, foundations, or any other interested organization.

Want to get involved?

Our work is only fully realized through collaboration with external partners. If you are interested getting further involved with DFS or DEEP on these initiatives, please contact us at partnerships@datafriendlyspace.org

Written by: Valentin Pistorozzi —Director of Development — Data Friendly Space