UNHCR-350px

United Nations High Commissioner for Refugees (UNHCR)

UNHCR-350px

United Nations High Commissioner for Refugees (UNHCR)

Description of Activities on AI

Project 1: Project Jetson

Project Jetson is a predictive analytics experiment aimed at providing predictions on the movement of displaced populations within and outside of Somalia. It’s a project initiated and launched by UNHCR’s Innovation Service. Jetson technology is machine learning-based and it provides predictions on the movement(s) of displaced people. This experiment also combines data science, statistical processes, design-thinking techniques, and qualitative research methods. Jetson actively seeks new data sources, new narratives, and new collaborations in order to keep iterating, and improving. It has further underlined the importance of partnership, of collaboration, and of transparency.

Project 2: ARiN – DHR Artificial Intelligence Project

ARiN is a web application developed by UNHCR Innovation Service for the affiliate partnerships and recruitment section (APRS) within the division of human resources (DHR). The application is machine-learning based and supports them with the screening process for external candidates coming from the UNHCR external talent pool applications. The talent pools are the most sought-after functional profiles within UNHCR, and they are dedicated to help respond urgently to forced displacement crises. There are approximately 29 talent pools that receive on average 8000 applications per month, which are majority text-based. Contrary to other off the shelf tools, ARiN was customized in order to comply with the internal policies and rules for talent acquisition within UNHCR, which includes transparency of process, gender and diversity elements.

Project 3: Remote protection monitoring: Syria

UNHCR Syria operation, as many other protracted conflict and other special situations ‘(e.g. COVID-19), have humanitarian access difficulties in accessing certain populations and/or geographical locations. For conducting protection monitoring, the team relies on partners reports and originally quantified the data using a Kobo form and a dashboard, counting manually the incidents in more than 700+ reports. UNHCR Innovation Service supported the team to experiment with AI using their original kobo form as training set to build a machine-learning based classification of protection incidents categories and subcategories from mission reports (free-text data) based on pretrained classified data.

Project 4: Refworld

Refworld is a UNHCR repository of legal documentation that contains most of the legal framework on humanitarian affairs and refugee law. Many stakeholders – including legal teams that process refugee sensitive cases – use it as reference for precedents in the law and other court decisions to advance their own legal research. The UNHCR Division of International Protection won UNHCR Innovation Fund 2019 and optimize Refworld navigation for the end-users by the use of artificial intelligence in order to extract metadata, citations and tags in an automated manner, in particular with regard to references in case law to other legal and policy documents. In this way, legal teams can find easier access to information and link to other cases.

Project 5: Reino

Refworld is a UNHCR repository of legal documentation that contains most of the legal framework on humanitarian affairs and refugee law. Many stakeholders – including legal teams that process refugee sensitive cases – use it as reference for precedents in the law and other court decisions to advance their own legal research. The UNHCR Division of International Protection won UNHCR Innovation Fund 2019 and optimize Refworld navigation for the end-users by the use of artificial intelligence in order to extract metadata, citations and tags in an automated manner, in particular with regard to references in case law to other legal and policy documents. In this way, legal teams can find easier access to information and link to other cases.

Project 6: Sahel Predictive Analytics

The UNHCR Special Advisor on Climate Action presented a proposal to members of the High-level Committee on Programmes regarding engaging in a pilot predictive analytics exercise, which aimed to use data, statistical algorithms and machine learning techniques to identify the likelihood of future outcomes based on historical data. The pilot focused on the United Nations system’s shared pressing challenge of tackling the interconnectedness of displacement, climate risks, food insecurity, increased violence and threats to livelihoods in the Sahel region.

Project 7: DEEP.io (inter-agency initiative)

The UNHCR Special Advisor on Climate Action presented a proposal to members of the High-level Committee on Programmes regarding engaging in a pilot predictive analytics exercise, which aimed to use data, statistical algorithms and machine learning techniques to identify the likelihood of future outcomes based on historical data. The pilot focused on the United Nations system’s shared pressing challenge of tackling the interconnectedness of displacement, climate risks, food insecurity, increased violence and threats to livelihoods in the Sahel region.

Project 8: Computer Vision Climate Change and Conflict (Partnership with University of Essex)

Inspired by UNHCR Innovation-led and Omdena Foundation challenge, and as a add-on to Project Jetson, the Human Rights, Big Data and Technology (HRBDT) project of the University of Essex is leading on a research using AI for automatic image classification – particularly satellite imagery – to identify drought patterns and conflict patterns in Somalia region. The aim is to be able to prove the correlation of conflict and climate change, by using computer vision analysis.

Project 9: Qualminer Project (UNHCR Ecuador)

UNHCR Ecuador won UNHCR Innovation Fund 2019 and bulit Qualminer. Qualminer explores qualitative indicators from an ActivityInfo database. ActivityInfo is an information management software that can be used for monitoring and evaluation, case management, inter-agency coordination in multiple contexts. This platform is used as the main monitoring system for the Venezuelan Response plan in Ecuador. The QualMiner project explores the qualitative data used for Venezuelan refugee response, particularly regarding project implementation, by applying text analysis (natural language processing) & other AI-based data mining techniques.

Challenges and Opportunities

Challenges
  • Lack of academic research accountability and ethics research on humanitarian decision-making
  • Lack of human rights due diligence and human rights-based impact assessment for AI applications
  • Lack of research on automated decision-making systems (ADMs) and or AI-based systems in humanitarian sector, including those without human supervision and/or oversight, including feedback loops
  • Data literacy: improve data literacy skills across the organization.
  • Lack of process-oriented AI applications (producing applications just for the sake of producing them and/or follow the UN system trends).
  • Availability of data readiness for A.I. applications (machine-readable data)
  • Managing expectations of A.I. uses and applications: not everything is automated and lots of data pre-work is needed. Data literacy: improve data literacy skills across the organization.
  • A.I. expertise: improve and introduce A.I. and data science expertise and skills and development of applications for internal uses or to adapt external applications for internal uses for business-as-usual processes/support.
  • Implementing standard support structures for A.I. – Organizational culture, more innovative and pushing for change. People cannot implement what they don’t fully understand its functioning and there is resistance for implementation.
  • Communication issues: to counter negative perceptions of technology.

Related Sustainable Development Goals (SDGs)

E-WEB-Goal-08-500px
E-WEB-Goal-10-500px
E-WEB-Goal-13-500px
E-WEB-Goal-16-500px
E-WEB-Goal-17-500px