Utilizing artificial intelligence for equitable and efficient volunteer selection

Utilizing artificial intelligence for equitable and efficient volunteer selection

Continuing with the UNV Digital Transformation journey, our next major milestone is the introduction of the state-of-the-art cloud-based volunteer recruitment and management platform: the Unified Volunteer Platform (UVP) in May 2021. The establishment of the UVP, coupled with the introduction of new UN Volunteer Conditions of Service (COS), will result in significant efficiency gains for partners, specifically the faster deployment of UN Volunteers worldwide, at a reduced operational cost.

UNV’s new Unified Volunteering Platform will replace several older volunteer recruitment and management applications. While UVP users will enjoy a number of major policy, process and system improvements, one innovation we are working on is the strategic utilization of Artificial Intelligence (AI).

Within the UNV Digital Transformation landscape, we are already using an AI-powered Chatbot, Voli. Voli has already engaged in over 150,000 conversations from the beginning of 2020, with less than 3 per cent of enquiries requiring intervention from the human-powered service team. Now, we are testing the use of AI for steps of the volunteer mobilization process, as part of the UVP. 

UNV fills thousands of new UN Volunteer assignments annually. An even larger number of Online Volunteers are supporting organizations worldwide via the UNV Online Volunteering platform. These volunteer engagements are the result of a rigorous selection process, which starts with application review and longlisting, or the first filtering, of candidates.

Overall, the number of ‘long-listed’ candidates UNV is processing annually is a five-digit number, which is growing from year to year. UNV is exploring the possibility of using AI to perform the longlisting process and provide a reliable list of candidates who best fit the requirements for a given volunteer assignment to be considered as part of the narrower shortlist. UNV is currently finalizing the integration of artificial intelligence talent API and associated machine learning. 

Last year was a remarkable year for UNV in terms of gender equality and women’s empowerment. For the first time in history, UNV reached full gender parity among UN Volunteers deployed in 2019. While considering the integration of AI in the UVP, important questions on AI fairness came up.

While AI seems neutral, it is programmed by humans, hence there is a risk that AI could internalize the same biases human may have. We are keen to ensure that AI does not introduce any gender bias that may endanger the gains we have achieved in gender parity. AI is like a child; it learns from you. If you teach your AI using a biased dataset, it becomes a biased tool. --Hae-Yeon Alice Jeong, Gender Focal Point at UNV

To further explore the topic and establish an effective monitoring approach, UNV’s Gender Action Team, the UVP team and AI vendor initiated a review of the measures in place and agreed on controls to guide the process going forward.

The topic of AI fairness and bias is a science of its own. In our collaboration, we have agreed which processes need to remain in the hands of humans who observe unbiased and fair selection principles, and which can be automated for efficiency gains. --Avetis Avagyan, Digital Transformation Project Manager

The current controls UNV has established include:

  • removing all gender-identifying data (“protected attributes”) from the datasets which are subject to Machine Learning;
  • ensuring that short-listing and final candidate selection remains in the hands of humans, who observe unbiased and fair selection principles;
  • excluding exceptional decisions from Machine Learning, as consistency in application of procedures is important to avoid confusion;
  • hiding photos of candidates on the Unified Volunteering Platform;
  • applying gender-neutral language on the platform by observing UN Globe recommendations on gender-inclusive language;
  • developing real-time reports on inclusiveness and fairness in AI and overall recruitment through business intelligence; and
  • monitoring relevant key performance indicators throughout.

We are grateful to the German Federal Ministry for Economic Cooperation and Development (BMZ) for its support to UNV’s Digital Transformation project, which has enabled our innovative pathway.