Candidate matching through AI takes place instantly upon a candidate’s application to an onsite or online assignment. Various data fields in the candidate profile and motivation statement are automatically matched against the description of assignment and a relevance score is awarded accordingly.
In this way, recruiters with UNV and UN host entities have an instant overview of all applications, listed by the AI matching score. The next step is human action to decide which applicant to shortlist for consideration for the assignment. Human control at this stage is essential, also to ensure that no AI bias or erroneous assumption goes undetected.
During the past 12 months, the Unified Volunteering Platform has registered over 800,000 matching operations. Over 30,000 shortlisted applicants and 9,000 eventual selections provided Machine Learning with sufficient data to optimize its engine using the current model.
At the same time, several opportunities to optimize have emerged. UNV is currently working with an AI vendor to refine the model to achieve higher matching precision, which will also enhance machine learning efficiency.
UNV has embarked on the project “AI Phase 2” which targets refining and retraining the model to optimize performance, and detailing a future AI roadmap for the Unified Volunteering Platform.
One of the building blocks of the roadmap is an AI-guided candidate suggestions feature to notify highly matching candidates about assignments relevant to their skills and experience to motivate them to apply. Another is map UNV's talent pool, which currently comprises over one million candidates, against different areas of expertise. This would be particularly useful for talent outreach campaigns.
The AI Phase 2 project will also highly benefit from and comply with the recently endorsed recommendations in the Principles for the Ethical Use of Artificial Intelligence in the United Nations System of the UN System Chief Executives Board for Coordination.