Proactive AI-Driven Benefit Assignment — Azerbaijan's State Social Protection Fund System
Azerbaijan
Azerbaijan's State Social Protection Fund uses AI-based data matching across 15+ agencies to proactively identify and enrol eligible citizens in …
Jordan · Amman · See the Jordan profile
Jordan's National Aid Fund uses a 57-factor algorithm to target 250,000+ households under Takaful, World Bank co-designed. HRW (2023) and CHRGJ (2024) documented opaque eligibility, systematic exclusion of non-Jordanians, and no appeals mechanism.
Jordan's National Aid Fund (NAF), under the Ministry of Social Development, operates Takaful — a World Bank-co-designed cash transfer programme that uses an algorithmic eligibility system to identify and rank vulnerable households. Launched in 2019, the programme had reached over 250,000 households by 2023, targeting the approximately 24% of Jordan's population living below the poverty line.
The algorithm evaluates households across 57 undisclosed factors, assessed at a single point in time to assign eligibility scores that determine cash transfer amounts. The World Bank provided technical co-design support, framing algorithmic targeting as a means of improving precision and reducing discretion compared to administrative caseworker decisions.
Two independent assessments document significant governance failures. Human Rights Watch (July 2023) found the system excessively inflexible — unable to respond to dynamic poverty conditions — and systematically excluding Syrian refugees, Palestinians without Jordanian passports, migrant workers, and non-Jordanian family members of Jordanian women. The Center for Human Rights and Global Justice (February 2024) further documented how the algorithm's full opacity — no factor weights or eligibility thresholds are publicly available — leaves affected households without meaningful grounds to challenge adverse decisions, creating what the report termed a 'transformer state' that obscures power under algorithmic neutrality.
Takaful illustrates a recurring dilemma in AI-assisted welfare delivery: national-scale coverage breadth is achievable, but rigid point-in-time targeting combined with opaque criteria can entrench exclusions of particularly vulnerable populations. It underscores the importance of transparent factors, dynamic data updates, and accessible appeals mechanisms as design requirements — not afterthoughts — in welfare AI.
Read the full analysis: https://chrgj.org/2024-02-19-transformer-states-jordan/
Implementation detail (cost, timeline, staffing, conditions for success) is not yet available for this practice.
Where this practice's information was retrieved from, and when.
Azerbaijan
Azerbaijan's State Social Protection Fund uses AI-based data matching across 15+ agencies to proactively identify and enrol eligible citizens in …
Canada
Canada's IRCC uses an AI eligibility model to triage and auto-approve routine IEC work permits. Launched Nov 2023 with a …
Kazakhstan
Kazakhstan's Digital Family Card uses predictive scoring across 100+ socio-economic indicators from 20+ agencies to proactively identify and serve vulnerable …
Pakistan
Pakistan's BISP operates a 35-million-household registry using Proxy Means Test scoring cross-referenced with NADRA biometric data to identify the poor; …
Open full copilot Grounded in cited practices — always check the sources.