Algorithm to Distribute Social Welfare Reported for Oversimplifying Economic Vulnerability

Takaful cash transfer program's algorithm which ranks families by their economic vulnerability level to determine financial assistance reportedly oversimplified people's economic situation, fueling social tension and perceptions of unfairness.

Risk classification

• Primary risk domain: 1 Discrimination & Toxicity
• Primary risk subdomain: 1.1 Unfair discrimination and misrepresentation

The Takaful algorithm directly reinforces gender-based discrimination by utilizing sexist citizenship laws to calculate household sizes, resulting in unequal treatment and lower aid rankings for women married to non-citizens.

Additional risk subdomains

• 7.3 Lack of capability or robustness: The algorithm fails to perform reliably or effectively as a poverty-targeting tool, relying on rigid and inaccurate proxies that exclude families in dire economic need.

Causal factors

• Entity: AI
• Intent: Unintentional
• Timing: Post-deployment

The exclusion of needy families and reinforcement of gender bias were unexpected and unintentional outcomes resulting from the deployment of the Takaful algorithmic ranking system.

EU AI Act risk tier

• Risk tier: 2 High Risk

High Risk: The system is used by a public authority to determine eligibility for essential public assistance and social security benefits, which has significant implications for fundamental rights and public interests.

AI system and alleged parties

• AI system: Takaful targeting algorithm
• AI purpose: Resource Allocation; Application Evaluation
• Behaviour type: Tool
• Alleged developer: World Food Programme, UNICEF, The World Bank
• Alleged deployer: National Aid Fund
• Alleged harmed parties: Jordanians in poverty

Harm severity

Highest direct severity in any category: Substantial. Severity is scored from Negligible to Catastrophic in each harm category, for harm the reports describe as caused directly or indirectly by the AI system.

• Physical: direct Negligible, indirect Negligible
• Infrastructure: direct Negligible, indirect Negligible
• Property: direct Negligible, indirect Negligible
• Financial: direct Negligible, indirect Substantial
• Environmental: direct Negligible, indirect Negligible
• Malicious content: direct Negligible, indirect Negligible
• Differential treatment: direct Substantial, indirect Substantial
• Civil rights: direct Substantial, indirect Substantial
• Democracy: direct Negligible, indirect Negligible
• Privacy: direct Negligible, indirect Negligible
• Psychological: direct Negligible, indirect Minor
• Epistemic: direct Negligible, indirect Negligible
• Child sexual exploitation and abuse: direct Negligible, indirect Negligible

Financial

Reported: The report describes indirect financial losses as eligible low-income families were excluded from a $1 billion cash transfer program.

Directly caused: N/A

Indirectly caused: Vulnerable families, such as those earning only $353 a month or needing cars for basic survival, were denied critical cash assistance due to algorithmic exclusion.

Inferred additional harm: Thousands of eligible impoverished households suffered significant financial deprivation and deepened poverty due to being incorrectly disqualified by the algorithm.

Differential treatment

Reported: The report explicitly describes systematic differential treatment of individuals and groups caused directly by the algorithm.

Directly caused: The algorithm treated applicants unequally based on rigid indicators like car ownership and systematically disadvantaged women married to non-citizens due to citizenship-based household size calculations.

Indirectly caused: Women with non-citizen spouses received lower rankings and less assistance, reinforcing existing societal and legal gender inequalities.

Inferred additional harm: Widespread unequal distribution of welfare benefits across Jordan, disproportionately impacting female-headed households and families with minor asset ownership.

Civil rights

Reported: The report explicitly describes violations of human rights, specifically the right to social security, caused directly by the incident.

Directly caused: The algorithm's flawed targeting formula directly deprived eligible individuals of their right to social security and social assistance.

Indirectly caused: The denial of financial aid indirectly undermined families' related rights to food, health, housing, and an adequate standard of living.

Inferred additional harm: Systemic human rights violations affecting thousands of vulnerable Jordanian citizens who were excluded from basic social protections.

Psychological

Reported: The report explicitly describes psychological distress, distrust, and social tension caused indirectly by the algorithm's decisions.

Directly caused: N/A

Indirectly caused: The opaque and rigid ranking methodology caused widespread distrust, confusion, and social tension among applicants who felt pitted against one another.

Inferred additional harm: Severe anxiety and emotional distress are highly likely among thousands of impoverished families who were arbitrarily denied essential financial support.

People affected

• Occurrences reported: 1
• People reportedly harmed: 350
• People reportedly exposed: 1100000

Potential causes

Management

• Prioritization of Fiscal Space: World Bank focused on targeting to restrict spending rather than universal aid.
• Inadequate Risk Assessment: Management failed to assess risks of excluding vulnerable populations.
• Lack of Accountability: NAF and World Bank avoided public disclosure and accountability for errors.

Technology

• Rigid Algorithmic Formula: Formula flattens economic complexity, losing crucial nuance.
• Inflexible Exclusion Thresholds: Automatic exclusion for asset ownership regardless of actual financial value.
• Sexist Legal Code Integration: Sexist legal codes reduce household size calculation for certain women.

Data Inputs

• Unreliable Proxy Indicators: Water and electricity usage are unreliable indicators of poverty.
• Infrequent Data Collection: Infrequent surveys and registration lead to outdated applicant data.
• Inaccurate Income Reporting: Applicants forced to alter financial data to bypass rigid system thresholds.

Human Factors

• Overreliance on Automation: Overreliance on automated systems to solve complex social problems.
• Applicant Confusion: Applicants lacked understanding of the secret ranking methodology and criteria.
• Human Data Compilation Errors: Inaccuracies and inconsistencies introduced during manual data collection.

Process and Methods

• Secret Ranking Methodology: Lack of transparency in the 57 indicators and how they are weighted.
• Inadequate Validation Methods: Lack of validation to verify if indicators matched economic reality.
• Flawed Expense Filtering: System rejected applications where expenses exceeded income by 20 percent.

Regulatory Environment

• Absence of Algorithmic Oversight: No regulatory framework to audit or challenge automated welfare decisions.
• Lack of Transparency Mandates: No legal requirements forcing the disclosure of algorithmic weights.
• Discriminatory Citizenship Laws: Sexist legal frameworks allowed to influence algorithm metrics.

Information quality

• Classification confidence: High
• Reason for confidence: The reports from Human Rights Watch and media outlets provide consistent, detailed accounts of the Takaful algorithm's deployment, its inputs, and its social impacts. While the exact mathematical weights of the 57 indicators remain secret, the qualitative impacts on applicants and the systemic biases (such as gender-based citizenship issues) are well-documented.
• Ambiguities identified: The exact mathematical formula and weights of the 57 indicators are kept secret by the Jordanian government and the World Bank.
• Alternative interpretations: The government and World Bank frame the system as an objective, efficient tool to maximize constrained fiscal space, whereas human rights organizations view it as an arbitrary and discriminatory exclusion mechanism.
• Missing information: The specific list of the 57 indicators and their exact mathematical weights used in the ranking algorithm.

The deployment of Jordan's Takaful welfare ranking algorithm resulted in systemic exclusion of vulnerable families and reinforced gender-based discrimination. While causing notable social tension and distrust in public institutions, the incident represents a minor, localized national security impact manageable through standard administrative and policy reforms.

• Overall national security impact: Minor
• Response level: Moderate
• Scope: Single nation
• Primary target: Jordan
• Alleged perpetrator: National Aid Fund

Threat characteristics

• Imminence: Long-term. Represents an ongoing administrative and societal concern rather than an imminent national security crisis.
• Autonomy: Human-supervised. The AI system ranks applicants autonomously based on 57 indicators, but operates under the administrative oversight of the National Aid Fund.
• Novelty: Established threat. Algorithmic bias and demographic exclusion errors in automated public welfare systems are well-documented global issues.

Impact by dimension

• Physical security: Negligible. No physical security threats, kinetic attacks, or critical infrastructure compromise reported in relation to the Takaful algorithm's deployment.
• Information security: Negligible. No evidence of information warfare, intelligence compromise, or state-sponsored disinformation campaigns associated with this incident.
• Sovereignty: Minor. The algorithmic system's failures compromised government decision-making for aid distribution, causing public distrust but not disrupting core constitutional processes.
• Economic security: Negligible. The economic impact is localized to individual welfare recipients and does not threaten Jordan's national economic stability or strategic industries.
• Societal stability: Minor. Flawed targeting led to systematic discrimination and social tension among applicants, representing a minor impact on societal stability.