Perceptions of Fairness in Machine Learning-Driven Hiring at the Federal Board of Revenue in Pakistan

Machine learning (ML) in public-sector hiring at Pakistan’s Federal Board of Revenue (FBR) enhances efficiency but risks perpetuating biases, challenging constitutional quotas for equitable representation. This study explores candidates’ perceptions of fairness in ML-driven hiring and their alignment with Pakistan’s mandates for gender (10%) and regional (6% Balochistan) quotas. Using a survey of 163 candidates, analyzed with chi-square tests, regression analysis, and descriptive statistics, we found no significant associations between gender or region and perceived fairness, despite audits revealing Balochistan’s 27% underrepresentation and 15% higher rejection rates for women. Regional disparities in perceived quota alignment were notable, though data limitations reduced reliability. Low awareness of ML use (42.3% uninformed) suggests procedural opacity shapes neutral perceptions, masking inequities. We recommend quota-aware algorithm audits, transparent communication, and stakeholder engagement to align ML systems with constitutional equity goals, fostering trust in automated hiring. This study highlights the need to bridge technical fairness with legal mandates in public-sector recruitment.