Algorithms and artificial intelligence are now organizing access to jobs, credit, welfare, policing, and justice and are also contemplating entrenching algorithmic discrimination and compromising foundational human rights like equality, non-discrimination, privacy, due process, and human dignity. Objectives include to determine the impact of AI systems on equality, non-discrimination, privacy, dignity, due process and social rights in different sectors; examine the application of the current non-discrimination, data protection and human rights norms to address the issue of algorithmic bias and discriminatory results and find out the particular doctrinal and enforcement gaps and reforms such as hybrid secured ground regimes, compulsory human rights, algorithmic audit, independent oversight organisms. The AI systems that are trained on biased data can be systematically disadvantaged to legally protected groups and black box and opacity architecture can render it hard to refute and show that they are discriminative. However, such regimes are seen to have severe flaws upon application to AI, such as enforcement gaps, information asymmetries. The mixed doctrinal and comparative case study design is an appropriate one and optionally supplemented by limited empirical work is to be done. The argument presented in this paper is the case of the enhanced legal protection based on human rights-grounded like more robust integration of anti-discrimination and data protection laws by means of impelled algorithmic audits, impact analysis and access to the meaningful explanations; sector-specific regulation of high-risk applications like recruitment, credit rating, welfare distribution, predictive policing, and biometric policing; a dignity-based system that instils active anti-discrimination responsibilities and good regulatory solutions both in domestic legislation and new AI tools.
Acharya et al. (Sun,) studied this question.