ARTICLE:5-ETHICAL AND LEGAL CHALLENGES OF AI IN HR: BIAS, COMPLIANCE, AND TRUST —SRI LANKAN VIEWPOINT

 

Artificial Intelligence (AI) is rapidly transforming Human Resource (HR) activities in sectors across the globe including Sri Lanka's evolving corporate economy. From AI recruitment to AI-driven performance appraisals and workforce analytics AI promises to be efficient and accurate. But AI introduces ethical and legal considerations most prominently around bias in AI decisions and compliance with labor legislation and data privacy rules.

The current paper critically assesses these issues in the context of Sri Lankan organizations and provides real world examples understandings and approaches to controlling risks.

1. Bias in HR Decisions from AI

AI Bias and Its Impact on HR Practices

AI technologies make projections about future results based on historical data. If historical data contain unconscious gender, ethnic, or age biases, then AI can propagate and amplify these biases into hiring, promotions, and compensation (Binnes, 2018).

In Sri Lanka ethnic and gender representation at organizational leadership remains in progress and AI may inadvertently mirror systemic biases if not properly supervised. For example, an AI system that looks at historical promotion records may favor male candidates for leadership positions as the current pool of leadership has a higher percentage of males.

Example: Gender Bias in Recruitment

A leading telecommunication company has also recently implemented an AI based recruitment system for managers where in its initial stage of implementation it was noted that the system was biased towards male candidates. It was found later that this bias represents historical hiring data. Lamplight the problem was realized, and the HR came in to curb these gender fairness issues by rewriting the AI from a better diversified dataset incorporating samples of high-performance female managers. This illustration explains the important demand for constant human monitoring and set auditing during releasing AI for its use in staffing to realize just and ethical output.

Theoretical Lens: Modern Approaches to Fairness and Equal Opportunity in AI

Recent work emphasizes the application of algorithmic ethics and fairness in ethical AI practices that deal with preventing AI systems from imposing discrimination or inequality into decision making processes. Binns (2018) explains fairness in algorithmic systems in his work as creating processes that avoid reinforcement of current prejudice and uphold principles of equal opportunities in results. Likewise, Mehrabi et al. (2021) contend that AI should be programmed to detect, reduce and avoid biases to provide fair and equitable results in sensitive domains such as recruitment and performance management.

Therefore, AI based HR decisions that benefit demographic groups or profiles contravene modern ethical principles and threaten to harm organizational trust and integrity.

Addressing Bias in AI Systems: Modern Approaches

·       Implement diverse and inclusive data sets used for training AI to account for multiple demographics and experiences, reducing the risk of systemic bias (Mehrabi et al, 2021).

·       Conduct regular bias testing and external testing to identify unfair trends and adjust AI models to rectify these (Binns, 2018).

·       Implement human oversight as an essential layer to screen AI generated decisions aligning them with the organizational justice and ethics (Raghavan et al, 2020).

2. Compliance with Labor Laws and Privacy of Data in Sri Lanka

AI and Labor Law Compliance

Sri Lanka's Shop and Office Employees Act and Industrial Disputes Act mandate fair labor practices like nondiscrimination, equality in treatment, and due process in grievances and terminations. AI based HR systems need to be following these laws as well to avoid unfair labor practices.

For example, if an AI system explicitly excludes applicants over a certain age for IT job positions (on the basis of flawed assumptions about being adaptable) it could amount to age discrimination in Sri Lankan law.

Real Example: AI-Based Monitoring and Workers' Rights

Some Sri Lankan BPO companies have begun using AI technology to track productivity levels of workers such as login/logout patterns, keystroke patterns, and communication frequency. While these applications assist in improving efficiency continuous monitoring without worker consent can be contrary to expectations of privacy and workers' rights.

Data Privacy and AI in Sri Lanka: Gaps and Risks

Although Sri Lanka passed the Personal Data Protection Act (PDPA) in 2022 the process of implementation is ongoing, and most organizations have yet to make the necessary changes. AI systems that handle employee sensitive data such as health data or biometric data are subject to data minimization, consent, and transparency requirements under PDPA.

Example: Financial Sector and AI-Driven HR Tools

Sri Lankan banks such as Sampath Bank and Commercial Bank are exploring AI for employee performance analysis. However, AI tools use for performance measurement without transparent communication and data protection could subject them to regulatory scrutiny in line with PDPA.

3. Theoretical and Ethical Foundations for AI in HR

Utilitarian vs Deontological Explanations of AI in HR

Integration of AI in HR functions such as recruitment performance evaluation and employee management present complex ethical issues. A utilitarian approach views how the ability of AI to enhance efficiency, reduce costs, and enhance decision making can deliver enormous net gains for companies (Taddeo & Floridi, 2018). This explains outcomes and the common good for the organization and stakeholders.

However deontological ethics that believe in duties and rights oppose that such efficiency cannot at any cost compromise fairness, dignity, and respect for human beings (Mittelstadt, 2019). Deontological teachings would for instance reject such complete automation choices such as the termination or giving promotions to employees without human judgment as they are made on the basis of people as devices towards organizational ends instead of responsible persons deserving respect. AI HR processes need to be transparent, equitable, and respectful of employees' rights and adhere to such ethical requirements (Jobin, Ienca & Vayena, 2019).

Contemporary Stakeholder Theory and the Impacts of AI on Employee Relations

Based on contemporary stakeholder theory organizations are increasingly being asked to manage the interests of various stakeholders from employees and customers to communities. Harrison et al. (2019) observe that organizations applying AI in HR are concerned not only with shareholder returns but also with AI's effects on employees' trust, wellbeing, and privacy. AI systems that undermine employee morale, trust, or equity can potentially weaken long-term organizational reputation and sustainability. As HR practices have a direct impact on employees a key stakeholder group AI uptake must be within ethical and socially responsible standards that promote inclusivity, respect, and openness (Dignum, 2019). This is particularly crucial where workplace culture and relationships matter most.

4. How can Sri Lankan Organizations prevent from AI-Related Ethical and Legal Challenges

 1. Implement AI Ethics Models Tailored to Local Environment

•                Develop organization-wide AI ethics models that align with Sri Lanka's PDPA and labor laws.
•                Weave local cultural sensitivities concerning privacy and equity in artificial intelligence models.

2. Explainability and Transparency (XAI)

•            Implement Explainable AI so employees understand how and why AI decision making occurs (Doshi-Velez & Kim, 2017).
•            Establish appeal mechanisms for HR decisions made using AI.

 3. Employee Awareness and Consent

•               Obtain explicit consent when collecting personal data to be used in AI.
•                Educate employees on AI use and PDPA rights.

 4. Cross Functional Committees to Manage AI

•            From committees of HR, legal, IT, and staff representatives to manage AI implementation and ensure ethical alignment.

5. Human AI Collaboration

•    Provide for human intervention in major decisions, such as hiring and termination, to balance AI efficiency with human justice and empathy.

CONCLUSION

Though AI is highly capable of changing HR, Sri Lankan companies need to steer clear of legal and ethical challenges. Addressing AI bias and conforming to labor and data protection legislation is imperative to creating trustworthy and unbiased AI-enabled HR processes. Adopting ethical AI mechanisms and ensuring openness, Sri Lankan business can be the trendsetters for ethical AI adoption, protecting workers' interests in addition to organizational interests.

REFERENCE LIST

Binns, R. (2018) ‘Fairness in machine learning: lessons from political philosophy’, Proceedings of the 2018 Conference on Fairness, Accountability, and Transparency (FAT), pp. 149–159.

Dastin, J. (2018) ‘Amazon scraps secret AI recruiting tool that showed bias against women’, Reuters, 10 October. Available at: https://www.reuters.com/article/us-amazon-com-jobs-automation-insight-idUSKCN1MK08G (Accessed: 10 March 2025).

Dignum, V. (2019) Responsible artificial intelligence: how to develop and use AI in a responsible way. Cham: Springer.

Doshi-Velez, F. and Kim, B. (2017) ‘Towards a rigorous science of interpretable machine learning’, arXiv preprint, arXiv:1702.08608.

Harrison, J.S., Barney, J.B., Freeman, R.E. and Phillips, R.A. (2019) ‘Stakeholder theory in the modern era’, Academy of Management Annals, 13(1), pp. 321–350.

Jobin, A., Ienca, M. and Vayena, E. (2019) ‘The global landscape of AI ethics guidelines’, Nature Machine Intelligence, 1(9), pp. 389–399.

Kant, I. (1785) Groundwork for the metaphysics of morals. Translated by M. Gregor. Cambridge: Cambridge University Press (1997).

Mehrabi, N., Morstatter, F., Saxena, N., Lerman, K. and Galstyan, A. (2021) ‘A survey on bias and fairness in machine learning’, ACM Computing Surveys (CSUR), 54(6), pp. 1–35.

Mittelstadt, B.D. (2019) ‘Principles alone cannot guarantee ethical AI’, Nature Machine Intelligence, 1(11), pp. 501–507.

Raghavan, M., Barocas, S., Kleinberg, J. and Levy, K. (2020) ‘Mitigating bias in algorithmic hiring: evaluating claims and practices’, Proceedings of the 2020 Conference on Fairness, Accountability, and Transparency, pp. 469–481.

Shop and Office Employees Act (Chapter 129). Parliament of Sri Lanka.

Sri Lanka Personal Data Protection Act, No. 9 of 2022. Available at: https://www.parliament.lk/uploads/acts/gbills/english/6207.pdf (Accessed: 10 March 2025).

Taddeo, M. and Floridi, L. (2018) ‘How AI can be a force for good’, Science, 361(6404), pp. 751–752.