Ethical Pitfalls of AI in HR

AI in HR is an ethical minefield. Nowhere else do algorithms make decisions that so directly affect people's lives — hiring, promotion, salary, termination. Mistakes here risk not only reputation but also legal consequences.

Anti-Discrimination Law

The General Equal Treatment Act (AGG)

The AGG prohibits discrimination based on:

Race and ethnic origin
Gender and gender identity
Religion and beliefs
Disability
Age
Sexual identity

AI-specific risk: Algorithms can discriminate indirectly, even when protected characteristics are not used as input (proxy discrimination).

Typical Discrimination Scenarios

Application screening:

AI favors graduates from certain universities → social selection
Language models rate "male-sounding" formulations higher
Resume gaps rated negatively → disadvantages parental leave, caregiving, migration

Salary algorithms:

AI suggests salary based on market data → reproduces gender pay gap
"Willingness to negotiate" as a feature → disadvantages cultures with less negotiation tradition

Performance evaluation:

Communication-intensive metrics → disadvantage introverted personalities
"Visibility" as a factor → disadvantages remote workers and part-time employees

Transparency Requirements

EU AI Act — High Risk

AI in HR is classified as high risk. Requirements:

Risk management: Documented process for identifying and mitigating risks
Data quality: Training data must be representative and low-bias
Transparency: Affected individuals must know AI is being used
Human oversight: A human must be able to override AI decisions
Accuracy & robustness: Regular testing and validation
Documentation: Technical documentation of the entire AI system

Explainability

Applicants and employees have the right to know:

That AI is being used (information obligation)
Which criteria the AI uses (transparency)
Why a specific decision was made (explainability)
How they can object (contestability)

Works Council and Co-Determination

Co-Determination Rights (§ 87 BetrVG)

In Germany, the works council has extensive co-determination rights regarding AI in HR:

§ 87 para. 1 no. 6: Technical devices for behavioral and performance monitoring
§ 94: Personnel questionnaires and assessment principles
§ 95: Selection guidelines for hiring and transfers

Consequence: No AI introduction in HR without a works council agreement.

Best Practices for Collaboration

Involve early: Inform works council from the beginning, not just at launch
Agree on pilot phase: Test phase with clear evaluation criteria
Audit rights: Works council can demand regular bias audits
Kill switch: Agreement that the system is immediately stopped if discrimination is proven

Checklist: Using AI in HR Ethically

Bias audit before go-live (by independent third parties)
Works council agreement completed
Applicants/employees informed about AI use
Human-in-the-loop for all significant decisions
Regular re-evaluation (quarterly)
Complaint mechanism established
Documentation EU AI Act-compliant

Conclusion: AI in HR can deliver enormous value — but only with an ethical compass, legal safeguards, and human control. Technology without ethics is not an option in HR.