Human-AI Interaction in Human Resource Management: Understanding Why Employees Resist Algorithmic Evaluation at Workplaces and How to Mitigate Burdens

Title of the Paper

Human-AI Interaction in Human Resource Management: Understanding Why Employees Resist Algorithmic Evaluation at Workplaces and How to Mitigate Burdens

Paper Information

• Subject Area: Human-AI Interaction in Artificial Intelligence and Human Resource Management
• Keywords: Artificial Intelligence (AI), Explainable AI (XAI), Algorithmic Management, Human Resource Management (HRM), User Burden, Transparency, Explainability, Human-AI Collaboration, Algorithmic Resistance, Algorithmic Fairness, Adoption, Trust, Responsible AI

Research Background and Issues

• Problem or Challenge:

  • The application of algorithms in human resource management is increasing, particularly in areas such as performance evaluation and decision-making. However, employees' resistance to algorithmic evaluation may limit the widespread adoption of this technology, especially in sensitive management and evaluation tasks.
  • Legal and ethical issues, data biases, and employees' complex psychosocial factors further exacerbate the challenges of adoption.

• Importance:

  • As companies increasingly adopt AI solutions, understanding employees' resistance and acceptance of algorithmic decision-making becomes critical. Addressing these issues can reduce unnecessary costs and conflicts caused by resistance while ensuring fairness and credibility in the system.

• Research Motivation and Related Work:

  • Existing studies focus more on algorithms in recruitment, with insufficient research on performance evaluation for current employees.
  • Lee et al. found that employees exhibit distrust and negative emotions toward algorithmic decisions in tasks involving "human capabilities," such as recruitment and work evaluation.
  • This study focuses on understanding the specific reasons behind employees' resistance to algorithmic decision-making and how to alleviate such resistance.

Solutions

• Methods or Solutions:

  • The authors conducted in-depth interviews with 21 employees from diverse professional backgrounds using scenario design to explore their perceptions of algorithmic evaluation and associated burdens.
  • Proposed design interventions centered on transparency, explainability, and human-AI collaboration to alleviate employee burdens.

• Innovations:

  • Introduced a classification of six types of burdens (emotional burden, psychological burden, bias burden, control burden, privacy burden, social burden) and corresponding design recommendations to mitigate these burdens.
  • Highlighted the importance of transparent processes and adjustments in the proportion of human-AI collaboration to reduce employee resistance.

• Implementation Steps and Key Techniques:

  • Designed specific scenarios for HRM and conducted interviews based on real-world cases.
  • Used thematic analysis to analyze interview data, identifying categories of employee burdens and mitigation strategies.
  • Proposed design recommendations, including explainable design, transparent management, collaborative decision-making, and emotional support.

Research Findings

• Specific Findings:

  • Identified six major types of burdens employees face in algorithmic work evaluations: emotional burden (e.g., dehumanization and uncanny valley effects), psychological burden (e.g., difficulty interpreting complex algorithmic decisions), bias burden (e.g., unfair decisions due to incomplete data and algorithms), control burden (e.g., ambiguity in algorithm ownership and data input control), privacy burden (e.g., invasion of personal privacy), and social burden (e.g., deterioration of workplace competition and team atmosphere).
  • Proposed mitigation strategies for these burdens:
    • Explainability: Algorithms should explain decision-making reasons in an understandable and interactive manner.
    • Transparency: Evaluation standards and processes should be disclosed, with clear roles for operators.
    • Human-AI Collaboration: Address control and bias issues through collaborative decision-making with balanced weight distribution and reverse monitoring.
    • Emotional Support: Avoid AI exhibiting "false empathy" and provide human-operated counseling services.
    • Social Supplement Mechanisms: Design reward systems that encourage team collaboration and relationship building.

• Advantages:

  • Compared to human evaluation systems, algorithmic evaluations are considered more objective and less biased, but they must overcome social and psychological barriers introduced by new technologies.
  • The proposed solutions reduce negative employee emotions caused by algorithmic opacity or mechanical nature, enhancing system trust and acceptance.

• Experimental or Evaluation Results:

  • Interview results showed that transparency and human-AI collaborative design significantly reduced employees' distrust of algorithmic evaluations.
  • Most participants preferred human-AI collaboration over fully algorithm-driven evaluations.

• Limitations and Future Directions:

  • Limitations:
    • The scenario design method may not fully reflect real employee experiences of resistance and acceptance toward AI.
    • The interview sample size is small, requiring further statistical testing to validate the generalizability of the findings.
  • Future Work:
    • Further validation of the study through real employee experience data (e.g., employees dismissed by algorithms) and expansion of findings.
    • Exploration of best practices for transparency and explainability in HRM, studying the specific impact of various design interventions on different burden types.