Knowledge Workers' Perspectives on AI Training for Responsible AI Use

Research Background and Issues

• Issues and Challenges: Knowledge workers are attempting to adopt artificial intelligence (AI) in rapidly changing work environments. However, it remains unclear whether they receive adequate education and support when using AI. Potential issues include workers abandoning AI tools, misinterpreting AI outputs, or unintentionally perpetuating biases.
• Significance: The lack of sufficient AI training may result in the untapped potential of AI tools, while neglecting workers' rights, privacy, and considerations of diversity, equity, and inclusion (DEI) could lead to serious problems.
• Research Motivation: Although there is existing research on AI-related skills, these studies often focus on the needs of management, overlooking the perspectives of ordinary knowledge workers (e.g., middle managers and individual contributors).

Solution

• Proposed Approach: Through workshops with 39 knowledge workers and 17 follow-up interviews, the authors identified four major challenges in AI training and proposed nine relevant training topics.
  • The four challenges include: lack of basic AI knowledge, blind trust in AI outputs, neglect of AI-related DEI risks, and potential violations of workers' rights and privacy by AI.
  • Each challenge corresponds to specific training topics, such as understanding basic AI functionalities, critically interpreting AI outputs, identifying dataset biases, and safeguarding worker privacy.
• Innovative Contributions:
  • Developed a multi-level AI training framework encompassing both technical and non-technical domains.
  • Focused on the needs of knowledge workers rather than designing training solely from leadership perspectives.
  • Advocated for integrating ethical and fairness considerations into AI training.
• Implementation Steps:
  • Research and Understanding: Identify the challenges and needs of knowledge workers through workshops and interviews.
  • Propose Training Topics: Develop nine training topics tailored to different knowledge domains and organizational contexts.
  • Method Exploration: Recommend leveraging existing HCI research prototypes as practical learning tools for knowledge workers.

Research Outcomes

• Specific Achievements:

  • Identified four core challenges faced by knowledge workers in using and training for AI tools.
  • Proposed nine actionable training topics, including foundational AI knowledge (e.g., what AI is and what it can do), critical thinking skills (e.g., how to interpret AI outputs), DEI-related education, and mechanisms to protect worker privacy and rights.

• Advantages Over Existing Solutions:

  • Unlike previous studies centered on technical developers, this project is based on the real needs of knowledge workers, contributing to a more inclusive and practical AI safety training system.
  • Goes beyond traditional technical training by incorporating social and ethical considerations.

• Experimental and Evaluation Results:

  • Workshops and interviews conducted across countries and industries provided diverse perspectives, covering 26 countries and multiple sectors.
  • Analysis of dataset conditions and localized AI tool performance further highlighted the complexity and multidimensionality of the issues.

• Limitations and Future Directions:

  • Limitations:
    • The sample primarily covers knowledge workers in specific fields, which may not fully represent the needs of other worker types.
    • Cultural and regional biases exist in the study, as some participants may have been limited in their expression due to language barriers.
  • Future Directions:
    • Expand to other industries and larger sample sizes to investigate AI training needs for non-knowledge workers and various work domains.
    • Develop a prototype tool repository to help organizations test, select, and deploy practical AI training tools tailored to their workers' needs.
    • Explore methods to enhance AI tool performance across cultural and linguistic boundaries while addressing DEI and technological fairness.

Conclusion

This study analyzed the needs and challenges of knowledge workers in safely and effectively using AI in workplace scenarios through workshops and interviews. The proposed nine training topics not only cover technical aspects but also address critical ethical and social issues. These findings provide valuable references for designing AI education and capacity-building initiatives for knowledge workers, while calling on the research community to create prototype tool repositories and more equitable technology deployment models to support diverse work environments.