Investigating accountability for Artificial Intelligence through risk governance: A workshop-based exploratory study

Area of Science:

• Computer Science
• Artificial Intelligence
• Risk Management

Background:

• The increasing prevalence of AI systems necessitates robust accountability frameworks.
• Existing regulations and standards highlight the importance of addressing AI consequences.
• A gap exists in practical strategies for AI accountability among practitioners.

Purpose of the Study:

• To investigate the applicability of risk governance methods for AI accountability.
• To identify challenges faced by AI practitioners in accountability and risk management.
• To contribute practical solutions for AI accountability in industry.

Main Methods:

• An exploratory workshop-based methodology was employed.
• AI practitioners from academia and industry participated.
• Interactive study design to gather insights on AI risk handling.

Main Results:

• Identified 5 essential characteristics for AI risk management methodologies: balance, extendability, representation, transparency, and long-term orientation.
• Highlighted the need for clearer definitions of risk and accountability.
• Demanded standardization in risk governance and management for AI.

Conclusions:

• Moving AI accountability from a conceptual to a practical stage requires defined characteristics and actions.
• Addressing practical challenges in AI risk management is crucial for industry adoption.
• Standardization and clear definitions are vital for effective AI governance.