Not Another Grocery List: Proposals for an Effective AI Ethics Implementation.

Area of Science:

• Computer Science
• Ethics
• Technology Policy

Background:

• The past decade has seen significant efforts to create ethical frameworks for artificial intelligence (AI).
• Current frameworks often rely on axiomatic lists of principles or values.
• These axiomatic approaches face inherent challenges in practical application.

Purpose of the Study:

• To critically analyze the limitations of current axiomatic AI ethical frameworks.
• To propose actionable solutions for developing more effective and implementable AI ethics.
• To reframe the understanding of AI ethics from static lists to dynamic processes.

Main Methods:

• Argumentative analysis of existing AI ethical frameworks.
• Identification and discussion of key challenges: principle divergence, structural incoherence, and ethics washing.
• Development of three core proposals for enhancing AI ethics frameworks.

Main Results:

• Axiomatic lists of AI ethical principles suffer from divergence, incoherence, and the risk of 'ethics washing'.
• Proposed solutions include framework parsimony, balancing structures for multiple principles, and viewing ethics as an ongoing process.
• Effective AI ethics requires moving beyond static principles to a more dynamic and integrated approach.

Conclusions:

• Current axiomatic AI ethical frameworks are insufficient for effective implementation.
• A parsimonious, balanced, and process-oriented approach is necessary for robust AI ethics.
• Reframing AI ethics as a continuous process ensures adaptability and genuine ethical practice.