What are AI perspectives in ISO 42001?

ISO 42001 is aimed at verifying the trustworthiness of an AI system, and this trustworthiness is expressed through distinct qualities or ‘perspectives’, as they are referred to in ISO 42005 (a related standard which sets out the content required for artificial intelligence impact assessments or ‘AIIAs’).  These perspectives are articulated in terms of benefits and harms to interested parties and are organised into eight categories, as follows:

• Accountability: Which individuals or entities are responsible for a decision made by the AI system
• Transparency: How straightforward it is for interested parties to obtain and understand information about the AI system’s activities, decisions and system properties
• Fairness and discrimination: Ensuring that the decisions rendered by the AI system are fair and free from discrimination, so that the AI system does not perform unjustifiably better or worse for different individuals, organisations, groups of individuals or societies (with the most commonly recognised issue being (actual or perceived) bias)
• Privacy: Ensuring the effective protection of any personally identifiable information (PII) that is used within the AI system
• Reliability: The ability of the AI system to perform as intended
• Safety: Relevant to both systems that involve mechanical automation, such as auto-drive vehicles, and that render decisions with the potential to impact a person’s health or wellbeing
• Explainability: The extent to which a human can understand how the AI system has arrived at the decisions it delivers
• Environmental impact: How the environmental impact of the AI system is managed.