Bias in AI
Systemic errors in AI output leading to unfair outcomes or discrimination.
How the laws define this
Both regulatory documents establish that bias consists of systematic or systemic, repeatable errors within a system that produce unfair outcomes. Their primary point of divergence lies in the scope of the technology covered and how the resulting unfairness is defined. The National Policy Framework for Artificial Intelligence (Legislative Recommendations) focuses specifically on an AI system's output, framing unfair outcomes as discrimination against certain demographic groups. In contrast, Responsible AI in action: Guidance for Regulators applies to any computer system and defines the harm more broadly as privileging one arbitrary group of users over others. Responsible AI in action: Guidance for Regulators also extends its focus beyond outputs by emphasizing the importance of addressing inherent biases in data and models to prevent the amplification of unfairness, a dimension absent from the National Policy Framework for Artificial Intelligence (Legislative Recommendations).
Synthesised from the 2 statutory definitions below. Each one is quoted in full, with its source.
What it triggers
- In New Zealand's Responsible AI in action: Guidance for Regulators, it recommends addressing inherent biases in data and models to prevent amplification of unfairness.source
Definitions (2)
Refers to systemic and repeatable errors in an AI system's output that create unfair outcomes, such as discrimination against certain demographic groups.
Refers to systematic and repeatable errors in a computer system that create unfair outcomes, such as privileging one arbitrary group of users over others. The guidance highlights the importance of addressing inherent biases in data and models to prevent amplification of unfairness.
Questions and answers
What does bias in AI mean under New Zealand's Responsible AI guidance?
Bias in AI refers to systematic and repeatable errors in a computer system that create unfair outcomes, such as privileging one arbitrary group of users over others. The guidance highlights the importance of addressing inherent biases in data and models to prevent amplification of unfairness.
How is AI bias defined in the National Policy Framework for Artificial Intelligence?
It refers to systemic and repeatable errors in an AI system's output that create unfair outcomes, such as discrimination against certain demographic groups.
What is the purpose of New Zealand's Responsible AI in Action guidance?
The guidance guides senior leadership at regulatory agencies on safely integrating artificial intelligence into public decision-making.