Develop
Identify and address ethical concerns as they arise
Develop the technology while implementing mechanisms for ongoing feedback, audits, and assessments to identify and address ethical concerns as they arise.
Continuous Ethical Evaluation
Assess and improve the AI system's ethical performance during development. Implement mechanisms for ongoing feedback, audits, and assessments to identify and address ethical concerns as they arise.
ICO AI and Data Protection Risk Toolkit
The UK Information Commissioner’s Office has an ICO AI and data protection risk toolkit, that is a practical excel spreadsheet mapping each of the stages in the AI Lifecycle to risk assessment recommendations, proposed controls, and even suggesting practical steps to reduce the risk. As it is UK based it will need modification to the New Zealand context (it follows the UK GDPR articles), but is an extremely simple and useful tool for ensuring you are taking a structured approach to your AI projects.
Cognitive Bias Codex
The Cognitive Bias codex Infographic is a handy visual tool that organises known biases (around 180) in a meaningful way. Cognitive biases can be a problem for humans and AI technologies alike and understanding them is the first step toward identifying biases in your data, models our outputs.
Microsoft AI fairness checklist
A tool to help developers identify and address potential fairness issues in their AI systems (https://www.microsoft.com/en-us/research/project/ai-fairness-checklist/)
AI risk management
offers a path to minimise potential negative impacts of AI systems, such as threats to civil liberties and rights, while also providing opportunities to maximise positive impacts. Addressing, documenting, and managing AI risks and potential negative impacts effectively can lead to more trustworthy AI systems.
The NIST AI Risk Management Framework (AI RMF 1.0) is designed to be voluntarily used and equip organisations and individuals with approaches that increase the trustworthiness of AI systems and help foster the responsible design, development, deployment, and use of AI systems over time. It is thorough, which may be too time or resource costly for a robust organisation’s situation, however it is valuable to understand NIST best practices. See here for a summary of how the NIST AI RMF is structured https://securiti.ai/nist-ai-risk-management-framework/
