Resource Planning

  1. Checklist / Questions to ask your team

    • Did you do a technology assessment to determine if the problem being solved needs AI technologies. Acting on predictions in an automated way should should only be used when risk is low. The following uses case best suited for AI technologies to predicting recommendations based on different data subjects, predicting future events eg. weather, natural language understanding, and image recognition. Other uses cases such as people don’t want this task automated, should be reconsidered more information when to use AI or not User Needs chapter

    • Did you create an Impact Statement which frames the problem to be solved and includes Risk Assessment chart with measurable thresholds for bias, fairness, transparency, Data Gathering, Data privacy, Training Data, Explainability, List key potential harms)

    • Did you create an AI BOM (Bill of Material) which includes: list of all components ( data, models and code)

    • What Ethics by Design approach did you use eg. Scoping, Mapping, Artifact Collection, Testing, and Reflection (SMACTR)

    • Did you create a Risk Management chart to include list of intent ‘harm/risk, probability and planned mitigation if mistakes are made eg. data drift occurs

    • Do you have an AI Ethics board in place to review and govern the ethical development and deployments

    • Are your teams trained on ethical deployment of AI (annual certification needed)

    • What standards / regulation are to be adhered to

  2. External Resources - Tools to use

    • Bias There are tools from IBM and Microsoft that evaluate this.

    • Data Drift: eg Dataiku those that help manage data drift or individual prediction explanations

    • Impact Assessment tool

  3. Case Studies

  4. Further Readings