: She actively promotes the democratization of AI and greater diversity within the technical workforce.
Model rejects women candidates due to historical bias in training data (previous hires were male). No traceability. Legal liability.
: Practical guidance on using Microsoft FairLearn , Google What-If Tool , IBM AI 360 Fairness , and Aequitas . heather dawe responsible ai in the enterprise
Heather Dawe argues that Responsible AI (RAI) is not a compliance checkbox or a set of abstract ethical principles, but a that must be embedded into the software development lifecycle (SDLC). Her core thesis is that without engineering rigour, AI ethics fails. This report outlines her key pillars: moving from principles to practice, the necessity of interdisciplinary teams, the role of continuous monitoring, and the governance of data provenance.
| Pillar | Enterprise Implementation | Key Question | |--------|--------------------------|---------------| | | Pre-deployment bias testing (e.g., using AIF360 , Fairlearn ); disparate impact analysis per demographic segment. | Does the model treat similarly situated individuals similarly? | | Explainability | Mandatory interpretability methods (LIME, SHAP, counterfactuals) for any model affecting customers/employees. | Can we justify this decision in plain language to a user? | | Traceability | Version control for data, code, and model hyperparameters; full data lineage from source to prediction. | Under what exact conditions was this prediction made? | | Observability | Live monitoring for concept drift, data drift, and prediction skew; automated alerts for fairness violations. | Is the model behaving in production as it did in testing? | : She actively promotes the democratization of AI
Dawe’s framework for responsible AI focuses on moving beyond abstract principles toward practical, hands-on risk management. In the enterprise setting, this involves several critical components: Go to product viewer dialog for this item. Responsible AI in the Enterprise
An enterprise builds an AI hiring tool to screen CVs. Legal liability
Dawe emphasizes using established tools from major cloud providers to operationalize these principles:
A valid, fair, and defensible system.