Introduction to Ethical AI
- Articulate the context and motivation for ethical AI. - Create an ethical AI perspective to understand the strengths and weaknesses of our thinking.
AI Ethics for Organizations
- Articulate the impact of bias and fairness on decision-making with AI.
- Identify and develop organizational ethical AI pipelines, guidelines, and frameworks.
- Articulate components of ethical governance initiatives.
Identifying Bias Towards Fairness
- Identify types of data and machine learning (ML) bias and where they are introduced.
- Identify harms in AI solutions.
- Define AI fairness problem statements and priorities.
- Apply methodologies for identifying data and AI models bias.
- Apply metrics for measuring AI bias and fairness.
Mitigating Bias Towards Fairness
- Identify bias and fairness with AI lifecycle phases and negative feedback loops.
- Assess the strengths and weaknesses of bias and fairness mitigation strategies and metrics.
- Implement mitigation strategies to improve fairness in AI models and solutions.
- Articulate considerations towards designing and building data and models with enhanced fairness.
Transparency, Trust, and Explainability
- Articulate elements of legal programs towards data privacy, AI security, and transparency.
- Articulate compliance metrics for responsible data governance.
- Articulate what explainability is in AI/ML and apply solutions.
- Communicate trust to customers and users of AI/ML systems using compliance and creating industry standard documentation. - Identify ethical AI auditing mechanisms.
Course Project: AI Ethics for Personalized Budget Prediction
Personalization is a central aspect of many core AI systems. In this project, learners will be tasked with designing an AI model for budget prediction and applying ethical AI considerations. They will use data exploration and bias and fairness measurement skills to complete the first phase of the project around harm quantification. They will then apply bias mitigation skills towards remediating the harm, and construct a model card articulating the ethical implications, quantitative analysis, and business consequences of the use case.
