Async review

Instruction and application

In Progress

Recap core topics:

• Unit 1: Ethical considerations in AI and machine learning

Unit 1: Ethical considerations in AI and machine learning

In Unit 1, you explored:

✨

Responsible AI development:

Ethical considerations in AI/ML model design, deployment and decision-making.

✨

Key ethical principles:

Fairness, accountability, transparency and explainability in AI-driven systems.

✨

Privacy and data protection:

Laws such as GDPR and CCPA, privacy-preserving techniques and balancing data utility with security.

✨

Ethical data collection:

Best practices for gathering, storing and managing data responsibly while minimising bias.

✨

Decision-making in AI/ML:

Navigating conflicts between business goals and ethical AI principles, ethics review processes and governance frameworks.

✨

AI governance frameworks:

Understanding global AI ethics standards.

Why do ethical frameworks matter in AI and ML?

• AI and ML systems must balance innovation with ethical responsibility to avoid harm and ensure trust.
• Ethical frameworks help guide data collection, model deployment and decision-making to prevent bias, privacy violations and unfair outcomes.

Key AI and ML ethical frameworks

FrameworkKey focusHow it helps AI/ML developmentUK Data EthicsGuidelines for ethical AI use in the public sector-Transparency: Clearly communicate AI decisions.

✨

Accountability:

Define responsibility for AI-driven decisions.

✨

Fairness:

Reduce bias and ensure equitable AI applications.

UK’s Ethics, Transparency and AccountabilityAutomated decision-making- Explainability: Make AI decisions interpretable.

✨

Impact assessment:

Analyse AI’s social, legal and ethical implications.

✨

Bias mitigation:

Identify and reduce algorithmic discrimination.

EU AI ActAI risk-based regulation- **Risk categorisation:**Determines AI governance based on application type.

✨

Compliance requirements:

Strictest regulations for high-risk AI (e.g. hiring, health care).

✨

Transparency and fairness:

Mandates auditability and user awareness.

IEEE AI Ethics and Governance StandardsGlobal AI best practices- **Trustworthiness:**Protects sensitive data.

✨

Fairness:

Prevents discrimination in AI decisions.

✨

Human-centred:

Ensures that AI benefits society.

Action item: Navigating AI and ML ethics poll

Let’s do a quick ethics-focused poll! This will help us gauge our understanding of key AI and ML ethical challenges in real-world applications. Just go with your best judgement!

A company is developing an AI system that analyses customer interactions to improve service recommendations. What is the most responsible approach to data collection?

Gather as much data as possible to maximise AI performance, ensuring it is stored securely.Provide clear information to users about data usage and allow them to opt in before collecting their data.Use third-party datasets that have been legally acquired, assuming compliance checks were done by the provider.Retain customer data indefinitely to ensure continuous AI model improvements.

Sign in to vote.

An AI model used for loan approvals has been flagged for disproportionately rejecting applications from a particular demographic. What should the company do?

Assume the model is objective since it was trained on real historical data.Remove demographic data from the dataset to prevent bias from influencing the model.Perform a fairness audit, retrain the model if necessary and document the bias mitigation process.Proceed with the current model unless external audits highlight compliance risks.

Sign in to vote.

A healthcare AI system relies on large datasets to improve diagnostic accuracy. What is the most ethical approach to managing and sharing sensitive patient data?

Apply privacy-preserving techniques, and comply with data protection regulations before sharing information.Share anonymised data across institutions, but retain original records for long-term analysis.Partner with external organisations to access additional medical data, as long as it benefits patient care.Collect and store all patient data centrally, ensuring future AI models have access to comprehensive training data.

Sign in to vote.

In progress