Bias and fairness in AI systems
Learn where AI bias comes from, how it shows up, and what can realistically be done about it
A taste of a lesson
Couldn't they fix a biased hiring model by just deleting the gender column?
That is the first fix most people try, and it usually fails. Other features act as proxies: names, certain clubs or sports, career gaps, even word choices in a CV can correlate with gender. A model trained on past hiring decisions that favoured men can rediscover the pattern through those proxies. This is roughly what happened with one widely reported recruiting tool. Better approaches test outcomes by group directly and examine the training labels themselves. Exercise: name three CV features that might act as proxies for gender.
Written by the teacher as an example. In your lesson the tutor answers your own questions, and like any AI it can be wrong.
What you will be able to do
- Trace how bias can enter an AI system at each stage from data to deployment
- Describe major documented cases accurately and what each teaches
- Explain the main fairness definitions and why they can conflict
- Evaluate mitigation options and their limits for a given system
Lesson plan
- 1 Where bias comes from Map the stages of an AI system where bias can enter. Start
- 2 Proxies and the 'just remove it' myth See why dropping protected attributes does not remove discrimination. Start
- 3 Cases that shaped the debate Learn what well known cases do and do not show. Start
- 4 Defining fairness Understand common fairness definitions and their trade offs. Start
- 5 Bias in generative AI Recognise how text and image generators reproduce stereotypes. Start
- 6 What can be done Evaluate realistic mitigations, including the option not to deploy. Start
Try asking
About this tutor
For professionals, students and curious citizens who want a solid understanding of bias in AI beyond headlines. You trace bias through the whole pipeline: who is in the data, how labels are made, what the model optimises, how it is deployed and how people respond to it. You study well known cases in hiring, criminal justice, health and image models, and learn the main ideas of fairness, including the fact that different definitions can conflict. You also see why language and image generators reproduce stereotypes and what mitigation can and cannot achieve. You finish able to ask sharp questions about any AI system that affects people.
Reviews
4.7
3 ratingsSample
- Nadia O.Sample
The proxy lesson was a revelation. Using the loan form exercise, I found two proxies in a scoring sheet at my own workplace.
- Yaw D.Sample
The point that fairness definitions conflict mathematically changed how I read every news story about algorithmic bias.
- Felix W.Sample
Accurate and balanced on COMPAS, which I had only heard about one sided before. Slightly heavy on cases, light on practical mitigation steps.
About the teacher
I teach AI ethics as practical judgment: privacy, fairness and accountability you can act on
9 tutors 327 lessons taught Sample
I teach the ethics of AI as something you do, not something you recite. My lessons cover privacy and personal data, bias and fairness, explainability and the effects of AI on work and high stakes decisions. I use real cases and simple methods so learners can reason through a new situation on their own. My background combines data analysis with...
See Aisha's profile and tutorsMore like this
Other tutors on the same or nearby topics.