AI 伦理与偏见
AI Ethics and Bias
Throughout this programme, we have explored how data, algorithms, and neural networks come together to create intelligent systems. But intelligence without responsibility can cause real harm. In this final lesson, we examine the human side of AI - the biases it inherits, the ethical dilemmas it raises, and what we can all do about it.
What Is AI Bias?
AI bias occurs when a system produces results that are systematically unfair to certain groups of people. The AI is not deliberately prejudiced - it simply reflects the patterns in its training data and the assumptions of its designers.
Real-World Examples
Amazon's Hiring Tool (2018) Amazon built an AI to screen job applications. It was trained on CVs submitted over the previous ten years - a period when the tech industry was overwhelmingly male. The AI learned to penalise CVs that contained the word "women's" (as in "women's chess club") and downgraded graduates from all-women's universities. Amazon scrapped the tool.
Facial Recognition Failures Research by Joy Buolamwini at MIT found that commercial facial recognition systems had error rates of up to 34.7% for darker-skinned women, compared to just 0.8% for lighter-skinned men. The training data simply did not represent all faces equally.
Why did Amazon's AI hiring tool discriminate against women?
Where Does Bias Come From?
Bias can enter an AI system at every stage:
- Data collection - If the data over-represents one group, the model learns to favour that group.
- Labelling - Human annotators bring their own unconscious biases when tagging data.
- Feature selection - Choosing which variables to include (or exclude) can embed assumptions.
- Evaluation - If we only test on certain demographics, we miss failures on others.
AI does not create bias from thin air. It amplifies the biases already present in human decisions, historical records, and societal structures. The data is a mirror - and sometimes we do not like what it reflects.
Discussion
Sign in to join the discussion