Explore the critical issue of algorithmic bias in AI and machine learning systems through this conference talk delivered at GOTO Copenhagen 2022. Delve into the origins of bias in algorithms, its widespread impact on decision-making processes, and strategies for addressing, avoiding, and mitigating these biases. Learn about heuristics, cognitive biases, and their influence on algorithmic design. Examine real-world examples of algorithmic bias and understand where bias enters the development process. Investigate concepts of fairness in algorithms, including trade-offs in protecting different groups or individuals. Discover techniques for combating bias, such as algorithmic auditing and screening methods. Gain insights into the challenges of addressing algorithmic bias and explore practical approaches to mitigate its effects in AI-driven decision-making systems.
Overview
Syllabus
Intro
Heuristics & biases
The cognitive bias codex
Algorithmic bias
Examples
Where does bias enter the algorithms?
Fair algorithms
Analysing trade-offs when choosing who to protect from algorithmic bias
Protecting groups, protecting individuals
Accuracy is fairness
Combating bias in algorithms
Screening algorithms for bias
What can we control?
Algorithmic auditing
Challenges with addressing algorithmic bias
Mitigating algorithmic bias
Outro
Taught by
GOTO Conferences
