Category
Ethical Dilemmas & Applied Ethics
15 free guides in this category.
Algorithmic Fairness & AI Ethics
Bias Mitigation Strategies
Techniques to reduce unfair discrimination in algorithmic decision-making systems through data preprocessing, model training adjustments, or output correction.
Read guideCausal Reasoning in Fairness
Using causal models and counterfactual reasoning to identify and eliminate algorithmic discrimination by modeling how protected attributes actually influence decisions.
Read guideDemographic Parity
A fairness metric requiring an algorithm's outcomes to be equally distributed across different demographic groups, regardless of their underlying qualifications.
Read guideEqualized Odds & Opportunity
Examines how to ensure an algorithm's prediction errors, like false approvals and false rejections, are balanced across different demographic groups.
Read guideFairness-Constrained Models
A method for building AI systems that are explicitly designed to minimize biased outcomes by balancing predictive accuracy with fairness goals.
Read guideFairness Through Unawareness
The approach of assuming algorithms are fair if they don't use protected attributes like race or gender, despite potential hidden biases in correlated data.
Read guideGroup vs. Individual Fairness
Two competing approaches to measuring fairness in AI systems: ensuring similar outcomes across demographic groups versus treating similar individuals equally regardless of group identity.
Read guideIntersectional Fairness
AI fairness that considers how multiple identities (race, gender, age, disability) combine to create unique forms of algorithmic harm.
Read guideLong-Term Fairness Impacts
How algorithmic decisions shape fairness outcomes over time through feedback loops that amplify, compound, or potentially reduce biases in society.
Read guideModel Interpretability Methods
Techniques for understanding how AI systems reach decisions, making opaque algorithms transparent and accountable.
Read guideReweighing for Fairness
A technique that adjusts training data weights to reduce discrimination by balancing representation across groups before model training begins.
Read guideSocio-Technical Systems
Systems where people and technology are deeply interconnected—changing one always affects the other.
Read guideAlgorithmic Bias Sources
Algorithmic bias originates from flawed training data, design choices, and societal inequalities embedded in AI systems.
Read guideBlack Box Problem
The challenge of understanding how AI systems make decisions when their internal reasoning is invisible to humans.
Read guideAlgorithmic Bias
Understand how automated systems can produce unfair or discriminatory outcomes, even when designed with good intentions.
Read guide