Discrimination

Accountability

Privacy

Inequity

AI systems are often used to allocate opportunities, such as jobs or education. AI may be trained to have biases on certain races/genders etc. Some think that just removing these physical attributes may solve the problem. A model can ignore race explicitly and still reconstruct it through proxies such as ZIP code, school history, browsing patterns, language style, purchase history, or social network structure. 

When AI inevitably causes harm and pain, whose fault is it and who has the power to prevent it? Accountability means that someone must be responsible when an AI system causes harm. A company, developer, government, or user cannot simply blame “the algorithm.” There must be clear responsibility for how the AI is trained, tested, deployed, monitored, and corrected.

Data privacy has always been an essential issue in the age of internet. Privacy in AI means protecting people’s personal data from being collected, used, shared, or exposed without proper consent. Since AI systems often need large amounts of data, privacy becomes a major issue when the data includes things like location, messages, medical records, faces, voices, search history, or personal preferences.

Inequity means AI benefits some groups more than others or worsens existing social inequalities. For example, wealthy schools may get better AI tutoring tools while underfunded schools get weaker systems, or AI hiring tools may favor applicants from elite backgrounds. Inequity is broader than discrimination: discrimination is unfair treatment, while inequity is an unfair distribution of benefits, risks, and opportunities.