AI ethics note

1.Ethics VS. Law

Ethics are the moral principles that govern a person or society’s behavior Laws are systems of rules that govern a society’s conduct, often enforced by a controlling authority • Laws are often informed by ethics, but not all ethical behaviors are codified in the law • The ethical effects of technology (like social media or AI) can take decades to understand, let alone translate to law

Unethical Behaviors Enforced by Law: Murder, Theft, Fraud, Reckless Driving

Unethical Behaviors NOT Enforced by Law: Lying, Being unfair to employees or children, Taking credit for others’ work, Eavesdropping or reading private messages.

SCENARIO: the trolley problem

A runaway trolley speeds down a track. If it stays on course, 5 people on the track will be hit and killed. By pulling a lever, you can divert the trolley to another track where a single bystander will be hit.

Do you pull the lever? • 90% of people say they would pull the lever, to save the most lives • But what if the one person was your spouse or best friend? What if it’s one child and five senior citizens?

SCENARIO:self-driven car

A self-driving car is facing a head-on collision with a reckless driver. If it stays on course, it estimates that the passenger and 4 people in the other car will die. If it veers off course, it will hit one person on the sidewalk.

How should the self-driving car be programmed to respond? • What type of algorithm would determine the “best” course of action? • Does the company that programmed the car decide how to solve this problem, or should there be laws that regulate how AI companies handle decisions like this?

2.Data Stewardship

Data stewardship is the practice of managing and using data ethically • Stewardship ultimately comes down to the golden rule: treat user data how you’d like your own to be treated • Ethical data stewardship requires a thoughtful approach to user consent, security, privacy and confidentiality

3.Data & Algorithmic bias

•What is bias?

Prejudice in favor of or against one thing, person, or group compared with another, usually in a way considered to be unfair.

•Types of bias

Sampling bias

Selection bias

Algorithmic bias

Confirmation bias