AI and Social Justice Part 1 and 2

To ensure that AI is used to promote social justice: teach ethics to data scientists

1. Selecting Training Data: Recognize and address historical biases in training data (EX: examine whether hiring data reflect historical discrimination against women)

2. Evaluating AI Systems: Examine whether AI systems produce unfair outcomes(EX: check whether a hiring algorithm systematically disadvantages certain groups)

Benefits:

• Raises awareness of ethical issues

• Encourages responsible decision-making

• Helps data scientists consider the social consequences of their work

Is teaching ethics enough

Not necessarily

Philosopher Green’s Alternative: Data Science as Political Action

Technical approach:

Data scientists are primarily trained to solve technical problems.

They may approach ethical problems as technical problems

They may also try to address these problems primarily through technical solutions

EX: Does ChatGPT Understand Language (understanding reference and semantics)

Removing sensitive information from training data can be technically difficult

Ethical Checklists:

Ethics may be reduced to a set of checklists

Meeting ethical requirements does not necessarily mean addressing deeper social and structural issues

Ethical issues are more complex than a checklist can capture

• equal treatment

• correcting past inequalities

Ethics Washing

Tech companies may use ethical guidelines and checklists to present themselves as ethical

but they may avoid making meaningful structural changes

EX: Microsoft lays off an ethical AI team as it doubles down on OpenAI

Political

Thinking about how society should be organized and what kind of society we should create

Doesn’t mean supporting a political party

Having ideas about what constitutes the social good

Data scientists are political actors

Their work can shape society and influence what kind of society we live in. They should have an idea of what kind of society they are helping to create

Five Technological Revolutions

1. Industrial Revolution

2. The Age of Steam and Railways

3. The Age of Steel, Electricity, and Heavy Engineering

4. The Age of Oil, Automobiles, and Mass Production

5. The Age of Information and Telecommunications

AI transformed society: Work and economy

AI may transform how we work and reshape the job market

AI transformed society: Education and Knowledge

AI may transform how we learn, teach, and produce knowledge

AI transformed society: Social and Political Life

AI may transform how decisions are made and how resources and opportunities are distributed

Objection 1: Data scientists should understand their work as political action

AI systems are tools. Tools can be used for good or bad purposes, and data scientists can’t always predict how they will be used.

• Tools are moral agents

• The moral consequences depend on how people use them

• Data scientists cannot fully predict or control how people use them

• Therefore, data scientists should not be held responsible for how others use the AI systems and should not be political actors

Green’s reply: technology shapes how society is organized and influences people’s behaviors; therefore, data scientists can’t separate their work from its social and political consequences

Objection 2: technology should be politically neutral

Even if technology affects society, engineers/ data scientists should try to keep their work politically neutral.

• Being a political actor does not necessarily mean supporting a political party

Green’s Reply: complete neutrality is impossible

• Our decisions are influenced by our values

• Remaining neutral can reinforce the status quo

EX: AI Hiring

• Historical hiring data reflects gender discrimination; an AI trained on that data may reproduce existing inequalities

Objection 3: Good Enough is Good Enough

AI already improves people’s lives, even if it isn’t perfect. We shouldn’t let the perfect be the enemy of the good.

Green’s Reply: Even if AI is good enough for some people.. is it good for everyone

Technological progress

developing more advanced technologies

Social progress

involves improving people’s lives and promoting social justice

Why should Data Scientist be political actors

1. AI has huge potential to transform society

2. Political neutrality is problematic

3. Data scientists play a role in shaping AI’s social impact

A challenge to green:

• social and political consequences of AI can’t be fully

• we may need to deploy to discover these consequences

• we cant expect data scientists to fully understand or address all the social and political consequences of AI in advance