# Session 1 Handout: Fair and Accountable AI: Ethics and Bias

<p class="meta">FAST-17 Responsible AI for All Staff · 7formations · Instructor: Wadood Chaudhary · All examples fictional · Not legal advice · Suggested reading: Svekis, AI Literacy for All Employees (2026), chapters 8 and 9</p>

## Why this session matters

Every employee who uses AI makes small ethical decisions: what to trust, what to paste, whether a result treats people fairly, and what to do when something looks wrong. You do not need to be a manager or a lawyer to make those calls well. You need a few principles, a quick check, and to know who to tell.

## Key concepts

### 1. Four principles of responsible AI

| Principle | In everyday terms |
| --- | --- |
| Fairness | People are not treated worse because of who they are |
| Accountability | A person or team owns the outcome, not the tool |
| Transparency | People know when AI affects them and can question it |
| Privacy | Personal information is protected and used only as needed |

### 2. The five-step ethical check

1. **Stakeholders:** who is affected?
2. **Risks:** what could go wrong, and for whom?
3. **Transparency:** do they know AI is involved? Can they question it?
4. **Oversight:** who reviews, and who can override?
5. **Compliance:** which law or policy applies?

### 3. How bias gets in

| Type | What happens | Example (fictional) |
| --- | --- | --- |
| Historical | Data reflects past unequal treatment | Résumé model trained on mostly male past hires |
| Representation | Some groups are missing or rare in the data | Voice tool fails on strong regional accents |
| Measurement | The thing measured is a poor stand-in | Call quality judged by call length |
| Evaluation | Tested on the wrong people or conditions | Safety detector tested only on the day shift |

Removing a protected attribute does not always remove bias: other fields (zip code, availability flags, career gaps) can stand in for it.

### 4. The law in Massachusetts (not legal advice)

- **G.L. c. 151B** prohibits discrimination in employment. A tool does not change who is responsible.
- The **Massachusetts Attorney General's advisory (April 2024)** says existing consumer protection, anti-discrimination, and data security laws apply to AI.
- Elsewhere, **NYC Local Law 144** requires bias audits for automated hiring tools, a sign of where rules are going.

## Worked example

North Shore Builders' (fictional) AI scheduler never offers weekend overtime to workers with a childcare flag. The flag acts as a stand-in for family status. The right response: write down examples with dates, report through the named channel, and do not quietly change the data yourself. The company, not the software, is accountable.

## Try this before session 2

- Run the five-step check on one AI use you have seen at work.
- Read Svekis chapters 8 and 9.