Summary of "L'éthique de l'IA - EP01 - C’est quoi l'éthique?"

Summary of L’éthique de l’IA - EP01 - C’est quoi l’éthique?

Main Ideas and Concepts

Introduction and Context

Joann Rochel, a teacher and researcher at EPFL Lausanne, introduces a video series focused on AI ethics. The series aims to explore the ethical questions raised by the use of artificial intelligence, particularly from an engineering perspective. A practical use case will be followed throughout the series: developing an AI tool for HR purposes that automates CV selection.

Definition of Ethics

Ethics is defined as a systematic analysis of the values guiding our choices and decisions in specific contexts.
Values refer broadly to personal convictions or principles that orient life decisions.
Ethics is not merely about labeling actions as good or bad but about seeking coherence and justified choices between alternatives.
Ethics is distinct from:
- Tradition: Long-standing practices are not inherently ethical without justification.
- Religion: Can be a source of ethical conviction but is separate from ethics itself.
- Law: Binding and enforced by institutions, unlike ethics which relies on personal or collective conviction.
Ethics operates at multiple levels:
- Individual ethics: Personal convictions.
- Organizational ethics: Shared values within a company or institution.
- Societal ethics: Collective values enshrined in constitutions or public norms.

Implications of Ethics in Engineering

No Neutral Technology: Every technological development involves value choices and trade-offs.
Ethics is Already Present: Engineers inherently make ethical decisions during development, often implicitly.
Ethics is Continuous: Ethical considerations arise throughout the innovation process, not just at the end.

Defining Artificial Intelligence (AI)

Defining AI is complex and involves choices beyond purely technical criteria.
Narrow definitions focusing on human-like intelligence or brain operations are limited.
A broader, contemporary definition includes current AI practices such as large language models.
Definitions have political and legal consequences, especially for regulation (e.g., the EU’s AI Act).
The EU AI Act defines AI systems as software using listed techniques to generate outputs (content, predictions, recommendations, decisions) based on human-defined objectives.
This dual approach combines technical methods and the effects or products generated by AI.
For this series, AI is defined as tools based on models (algorithms trained on large datasets) that achieve goals through statistical and mathematical operations.
AI tools in HR are considered high-risk under EU regulations due to the sensitive nature of their application.

Outline of Upcoming Content

The series will dissect ethical questions at different stages of AI tool creation, focusing on three main families of ethical questions:

Ethics in Technology: Ethical issues during tool development.
Social Justice: Impact on society and human groups.
Narrative Ethics: The stories and discourse surrounding technology.

The next video will focus on ethics in technology.

Methodology and Instructions Presented

Use a case study (AI tool for automated CV selection in HR) to explore ethical questions.
Adopt a working definition of ethics as systematic value analysis guiding choices.
Recognize ethics as distinct from tradition, religion, and law.
Understand that ethical considerations are embedded throughout the engineering process, not just at the end.
Use a working definition of AI that includes:
- Algorithms trained on large datasets,
- Statistical and mathematical operations,
- Producing outputs like recommendations or decisions.
Examine ethical questions in three categories:
1. Ethics during technology creation,
2. Social justice and societal impacts,
3. The narrative and discourse around technology.