Lesson List Browse Q&A

Introduction

Background and Major Conflicts in AI Ethics

Discussion/Reflection Topics

Quiz

Human-led AI Development

Human Agency and Machine Agency

Discussion/Reflection Topics for Students 2

Human-centered Design Principles

Go to Course Home
Understanding When and When Not to Use AI

Overview

Artificial Intelligence (AI) can solve many problems. But it is not always the right tool. This lesson will help you:

  • Think critically about when AI might cause more harm than good.
  • Learn how to define a problem that AI can solve well.

Part 1: When AI Should Not Be Used

1. Ethical Concerns

  • Unfair Treatment: If an AI system might discriminate against certain groups, it should not be used.
  • Privacy Risks: If the system collects sensitive information without consent, it can hurt people’s privacy.

 

2. Safety Issues

  • High Stakes: In areas like healthcare or driverless cars, AI mistakes can cause serious harm.  There maybe cases where human decision making, or human-in-the-loop is necessary instead of automating the complete system with AI.
  • No Reliable Oversight: If you cannot monitor or control the AI properly, it may become unsafe.

 

3. Lack of Trust or Clarity

  • Opaque Decisions: If you cannot explain how an AI makes decisions, people may not trust it.
  • Insufficient Data: If there isn’t enough good data, the AI might not be accurate or fair.

 

4. Alternatives Are Better

  • Simple Solutions: Sometimes a simple system (like a basic computer program or a checklist) works just as well.
  • Human Expertise Needed: Some tasks need a human’s empathy or creativity that AI cannot replace.

Part 2: Defining a Problem for AI

1. Clearly Identify the Problem

  • Write It Down: Use short sentences to describe what needs solving.
  • Ask Questions: Is the problem about sorting objects, finding patterns, or making predictions?

 

2. Check If AI Is Suitable

  • Data Availability: Do you have the right data to train an AI?
  • Complexity: Is the problem too simple or too complex for AI to handle?

 

3. Set Clear Goals and Limits

  • Success Criteria: How will you know if the AI works?
  • Boundaries: What should the AI not do? Protect privacy, limit data collection, etc.

 

4. Think About Ethics and Impact

  • Fairness: Will the AI treat everyone equally?
  • Safety and Responsibility: Who will fix problems if the AI fails

Conclusion

  • Not every problem needs AI.
  • AI can be powerful, but it can also raise ethical and safety concerns.
  • Before using AI, always scope the problem carefully, ensure data is available, and consider the impact on people.

 

Use these ideas to make smart choices about when to use AI and how to design it responsibly.

← Previous Next →
Skip to toolbar