Why AI sometimes makes confident mistakes in AI for Everyone - See It in Action

Why AI Sometimes Makes Confident Mistakes

📖 Scenario: You are learning about how artificial intelligence (AI) systems work and why they sometimes give answers that sound very sure but are actually wrong.Imagine you are explaining this to a friend who is curious about AI but does not know much about it.

🎯 Goal: Build a simple explanation using a list of reasons why AI can make confident mistakes. You will create a list of reasons, add a threshold for confidence, filter the reasons based on confidence, and then finalize the explanation.

📋 What You'll Learn

Create a list called reasons with exact entries about AI mistakes and confidence

Create a variable called confidence_threshold with the value 0.7

Use a list comprehension to create a new list confident_mistakes that includes reasons with confidence above the threshold

Add a final statement variable called summary explaining the concept

💡 Why This Matters

🌍 Real World

Understanding why AI sometimes makes confident mistakes helps people trust and use AI wisely in daily life.

💼 Career

This knowledge is important for anyone working with AI, from developers to managers, to explain AI behavior clearly.

1

Create the list of reasons

Create a list called reasons with these exact dictionaries: {'reason': 'Limited training data', 'confidence': 0.8}, {'reason': 'Ambiguous input', 'confidence': 0.9}, {'reason': 'Model bias', 'confidence': 0.6}, {'reason': 'Overgeneralization', 'confidence': 0.75}, {'reason': 'Lack of common sense', 'confidence': 0.65}

AI for Everyone

# Create the list called reasons with the exact dictionaries
# Your code here

Hint

Use a list with dictionaries. Each dictionary has keys 'reason' and 'confidence' with the exact values.

2

Set the confidence threshold

Create a variable called confidence_threshold and set it to 0.7

AI for Everyone

reasons = [
    {'reason': 'Limited training data', 'confidence': 0.8},
    {'reason': 'Ambiguous input', 'confidence': 0.9},
    {'reason': 'Model bias', 'confidence': 0.6},
    {'reason': 'Overgeneralization', 'confidence': 0.75},
    {'reason': 'Lack of common sense', 'confidence': 0.65}
]
# Create confidence_threshold variable and set it to 0.7
# Your code here

Hint

Just create a variable with the exact name and value.

3

Filter reasons by confidence

Use a list comprehension to create a new list called confident_mistakes that includes only the reason strings from reasons where the confidence is greater than confidence_threshold

AI for Everyone

reasons = [
    {'reason': 'Limited training data', 'confidence': 0.8},
    {'reason': 'Ambiguous input', 'confidence': 0.9},
    {'reason': 'Model bias', 'confidence': 0.6},
    {'reason': 'Overgeneralization', 'confidence': 0.75},
    {'reason': 'Lack of common sense', 'confidence': 0.65}
]
confidence_threshold = 0.7
# Create confident_mistakes list using list comprehension
# Your code here

Hint

Use a list comprehension with the condition on confidence and extract the reason string.

4

Add a summary explanation

Create a variable called summary and set it to the string 'AI can make confident mistakes because it relies on patterns in data, which can be limited or biased.'

AI for Everyone

reasons = [
    {'reason': 'Limited training data', 'confidence': 0.8},
    {'reason': 'Ambiguous input', 'confidence': 0.9},
    {'reason': 'Model bias', 'confidence': 0.6},
    {'reason': 'Overgeneralization', 'confidence': 0.75},
    {'reason': 'Lack of common sense', 'confidence': 0.65}
]
confidence_threshold = 0.7
confident_mistakes = [r['reason'] for r in reasons if r['confidence'] > confidence_threshold]
# Create summary string explaining why AI makes confident mistakes
# Your code here

Hint

Assign the exact string to the variable summary.

Practice

1. Why can AI sometimes make mistakes even when it seems very confident?

easy

A. Because it only answers questions it knows perfectly

B. Because it always checks facts before answering

C. Because it never learns from data

D. Because it guesses based on patterns, not understanding true facts

Solution

1. Step 1: Understand how AI guesses answers

   AI looks for patterns in data it has seen before to make guesses.

2. Step 2: Recognize the difference between confidence and correctness

   Confidence means the guess fits known patterns well, not that it is always right.

3. Final Answer:

   Because it guesses based on patterns, not understanding true facts -> Option D

4. Quick Check:

   AI confidence = pattern matching, not guaranteed truth [OK]

Hint: Confidence means pattern match, not always correct [OK]

Common Mistakes:

• Thinking AI always checks facts before answering
• Believing AI understands facts like humans
• Assuming AI never makes mistakes

2. Which statement correctly describes AI confidence in its answers?

easy

A. AI confidence means it has verified the answer is true

B. AI confidence means the answer fits patterns in its training data

C. AI confidence means the answer is always 100% accurate

D. AI confidence means the answer is random

Solution

1. Step 1: Define AI confidence

   AI confidence shows how well an answer matches patterns learned from data.

2. Step 2: Compare options to this definition

   Only AI confidence means the answer fits patterns in its training data correctly states confidence relates to pattern fit, not truth verification.

3. Final Answer:

   AI confidence means the answer fits patterns in its training data -> Option B

4. Quick Check:

   Confidence = pattern fit, not truth check [OK]

Hint: Confidence = pattern fit, not truth check [OK]

Common Mistakes:

• Confusing confidence with truth verification
• Thinking confidence means 100% accuracy
• Believing AI answers randomly when confident

3. An AI model trained mostly on photos of cats and dogs confidently labels a photo of a fox as a dog. Why did this happen?

medium

A. The AI guessed based on similar patterns it learned from dogs

B. The AI randomly chose an answer

C. The AI recognized the fox perfectly as a dog

D. The AI has no confidence in its answer

Solution

1. Step 1: Understand AI training data influence

   The AI learned mostly from cats and dogs, so it knows those patterns well.

2. Step 2: Explain why it labeled a fox as a dog

   The fox looks similar to dogs, so AI matched it to the closest known pattern, dog.

3. Final Answer:

   The AI guessed based on similar patterns it learned from dogs -> Option A

4. Quick Check:

   AI matches closest known pattern = dog [OK]

Hint: AI matches new data to closest known pattern [OK]

Common Mistakes:

• Assuming AI perfectly recognizes all animals
• Thinking AI guesses answers randomly
• Believing AI has no confidence in wrong answers

4. An AI chatbot confidently gives a wrong answer about a recent event. What is the most likely cause?

medium

A. The AI intentionally gives wrong answers

B. The AI cannot generate confident answers

C. The AI has outdated or incomplete data about the event

D. The AI always knows recent events perfectly

Solution

1. Step 1: Identify AI data limitations

   AI only knows what data it was trained on, so recent events may be missing.

2. Step 2: Explain why AI gives wrong confident answers

   Without updated data, AI guesses based on older patterns, leading to confident mistakes.

3. Final Answer:

   The AI has outdated or incomplete data about the event -> Option C

4. Quick Check:

   Outdated data causes confident mistakes [OK]

Hint: Outdated data causes confident AI mistakes [OK]

Common Mistakes:

• Thinking AI intentionally lies
• Believing AI always knows recent events
• Assuming AI cannot be confident in wrong answers

5. You want to reduce confident mistakes by an AI that classifies emails as spam or not. Which approach helps most?

hard

A. Train the AI on a wide, updated dataset including new spam types

B. Make the AI guess randomly to avoid confidence

C. Limit the AI to only old spam examples

D. Ignore AI confidence scores completely

Solution

1. Step 1: Understand how training data affects AI accuracy

   AI learns patterns from its training data, so more diverse and current data improves guesses.

2. Step 2: Choose the best method to reduce confident mistakes

   Training on updated data helps AI recognize new spam types and avoid wrong confident answers.

3. Final Answer:

   Train the AI on a wide, updated dataset including new spam types -> Option A

4. Quick Check:

   Updated data reduces confident mistakes [OK]

Hint: Use updated, diverse data to reduce confident mistakes [OK]

Common Mistakes:

• Thinking random guessing improves AI
• Using only old data limits AI accuracy
• Ignoring confidence scores without improving data