When AI is wrong vs when AI is uncertain in AI for Everyone - Hands-On Comparison

Understanding When AI is Wrong vs When AI is Uncertain

📖 Scenario: You are learning how artificial intelligence (AI) systems make decisions and how to tell the difference between when an AI is wrong and when it is uncertain.Imagine you use a voice assistant or a chatbot that sometimes gives answers. Sometimes it confidently gives a wrong answer, and other times it says it is not sure. Understanding this helps you trust AI better.

🎯 Goal: Build a simple example that shows a list of AI answers with a label for each one saying if the AI is wrong or uncertain.This will help you see how AI can be wrong (confident but incorrect) or uncertain (not confident).

📋 What You'll Learn

Create a list of AI answers with their confidence scores and correctness

Add a confidence threshold to decide if AI is certain or uncertain

Use a loop to label each answer as 'Wrong' if incorrect and confident, or 'Uncertain' if confidence is low

Add a final summary count of wrong and uncertain answers

💡 Why This Matters

🌍 Real World

Understanding when AI is wrong or uncertain helps users trust AI systems and make better decisions when using AI tools like chatbots, voice assistants, or recommendation systems.

💼 Career

This knowledge is useful for AI developers, product managers, and anyone working with AI systems to improve AI transparency and user experience.

1

Create the AI answers data

Create a list called ai_answers with these exact dictionaries: {'answer': 'Paris', 'confidence': 0.9, 'correct': True}, {'answer': 'London', 'confidence': 0.6, 'correct': False}, {'answer': 'Berlin', 'confidence': 0.4, 'correct': True}, {'answer': 'Rome', 'confidence': 0.3, 'correct': False}

AI for Everyone

# Create the list ai_answers with the given dictionaries
# Your code here

Hint

Use a list with dictionaries. Each dictionary has keys 'answer', 'confidence', and 'correct'. Use True and False for correctness.

2

Set the confidence threshold

Create a variable called confidence_threshold and set it to 0.5 to decide when AI is confident.

AI for Everyone

ai_answers = [
    {'answer': 'Paris', 'confidence': 0.9, 'correct': True},
    {'answer': 'London', 'confidence': 0.6, 'correct': False},
    {'answer': 'Berlin', 'confidence': 0.4, 'correct': True},
    {'answer': 'Rome', 'confidence': 0.3, 'correct': False}
]
# Set confidence_threshold to 0.5
# Your code here

Hint

Just create a variable named confidence_threshold and assign 0.5 to it.

3

Label answers as Wrong or Uncertain

Create a list called labels. Use a for loop with variables item to go through ai_answers. For each item, if item['confidence'] >= confidence_threshold and item['correct'] == False, add the string 'Wrong' to labels. Otherwise, if item['confidence'] < confidence_threshold, add 'Uncertain' to labels. If none of these, add 'Correct'.

AI for Everyone

ai_answers = [
    {'answer': 'Paris', 'confidence': 0.9, 'correct': True},
    {'answer': 'London', 'confidence': 0.6, 'correct': False},
    {'answer': 'Berlin', 'confidence': 0.4, 'correct': True},
    {'answer': 'Rome', 'confidence': 0.3, 'correct': False}
]
confidence_threshold = 0.5
# Create labels list and fill it using a for loop
# Your code here

Hint

Use a for loop over ai_answers. Check confidence and correctness to decide the label. Append the label to labels list.

4

Count and summarize wrong and uncertain answers

Create two variables: wrong_count and uncertain_count. Use the count() method on labels to count how many times 'Wrong' and 'Uncertain' appear and assign these counts to the variables.

AI for Everyone

ai_answers = [
    {'answer': 'Paris', 'confidence': 0.9, 'correct': True},
    {'answer': 'London', 'confidence': 0.6, 'correct': False},
    {'answer': 'Berlin', 'confidence': 0.4, 'correct': True},
    {'answer': 'Rome', 'confidence': 0.3, 'correct': False}
]
confidence_threshold = 0.5
labels = []
for item in ai_answers:
    if item['confidence'] >= confidence_threshold and item['correct'] == False:
        labels.append('Wrong')
    elif item['confidence'] < confidence_threshold:
        labels.append('Uncertain')
    else:
        labels.append('Correct')
# Count wrong and uncertain labels
# Your code here

Hint

Use labels.count('Wrong') and labels.count('Uncertain') to get the counts.