Ethics and bias in AI in Intro to Computing - Flowchart & Logic Diagram

Process Overview

This flowchart explains how ethics and bias can affect AI systems and the steps taken to identify and reduce bias to ensure fair and responsible AI use.

Flowchart

Collect Data

Yes No↓

Train AI Model

↓

Evaluate Model

↓

Adjust Data or Model

↓

Retrain AI Model

↓

Evaluate Model

↓

Repeat Adjustment

This flowchart shows the step-by-step process of creating an AI system while checking for ethical concerns and bias. It starts with data collection, checks if the data is diverse, trains the AI, evaluates it for bias, and if bias is found, adjusts and retrains until bias is minimized before deployment.

Step-by-Step Trace - 10 Steps

Step 1: Collect data from various sources.

Step 2: Check if the data is diverse.

Step 3: Since data is not diverse, decide to improve it.

Step 4: Adjust data to include missing groups.

Step 5: Train AI model using improved data.

Step 6: Evaluate AI model for bias.

Step 7: Adjust model or data to reduce bias.

Step 8: Retrain AI model with adjustments.

Step 9: Re-evaluate AI model for bias.

Step 10: Deploy AI system for real-world use.

Diagram

 +-----------------+      +-----------------+      +-----------------+
 |                 |      |                 |      |                 |
 |   Raw Data      | ---> |  Data Cleaning  | ---> |  Balanced Data  |
 |                 |      |                 |      |                 |
 +-----------------+      +-----------------+      +-----------------+
          |                        |                        |
          v                        v                        v
 +-----------------+      +-----------------+      +-----------------+
 |                 |      |                 |      |                 |
 |  Train AI Model | ---> | Evaluate Bias   | ---> |  Bias? (Yes/No) |
 |                 |      |                 |      |                 |
 +-----------------+      +-----------------+      +-----------------+
                                                          |
                                                Yes <-----+-----> No
                                                          |           
                                                          v           
                                               +-----------------+  
                                               |                 |  
                                               | Adjust Data or   |  
                                               | Model & Retrain  |  
                                               |                 |  
                                               +-----------------+  

This diagram shows how raw data is cleaned and balanced before training the AI. After training, the AI is evaluated for bias. If bias is found, data or model adjustments are made and the AI is retrained until bias is minimized.

Flowchart Quiz - 3 Questions

Why is it important to check if the data is diverse before training an AI model?

A To reduce the size of the data

B To make the training faster

C To ensure the AI learns fairly about all groups

D To avoid using any data at all

Key Result

Ensuring AI fairness requires checking data diversity, detecting bias, and adjusting the model before deployment.

Practice

1. What does ethics in AI mainly focus on?

easy

A. Making fair and safe choices in AI systems

B. Improving AI speed and performance

C. Adding more features to AI software

D. Reducing the cost of AI development

Solution

1. Step 1: Understand the meaning of ethics in AI

   Ethics in AI means making decisions that are fair and safe for everyone.

2. Step 2: Compare options with the definition

   Only Making fair and safe choices in AI systems talks about fairness and safety, which matches ethics in AI.

3. Final Answer:

   Making fair and safe choices in AI systems -> Option A

4. Quick Check:

   Ethics = Fair and safe choices [OK]

Hint: Ethics means fairness and safety in AI decisions [OK]

Common Mistakes:

• Confusing ethics with technical improvements
• Thinking ethics is about cost or speed
• Ignoring fairness and safety aspects

2. Which of the following is the correct way to describe bias in AI?

easy

A. A feature that makes AI faster

B. Unfairness that sneaks into AI from data or design

C. A tool to improve AI accuracy

D. A method to reduce AI errors

Solution

1. Step 1: Define bias in AI context

   Bias means unfairness that can come from the data or how AI is designed.

2. Step 2: Match options with bias definition

   Unfairness that sneaks into AI from data or design correctly describes bias as unfairness sneaking into AI.

3. Final Answer:

   Unfairness that sneaks into AI from data or design -> Option B

4. Quick Check:

   Bias = Unfairness in AI [OK]

Hint: Bias means unfairness in AI data or design [OK]

Common Mistakes:

• Thinking bias improves AI
• Confusing bias with speed or error reduction
• Ignoring the unfairness aspect

3. Consider this simple AI decision flow:

Input: Age
If Age < 18 then Output = "Minor"
Else Output = "Adult"

What is the output if the input age is 20?

medium

A. "Minor"

B. "Unknown"

C. "Error"

D. "Adult"

Solution

1. Step 1: Check the input value against the condition

   Input age is 20, which is not less than 18.

2. Step 2: Follow the decision flow for age 20

   Since 20 ≥ 18, output is "Adult".

3. Final Answer:

   "Adult" -> Option D

4. Quick Check:

   Age 20 ≥ 18 -> Output = "Adult" [OK]

Hint: Check if age is less than 18 or not [OK]

Common Mistakes:

• Choosing "Minor" for age 20
• Assuming error without reason
• Ignoring the else condition

4. An AI system shows bias by unfairly favoring one group. Which fix helps reduce this bias?

medium

A. Use more diverse and balanced data for training

B. Make the AI run faster

C. Add more layers to the AI model

D. Ignore the bias since it is small

Solution

1. Step 1: Identify cause of bias

   Bias often comes from unbalanced or non-diverse training data.

2. Step 2: Choose the fix that addresses data bias

   Using diverse and balanced data helps AI treat all groups fairly.

3. Final Answer:

   Use more diverse and balanced data for training -> Option A

4. Quick Check:

   Diverse data reduces bias [OK]

Hint: Fix bias by improving training data diversity [OK]

Common Mistakes:

• Thinking speed or model size fixes bias
• Ignoring bias problems
• Assuming bias is acceptable if small

5. A company wants to build an AI hiring tool that avoids bias. Which approach best ensures fairness?

hard

A. Use AI to select candidates faster without checks

B. Train AI only on resumes from one city

C. Regularly test AI decisions for fairness and update data

D. Exclude all personal information from resumes

Solution

1. Step 1: Understand fairness in AI hiring

   Fair AI must be tested regularly to catch bias and updated with good data.

2. Step 2: Evaluate options for fairness

   Regularly test AI decisions for fairness and update data includes testing and updating, which helps maintain fairness over time.

3. Step 3: Consider why other options fail

   Train AI only on resumes from one city limits data diversity; C ignores fairness; D may help but alone is not enough.

4. Final Answer:

   Regularly test AI decisions for fairness and update data -> Option C

5. Quick Check:

   Testing and updating ensures fairness [OK]

Hint: Fair AI needs ongoing testing and data updates [OK]

Common Mistakes:

• Using limited data causing bias
• Ignoring fairness checks
• Assuming removing info alone fixes bias