Page 66 - CT_AI_Class-8

How Can We Make AI Fair?

One can make AI fair by following these methods:
 Bias check: It’s important to test AI programs to ensure they do not treat different groups or

categories unfairly. AI should be evaluated to see if it favours certain groups over others.
 Human supervision: Human beings should review AI decisions rather than trusting the system
blindly. This ensures that any errors or biases in AI can be caught and corrected.

 Transparency: It’s crucial to understand how an AI system works and how it makes its decisions.
The more we know about the AI’s processes, the better we can ensure it operates fairly.

Many websites and apps depend on AI to recommend content that users are likely to enjoy. These
recommendations are made by studying user behaviour, including watched videos, liked posts
and shared content. Based on this, it recommends similar videos that are popular. However, when
the system suggests videos to new users, it may mostly recommend English or Hindi language
videos because these languages have very high total view counts. As a result, videos in smaller

regional languages might not appear as often in the recommendations. Some of the questions
that arises in mind are as follows:
 Do you think this could impact creators who produce videos in regional languages?

 What steps can the platform take to make sure that content in different languages gets equal
and fair representation?

Activity: Supervised Learning & Bias
The steps to use the Google Teachable Machine are as follows:

Step 1 Open the website https://teachablemachine.withgoogle.com/.

Step 2 Click on Get Started and select Image Project.
Step 3 Choose the Standard image model option to create an image classification project.

Step 4 In the first class, rename the category as Lion.

Step 5 Click on Upload, then select Import images from Google Drive.
Step 6 Open the Teachable Machine Lion-Bear Dataset saved in your Google Drive or computer.

Step 7 Open the Lion folder and select all the images from the training dataset folder at once.
Click Select to upload them.

Step 8 Create a second class and rename it as Bear.
Step 9 Repeat the same process for the Bear images by uploading all images from the training
dataset folder.

Step 10 After uploading both datasets, click on the Train Model button.

Wait for the AI model to complete the training process.

Once training is complete, test the model by uploading new lion or bear images and observe how
the AI predicts the correct category.

64 Artificial Intelligence (CT & AI)-VIII