AI ethics and responsible usage in Prompt Engineering / GenAI - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Understanding Bias in AI Models

Which of the following best describes a common source of bias in AI models?

A Using a training dataset that does not represent all groups fairly

B Choosing a model with too many layers

C Running the model on a slow computer

D Using a programming language that is outdated

💡 Hint

Think about what happens if the data the AI learns from is not balanced.

Bias often comes from training data that lacks fair representation of all groups, causing the AI to perform poorly or unfairly for some people.

❓ Metrics

intermediate

Evaluating Fairness Metrics

You have an AI model that predicts loan approvals. The model approves 90% of applications from one group but only 50% from another. Which metric best helps identify this fairness issue?

A Accuracy

B Demographic parity

C Mean squared error

D Precision

💡 Hint

Look for a metric that compares outcomes across different groups.

Demographic parity measures if different groups receive positive outcomes at similar rates, highlighting fairness issues.

❓ Model Choice

advanced

Choosing Models for Privacy Preservation

You want to build an AI system that respects user privacy by not storing personal data. Which approach is best?

A Use federated learning to train models locally on user devices

B Collect all data and anonymize it after training

C Train a centralized model with all user data stored on one server

D Use a simple linear regression model to reduce data needs

💡 Hint

Consider methods that keep data on user devices instead of sending it to a central place.

Federated learning trains models on user devices without sending personal data to a central server, enhancing privacy.

🔧 Debug

advanced

Identifying Ethical Risks in AI Deployment Code

Consider this pseudocode for deploying an AI chatbot:

if user_input contains sensitive_topic:
    respond with generic answer
else:
    respond with AI-generated answer

What ethical risk does this code most likely introduce?

A It stores all user inputs without consent

B It will always give detailed answers, risking privacy

C It uses too much computing power for simple responses

D It may ignore important user needs by avoiding sensitive topics

💡 Hint

Think about how avoiding sensitive topics might affect users seeking help.

By giving generic answers to sensitive topics, the chatbot may fail to support users who need specific help, causing harm.

🧠 Conceptual

expert

Balancing Transparency and Security in AI Systems

Which statement best explains the challenge of transparency in AI while maintaining security?

A Security requires hiding all AI decision processes from users

B Transparency means AI systems must never be updated after deployment

C Making AI models fully transparent can expose sensitive data or system vulnerabilities

D Transparency and security are unrelated and can be handled separately

💡 Hint

Consider what happens if you reveal everything about how an AI works.

Sharing full details of AI models can risk exposing private data or ways to attack the system, so transparency must be balanced with security.