Why responsible AI development matters in Prompt Engineering / GenAI - Challenge Your Understanding

Challenge - 5 Problems

🧠 Conceptual

intermediate

Why is fairness important in AI systems?

Imagine an AI system used to decide who gets a loan. Why must this system be fair?

A Because unfair AI can deny loans to certain groups without good reason, causing harm.

B Because fairness makes the AI run faster and use less memory.

C Because fairness means the AI always gives loans to everyone regardless of risk.

D Because fairness allows the AI to ignore user data and make random decisions.

💡 Hint

Think about the impact of biased decisions on people's lives.

Fairness ensures AI treats people equally and avoids harm from biased decisions.

❓ Metrics

intermediate

Which metric helps detect bias in classification models?

You have a model that predicts if a person qualifies for a job. Which metric helps check if the model is biased against a group?

A Accuracy score over the entire dataset.

B Number of layers in the model.

C False positive rate difference between groups.

D Training loss value.

💡 Hint

Bias often shows as different error rates for different groups.

False positive rate difference reveals if one group is wrongly flagged more often, indicating bias.

🔧 Debug

advanced

What error occurs with this AI fairness check code?

Consider this Python code snippet that tries to compute demographic parity difference but fails:

group_0 = predictions[labels == 0]
group_1 = predictions[labels == 1]
parity_diff = abs(group_0.mean() - group_1.mean())
print(parity_diff)

What error will this code raise if labels is a list, not a numpy array?

Prompt Engineering / GenAI

import numpy as np
predictions = np.array([0,1,1,0,1])
labels = np.array([0,1,0,1,0])
group_0 = predictions[labels == 0]
group_1 = predictions[labels == 1]
parity_diff = abs(group_0.mean() - group_1.mean())
print(parity_diff)

A IndexError: list index out of range

B AttributeError: 'list' object has no attribute 'mean'

C No error, prints the parity difference

D TypeError: list indices must be integers or slices, not list

💡 Hint

Think about what happens when you use a boolean mask on a Python list.

Boolean indexing works on numpy arrays but not on Python lists, causing a TypeError.

❓ Model Choice

advanced

Which model type is best to improve AI transparency?

You want an AI model that is easy to understand and explain to users. Which model type is best?

A Deep neural network with many hidden layers

B Decision tree with clear branching rules

C Ensemble of random forests

D Support vector machine with RBF kernel

💡 Hint

Think about which model shows decisions in simple steps.

Decision trees show clear rules, making them easier to explain than complex models.

🧠 Conceptual

expert

Why is continuous monitoring important in responsible AI?

After deploying an AI system, why must we keep monitoring its behavior over time?

A Because AI models can degrade or become biased as data changes over time.

B Because monitoring increases the AI model's training speed.

C Because monitoring allows the AI to automatically fix bugs without human help.

D Because once deployed, AI models never change and need no checks.

💡 Hint

Think about how real-world data can shift and affect AI decisions.

Data and environments change, so AI models can become less accurate or unfair without monitoring.