Bias and over-reliance when using AI to program

Name:

Date:

Class:

Quick Questions

Shade in the A or B circle for each question.

1. A student pastes AI-generated code into their project and submits it without reading it. Is this good practice?

A No. The code could be biased, outdated or insecure, so it must be reviewed first.

B Yes. AI code is always correct, so checking it is a waste of time.

2. An AI writes a loan approval function that gives lower scores to people from certain postcodes. What is the problem?

A This is fine because a computer cannot be unfair.

B The code shows bias and could treat groups of people unfairly.

3. Which is a sign that AI-generated code may be insecure?

A It uses clear variable names and comments.

B It builds an SQL query by joining user input straight into the string.

4. An AI suggests a Python method that was removed in newer versions. What is the risk?

A The code may be outdated and fail or be unsafe in the current version.

B There is no risk because AI always uses the latest version.

5. A student lets AI write every program and never practises writing code. What is the likely result?

A De-skilling: their own programming skills and problem-solving get weaker.

B Their skills improve faster because they write less code.

Match the Pairs

Draw a line connecting each item on the left to its match on the right.

Left column		Right column
1query = "SELECT * FROM users WHERE name = '" + name + "'"		ABiased or unfair output: the rule rejects people by where they live, which could unfairly disadvantage a group
2if applicant_postcode in blocked_postcodes: reject()		BLetting AI do all the coding causes de-skilling. You lose your own problem-solving skills and judgement. Use AI as a helper and write and understand code yourself.
3Outdated code: AI may suggest old methods that are no longer recommended, so check the documentation		Cassert calculate_total([2, 3]) == 5 # I tested it
4Keeping a human in control: read the code, test it and be able to explain every line before you use it		Dimport md5 # old, weak hashing library
5# AI wrote it, so I will paste it in without reading		EInsecure AI-generated code: joining user input into an SQL string allows SQL injection, so it must be reviewed and fixed
6# Let AI write all my functions so I never need to learn how		FPasting AI code without reviewing it means bias, outdated methods or security flaws go unnoticed. Review and test it first.

Definitions

AI over-reliance

Fill in the Blanks

Complete each line of code by filling in the missing parts.

# AI hardcoded it; ask the user instead password = _____("Enter password: ")

if reviewed ___ tests_passed: deploy()

# Equal skills, so scores must match assert score(a) __ score(b)

# I wrote this myself, no AI ___ average(marks): return sum(marks) / len(marks)

answer = input("Accept AI code? y/n: ") __ answer == "y": print("Merged after review")

Spot the Error

Shade in the letter of the line that contains an error.

Question 1

A def pick(skills, years, age):

B score = skills * 2

C score -= age // 10

D return score >= 10

Question 2

A import md5

B pw = input('Password: ')

C data = pw.encode()

D h = hashlib.sha256(data)

Question 3

A code = ask_ai('sort list')

B review(code)

C if code:

D deploy(code)

Question 4

A attempts = 3

B while attempts < 3:

C try_myself(task)

D attempts += 1

E ask_ai_hint(task)

Question 5

A def rank(applicant):

B return applicant.age < 30

C best = max(pool, key=rank)

Answer Key

Bias and over-reliance when using AI to program

Section A: Quick Questions

A — No. The code could be biased, outdated or insecure, so it must be reviewed first.

B — The code shows bias and could treat groups of people unfairly.

B — It builds an SQL query by joining user input straight into the string.

A — The code may be outdated and fail or be unsafe in the current version.

A — De-skilling: their own programming skills and problem-solving get weaker.

Section B: Match the Pairs

query = "SELECT * FROM users WHERE name = '" + name + "'" → Insecure AI-generated code: joining user input into an SQL string allows SQL injection, so it must be reviewed and fixed

if applicant_postcode in blocked_postcodes: reject() → Biased or unfair output: the rule rejects people by where they live, which could unfairly disadvantage a group

Outdated code: AI may suggest old methods that are no longer recommended, so check the documentation → import md5 # old, weak hashing library

Keeping a human in control: read the code, test it and be able to explain every line before you use it → assert calculate_total([2, 3]) == 5 # I tested it

# AI wrote it, so I will paste it in without reading → Pasting AI code without reviewing it means bias, outdated methods or security flaws go unnoticed. Review and test it first.

# Let AI write all my functions so I never need to learn how → Letting AI do all the coding causes de-skilling. You lose your own problem-solving skills and judgement. Use AI as a helper and write and understand code yourself.

Section C: Definitions

AI over-reliance — Trusting AI-generated code without checking it. The code may be biased, outdated or insecure, and you lose your own skills and judgement. Always review it and stay in control.

Section D: Fill in the Blanks

# AI hardcoded it; ask the user instead
password = input("Enter password: ")

if reviewed and tests_passed:
    deploy()

# Equal skills, so scores must match
assert score(a) == score(b)

# I wrote this myself, no AI
def average(marks):
    return sum(marks) / len(marks)

answer = input("Accept AI code? y/n: ")
if answer == "y":
    print("Merged after review")

Section E: Spot the Error

C — This AI-style scoring code penalises older applicants, which is biased and unfair. A fair score should only use relevant factors, so the line should be 'score += years'.

A — The old md5 module is outdated and insecure, and AI tools can still suggest it. It should be 'import hashlib' so the later hashlib.sha256() call works.

C — Any non-empty AI code would pass this check, so the human is no longer in control. The condition should be 'if tests_pass(code):' so the code only deploys after it passes tests.

A — Starting attempts at 3 means the loop never runs, so the student never tries the task and goes straight to the AI. This is de-skilling. It should be 'attempts = 0'.

B — Ranking by age builds unfair bias into the program. The rank should be based on relevant data, so the line should be 'return applicant.skills'.