AI Use in Programming and Ethics

This is optional enrichment. AI code assistants and chatbots are very relevant to modern programming, but detailed AI-tool use is outside H2 exam-core scope unless a question explicitly gives an AI-use scenario.

The main message is simple:

AI is a tool for support.
It is not a substitute for understanding, testing, or academic honesty.

Beginner Mental Model

An AI assistant can generate explanations, code, test ideas, and debugging suggestions. It does not automatically know whether the answer is correct for your exact task, syllabus, file structure, or marking scheme.

Caption: Safe AI use means inspecting, testing, explaining, and acknowledging assistance instead of copying blindly.

What AI Can Help With

AI can be useful when used carefully.

Student need	Helpful AI use
understand an error message	ask for a plain-language explanation
plan a function	ask for pseudocode or edge cases
design tests	ask for normal, abnormal, and extreme cases
compare approaches	ask for trade-offs between simple solutions
improve explanation	ask for a clearer analogy or step-by-step trace
review code	ask what assumptions or failure cases to check

Good use:

Ask AI to explain why my loop misses the last item.
Then inspect my code, fix it, and test boundary cases myself.

Weak use:

Paste the task, copy the whole answer, and submit it without understanding.

Hallucinated Code

AI can produce code that looks confident but is wrong. This is often called hallucination.

Examples:

• imports a library that is not installed;
• calls a function that does not exist;
• assumes a file has different columns or keys;
• writes SQL for a table name that is not in the database;
• solves a similar problem, not the exact problem;
• passes one sample case but fails boundary cases.

This is why generated code must be treated as untrusted draft code.

Safe rule:

Generated code is not correct until it has been read, understood, and tested against the actual task.

Testing Generated Code

Use the same testing discipline as for human-written code.

Check	Question
syntax	does the code run at all?
task fit	does it solve the exact requirement?
data fit	does it match the actual file/database schema?
normal test	does it work for ordinary input?
abnormal test	does it reject or handle invalid input?
extreme test	does it work at boundaries?
explanation	can the student explain why it works?

For example, if AI writes a validation function, do not only test one valid input. Test missing input, wrong format, boundary values, and unexpected types where relevant.

Plagiarism and Academic Honesty

AI assistance can create plagiarism risk when a student submits generated work as if it were entirely their own.

Different schools may have different rules, so follow the exact policy given by the teacher or exam authority. When in doubt, ask before using AI for assessed work.

Responsible behaviour:

• use AI to clarify concepts, not to hide lack of understanding;
• acknowledge AI assistance when required;
• do not submit generated code that you cannot explain;
• do not ask AI to produce answers to restricted assessments;
• do not upload private data, school credentials, or confidential files into public tools;
• keep drafts and final submitted work clearly separated.

Important distinction:

Learning from an explanation can be legitimate.
Submitting ununderstood generated work can be dishonest.

Explainability

Explainability means being able to explain how an answer, decision, or program works.

For student programming, explainability has a practical meaning:

Can I trace the code?
Can I explain each variable?
Can I predict output for a test case?
Can I justify the algorithm choice?
Can I modify the code if the requirement changes?

If the answer is no, the AI-generated code is not yet your working knowledge.

Weak defence:

The AI said it works.

Stronger defence:

I tested the function with these normal, abnormal, and boundary inputs.
The trace shows why the loop terminates and returns the correct value.

Bias and Fairness

AI systems can reflect biases in their training data, design, or use context.

In programming help, bias may appear as:

• assuming a particular language, country, school system, or naming convention;
• giving examples that exclude some users;
• suggesting solutions that work for one group but not another;
• producing unfair or stereotyped descriptions;
• treating majority cases as if they cover all users.

In AI decision systems, bias can be higher risk because real people may be affected by decisions about access, ranking, recommendation, or classification.

Ethical questions to ask:

Who benefits from this AI system?
Who may be harmed?
What data was used?
Can affected users understand or challenge the result?
Is human review needed?

Privacy and Data Leakage

Students should be careful about what they paste into AI tools.

Avoid sharing:

• passwords, API keys, tokens, or .env files;
• private student data;
• unpublished exam questions if rules forbid sharing;
• database files containing personal data;
• proprietary school or project material;
• confidential deployment settings.

Safe alternative:

Replace real names, keys, and private data with small artificial examples before asking for help.

Worked Scenario

Scenario:

A student asks an AI assistant to write a Flask route that saves attendance into SQLite.
The generated code runs locally, but it inserts into a table called Attendance.
The supplied database actually contains a table called CCA_Attendance with different column names.

Reasoning:

Issue	Why it matters
hallucinated table name	code does not match actual database
no schema check	student did not inspect the supplied resource
local run may still fail later	route may crash only when form is submitted
student cannot explain SQL	debugging becomes difficult

Responsible next step:

Inspect the real database schema.
Adjust the SQL to match actual table and column names.
Run a small insertion test.
Query the database to confirm the stored row.
Explain the route before using it in the project.

Safe AI Workflow for Student Code

Use this workflow:

1. Read the task yourself.
2. Ask AI for explanation, planning, or a draft only where allowed.
3. Compare the draft with the actual files, schema, and syllabus.
4. Run the code.
5. Test normal, abnormal, and extreme cases.
6. Explain the solution in your own words.
7. Acknowledge AI help if required.

This makes AI part of the learning process instead of a replacement for it.

How to Use This in Exam Answers

For H2 exam answers, use syllabus terms first:

• ethics and professional conduct;
• privacy and data protection;
• social and economic impact;
• validation, testing, and debugging;
• error types and test cases;
• correctness and reliability.

Mention AI code assistants, hallucination, prompt use, or generated code only if:

• the question gives an AI-use scenario;
• you are writing optional enrichment;
• you are discussing responsible computing practice beyond the core syllabus.

Avoid:

• saying AI output is automatically correct;
• saying AI use is automatically dishonest in every context;
• submitting generated work you cannot explain;
• using AI to bypass learning, testing, or citation rules;
• uploading private data or secrets into an AI tool.

Connect Back to Topics

Topic	Connection
Social, Ethical, Legal, and Economic Issues	academic honesty, bias, privacy, accountability, responsible use
Data Validation, Testing, and Debugging	generated code still needs syntax, runtime, logic, and boundary testing
Lab Exam and Project Skills	inspect resources, test incrementally, keep final work explainable

Final Takeaway

AI can speed up learning and drafting, but it does not remove responsibility.

Use this reasoning pattern:

What did AI help produce?
Do I understand it?
Have I tested it with the actual task data?
Is it allowed for this assessment?
Have I protected private data?
Can I explain and defend the final work?