EPQ Computer Science project ideas and research plans
Choose a focused AI, data science, design or cybersecurity question and plan a practical, ethical investigation.
Free research planning · UK students
Choose a project question you can actually investigate
A good EPQ question has a manageable scope, clear evidence, a feasible method and enough room for evaluation. Avoid choosing a fashionable technology and only afterwards trying to find an argument. Make sure your project complies with your school's qualification rules, safeguarding and research ethics.
AI model reliability
How do different training-set sizes affect the accuracy of a small image classification model on an unseen test set?
Data bias
How much does an unbalanced fictional dataset change the conclusions from a simple model?
Accessibility
Which text-size and contrast choices make a sample revision interface easiest for volunteers to use?
Cyber hygiene
How do different password-education messages influence choices in a consent-based, fictional quiz?
Algorithm efficiency
How does input size affect the time taken by two sorting algorithms under controlled tests?
Digital sustainability
How do different image formats change the file size and perceived quality of a school website's illustrations?
Human-centred design
What usability problems emerge when students try two versions of a fictional homework dashboard?
Computer vision
How does lighting affect a small image recognition system using non-personal objects?
Data visualisation
Which of three chart types helps readers answer a simple statistical question most accurately?
Worked planning example: compare two simple classifiers
| Planning question | Student's draft |
|---|---|
| What is the question? | How does training-data quality affect the reliability of a simple classifier? |
| What will I vary? | Proportion of deliberately misplaced labels in a fictional training dataset. |
| What will stay the same? | Model type, test set, data features and evaluation measure. |
| What evidence will I collect? | Accuracy, an error table, examples of mistakes and a reflective log. |
| How will I evaluate? | Compare a baseline, describe limitations and avoid claiming wider generalisability from a tiny sample. |
A simple eight-week planning framework
- Week 1: define a question and agree the scope with your supervisor.
- Week 2: locate trustworthy sources and record why you chose them.
- Week 3: design a safe, reproducible research method and pilot it.
- Weeks 4–5: collect results, noting failures and changes.
- Week 6: analyse results, check limitations and compare explanations.
- Week 7: write conclusions and organise citations, figures and appendices.
- Week 8: review and present the project as required by your centre.
This is an illustrative planning sequence, not an awarding-body timetable. Projects can require different lengths and assessment arrangements.
What counts as convincing evidence?
Record the initial hypothesis, method, observations, test conditions and any adjustments you made. A small study with well-explained limitations is usually more defensible than a large claim without reliable evidence. Keep identifiable personal information out of datasets unless your centre has approved an appropriate, lawful and ethical approach.
Self-check: is the question too broad?
“Will AI change the world?” is not a manageable investigation by itself. “How does a chosen data-quality problem change an image classifier's results on a fixed student-created dataset?” provides a narrower testable scope.