The best IB Computer Science EE topics are focused technical problems that allow you to collect reproducible evidence, apply computer science concepts, and reach a qualified conclusion within 4,000 words. A fashionable or complicated topic is not automatically a strong one.
Promising areas include lightweight machine learning, post-quantum cryptography, pathfinding, compression, cybersecurity, database indexing, recommender systems, and concurrency. The examples below are starting points. Your final question should reflect your interests, available resources, preliminary research, and supervisor discussions.
What makes a strong Computer Science EE topic?
The IB describes the Extended Essay as an independent research project culminating in a 4,000-word paper. A Computer Science EE should therefore investigate a question, not merely describe technology or document an application you built.
A strong topic normally includes:
- A technical focus, such as algorithmic efficiency, model performance, security, memory use, or computational complexity.
- Defined variables, technologies, datasets, and testing conditions.
- Accessible evidence from experiments, simulations, benchmarks, or critical analysis.
- Suitable metrics, such as execution time, accuracy, F1 score, compression ratio, latency, or throughput.
- Scope for evaluation, including trade-offs, anomalies, uncertainty, and limitations.
Programming can generate evidence, but code alone is not research. The essay must explain why results occurred and what they mean. Check the official IB Extended Essay overview and current Extended Essay updates alongside your school's guidance, particularly for the May 2027 assessment cycle or later.
IB Computer Science EE topics to investigate
| Research area | Possible research question | Method and metrics |
|---|---|---|
| Lightweight machine learning | To what extent does 8-bit quantization affect a selected image classifier on a defined device? | Compare accuracy, latency, model size, and memory use. |
| Pathfinding | How does obstacle density affect A* and Dijkstra's algorithm on fixed-size weighted grids? | Measure path cost, nodes explored, and execution time. |
| Post-quantum cryptography | How do ML-DSA and SLH-DSA differ on a specified computer? | Compare signing time, verification time, and signature size. |
| Data compression | When does Brotli provide a better time-compression trade-off than gzip for web files? | Test a fixed corpus using compression ratio and processing time. |
| Phishing classification | How does class-imbalance treatment affect two classifiers on a selected dataset? | Compare precision, recall, F1 score, and confusion matrices. |
| Database indexing | How does dataset size affect B-tree and hash-index performance? | Test equality and range queries, latency, and storage overhead. |
| Concurrency | How do asynchronous and multithreaded approaches compare for input-output-bound Python tasks? | Measure throughput, latency, CPU use, and memory. |
| Recommender systems | How does matrix sparsity affect user-based and item-based collaborative filtering? | Compare RMSE, precision at , or recall at . |
Choosing a current but manageable area
Lightweight machine learning
Model efficiency creates a useful trade-off between predictive performance and computational cost. Investigate one intervention, such as quantization, pruning, or input resolution, rather than combining several changes.
A small pre-trained model and a documented dataset such as CIFAR-10 are more manageable than training a generative model. Record software versions, random seeds, settings, hardware, and repeated results so your method can be reproduced.
Post-quantum cryptography
NIST finalized its first three principal post-quantum cryptography standards in 2024: ML-KEM, ML-DSA, and SLH-DSA. This provides authoritative specifications and maintained implementations for investigation.
Avoid speculative questions such as “Will quantum computers break encryption?” Instead, compare measurable implementation costs at a fixed security level on one platform. The NIST Post-Quantum Cryptography Project is an appropriate starting source.
Cybersecurity
Cybersecurity works well when testing is ethical, contained, and technically narrow. Suitable areas include phishing classification, password-hashing performance, intrusion detection, or a defined input-validation method.
OWASP identifies prompt injection as a major risk, but commercial models may change during an investigation. If studying this area, use a fixed open-source model in an isolated environment, define a limited attack taxonomy, and obtain supervisor approval.
Turning an idea into a research question
Use the structure comparison or intervention + defined context + measurable outcomes. “AI efficiency” is only an area, while “To what extent does 8-bit quantization affect accuracy and median inference time for Model X on Dataset Y using Device Z?” identifies boundaries and variables.
Before committing:
- Find 5-8 credible sources, including papers, specifications, and dataset documentation.
- Confirm that the required software, hardware, and data are accessible.
- Run a pilot that produces usable evidence.
- Check whether trials can be repeated consistently.
- Identify at least two likely limitations.
- Confirm with your supervisor that the question remains grounded in Computer Science.
RevisionDojo's guide to writing an Extended Essay research question can help you refine the wording. You can also study Computer Science EE examples to see how methods, evidence, and conclusions connect. Use exemplars to understand standards, never as templates to copy.
Designing a defensible experiment
Control factors that could distort the comparison. For algorithm benchmarks, keep the computer, programming language, software version, input-generation method, and measurement procedure consistent.
Timing data are noisy, so conduct repeated trials and report a suitable summary, often the median, plus a measure of spread. Consider background processes, caching, warm-up effects, and measurement precision before treating small differences as meaningful.
For machine-learning investigations, accuracy may be misleading when classes are imbalanced. Precision, recall, F1 score, confusion matrices, and stratified validation can reveal weaknesses that accuracy hides. The UCI Machine Learning Repository offers documented datasets, but you must still examine licensing, missing values, class balance, and original purpose.
Common topic mistakes
- Choosing an overly broad question: “Which programming language is fastest?” lacks a defined task, environment, implementation, and metric.
- Producing an IA-style product: An app, game, or website should function as an experimental instrument, not become the investigation's main purpose.
- Using unstable technology: Prefer versioned open-source software, archived datasets, and local experiments over changing commercial platforms.
- Reporting without explaining: Connect patterns to computational principles, investigate anomalies, and state where conclusions do not apply.
- Reusing IA material: Keep questions, datasets, code, analysis, and writing distinct. Review RevisionDojo's guidance on EE and IA material reuse, then confirm your plan with your coordinator.
A practical fall scoping plan
During the first two weeks, shortlist three areas and read enough to explain their central concepts. In weeks 3-4, create candidate questions and run a pilot for the most feasible option. By weeks 5-6, aim to have a provisional question, source log, reproducible method, and sample results.
Record why you rejected alternatives, changed metrics, or narrowed the scope. These decisions can support genuine reflection because they show how your methodology developed.
Use the RevisionDojo EE Guides to check question design and structure. Jojo AI can challenge assumptions and help identify uncontrolled variables, while the IB Coursework Grader can provide diagnostic feedback on work written in your own words. Automated feedback is not an official IB grade and does not replace supervisor guidance.
Conclusion
Strong IB Computer Science EE topics combine current relevance with a narrow, controllable investigation. Choose a question for which you can access credible sources, run a pilot, collect reproducible data, and evaluate meaningful trade-offs.
RevisionDojo can support this process through its EE Guides, Computer Science exemplars, Jojo AI, and Coursework Grader. Begin with question refinement and a small feasibility test before committing to a full investigation.
Sources and referenced URLs
- International Baccalaureate: Extended Essay overview
- International Baccalaureate: Extended Essay updates
- NIST Post-Quantum Cryptography Project
- OWASP prompt injection guidance
- UCI Machine Learning Repository
- UCI CIFAR-10 dataset
- RevisionDojo Extended Essay research-question guide
- RevisionDojo Computer Science EE examples
- RevisionDojo guide to EE and IA material reuse
- RevisionDojo Extended Essay Guides
- RevisionDojo IB Coursework Grader