Step 4 · Writing (Criteria A and B)
Structure carries your argument; knowledge shows you understand the computer science behind it.
By the end of this step
- A first rough draft: structure in place, arguments may be incomplete. Useful (not required) for your interim reflection session.
The structure of a CS essay
| Section | What goes in it | Words (plan) |
|---|---|---|
| Title page | Title, research question, subject, word count (see the guide for the full list) | – |
| Contents | Headings with page numbers | – |
| 1 Introduction | The RQ, why it matters (with sources), what you expect and why | 350 |
| 2 Background / theory | Only the CS concepts the investigation needs, applied to your question | 800 |
| 3 Methodology | Variables, set-up, controls, repeats, justification of choices | 600 |
| 4 Results | Tables and charts (not counted) with a short commentary | 300 |
| 5 Analysis and discussion | What the results show, why, how they compare with the literature, evaluation throughout | 1,450 |
| 6 Conclusion | Answer the RQ directly; limitations; further research | 400 |
| References | Every source you cite, in one consistent style | – |
| Appendices | Code, raw data. Nothing essential | – |
Introduction · Background · Methodology · Results · Analysis and discussion · Conclusion.
Sections in order Order drill
Without looking up: put the parts of a CS essay in order.
Plan your 4,000 words
The word budget Simulator
Drag the sliders. Watch what happens when you go over, and where Bit thinks the words should go.
Conventions examiners expect in CS
- Number and caption every table and figure, and refer to it in the text (“see Figure 3”).
- Units on every axis and column (µs, ms, MB); SI units.
- Explain any maths in computing terms: what the expression represents and why it matters here.
- Formal, precise language. Use “I” only when describing what you did.
- One referencing style (APA or IEEE), used consistently; cite datasets and libraries too.
Criterion B: show you understand, don't just define
“Big O notation describes how the running time of an algorithm grows. O(1) means constant time. O(n) means linear time.” (a textbook definition, never used again)
“With separate chaining, a lookup checks on average 1 + α/2 entries, where α is the load factor, so the cost should rise linearly with α. For linear probing the expected number of probes is about ½(1 + 1/(1−α)) (Knuth, 1998), so it should stay fast at low α but degrade sharply above about 0.7. These predictions are tested in section 5.”
Knowledge that is used to make a prediction, design the method or explain a result earns marks.
- Knowledge: relevant, comprehensive sources, mainly scholarly.
- Terminology: accurate and consistent (don't switch between “probe”, “check” and “step”).
- Concepts: explained and applied: complexity, data structures, caching, entropy, overfitting…
Finding good sources
- Google Scholar, IEEE Xplore, ACM Digital Library, arXiv (preprints: check whether they were later published).
- Textbooks (e.g. Introduction to Algorithms) for core theory.
- Official documentation for libraries and languages.
- AI tools may help you find sources, but check each one exists and says what you claim: AI can invent references.
How much can you rely on it? Self-check drill
Rate each source for a CS EE.