Artificial intelligence for A-Level Computer Science
Artificial intelligence (AI) is the branch of computer science that builds systems capable of performing tasks normally requiring human intelligence. For OCR A-Level Computer Science (specification H446), AI is named in only one place: Section 1.5.2, Moral and ethical issues. The spec treats AI as an ethics topic, not as a technical one, machine learning and neural networks are not named in the spec at all.
This guide covers what OCR actually asks you to do (discuss the ethical and legal issues AI raises), gives the broader technical background as context, and shows how to structure the extended-response ethics question on Paper 1 (Component 01).
AI sits in topic 1.5.2
OCR examines AI as a moral and ethical issue, not as a technical topic. Bias, accountability, privacy, and employment are the headline issues.
ML and neural networks are context
Machine learning and neural networks are not named in the H446 spec. Useful background for stronger answers, but not directly examined.
Search algorithms are separate
Tree traversals (BFS, DFS) sit in 2.3.1(e). Dijkstra, A*, binary search, and linear search sit in 2.3.1(f). These are not AI-specific.
Defining artificial intelligence
Artificial intelligence is the simulation of human intelligence by computer systems, including learning, reasoning, and problem solving. OCR mark schemes accept any answer that includes the idea of a machine performing a task that would normally require human intelligence.
OCR distinguishes between narrow AI (systems built for one specific task, such as image recognition or chess) and general AI (a hypothetical system that can perform any intellectual task a human can). All current AI is narrow. General AI does not yet exist.
Two definitions to memorise Narrow AI: A system designed to perform one specific task, such as voice recognition or playing chess. General AI: A hypothetical system capable of performing any intellectual task a human can. All real-world AI today is narrow.
Machine learning and neural networks: useful background
Not on the OCR H446 specification Machine learning and neural networks are not named in the H446 specification. The technical detail below is background context for stronger ethics answers and for wider reading. Do not rely on it appearing in a knowledge question, OCR examines AI through the ethics lens in section 1.5.2.
Machine learning is a sub-field of AI where systems learn from data instead of being explicitly programmed. The system is given a large dataset, builds a model of the patterns it sees, and uses that model to make predictions on new data. Three categories show up in most introductions to ML.
| Type | What it does | Example use case |
|---|---|---|
| Supervised learning | Learns from labelled training data | Spam email classifier trained on emails labelled spam or not spam |
| Unsupervised learning | Finds patterns in unlabelled data | Customer segmentation grouping shoppers by behaviour |
| Reinforcement learning | Learns by trial and error using rewards and penalties | Game-playing agents like AlphaGo |
A neural network is one common machine learning model, loosely based on the structure of the human brain. It is built from layers of artificial neurons connected by weighted edges. An input layer takes in raw data, one or more hidden layers transform the data via weights and activation functions, and an output layer produces the final prediction. Deep learning is the term used when there is more than one hidden layer.
Again, this is wider-reading context. OCR will not ask you to define a hidden layer or describe backpropagation in H446.
Search algorithms on the OCR spec
Search algorithms are sometimes loosely associated with AI, but on the OCR H446 spec they sit in section 2.3.1, separately from the ethics topic. Be careful about two different parts of the spec.
Section 2.3.1(e) covers tree traversal: depth-first (post-order) and breadth-first. Section 2.3.1(f) covers standard search algorithms: Dijkstra's shortest path, A*, binary search, and linear search. BFS and DFS are tree traversals on the spec, not 'standard searches'.
| Spec reference | Algorithms | What it covers |
|---|---|---|
| 2.3.1(e), Tree traversal | Depth-first (post-order), breadth-first | Visiting every node of a tree in a defined order |
| 2.3.1(f), Standard search algorithms | Dijkstra's, A*, binary search, linear search | Finding a target or shortest path, with A* using a heuristic |
Ethical and legal issues
OCR Paper 1 often includes a nine or twelve-mark extended-response question on the ethical issues raised by AI. A useful framing is the BCS Code of Conduct, which sets out four principles for computing professionals: Public Interest; Professional Competence and Integrity; Duty to Relevant Authority; and Duty to the Profession.
Four issues come up most often in mark schemes: Bias in training data, accountability when AI makes a wrong decision, privacy of personal data used for training, and the impact on employment as AI takes over tasks previously done by humans.
| Issue | Why it matters | Real-world example |
|---|---|---|
| Bias | Models trained on biased data produce biased outputs | Facial recognition systems with higher error rates on darker skin tones |
| Accountability | Hard to assign blame when an AI makes a mistake | Self-driving car accidents: Driver, manufacturer, or programmer? |
| Privacy | Large training datasets often include personal information | Voice assistants storing recordings of private conversations |
| Employment | AI automates tasks previously done by people | Customer service chatbots replacing call-centre staff |
Structuring a twelve-mark ethics answer
A twelve-mark OCR extended-response question rewards a balanced answer with named examples and a clear conclusion. The marking grid rewards three things: Knowledge and understanding, application to context, and evaluation.
Use a four-paragraph structure. Paragraph one defines the key terms. Paragraph two gives two arguments in favour of using the AI system. Paragraph three gives two arguments against. Paragraph four reaches a clear judgement and justifies it with evidence.
What examiners reward in ethics answers Clear definitions of the AI system in question. Specific real-world examples (not "some companies"). Reference to a recognised framework like the BCS Code of Conduct. A reasoned judgement at the end, not a fence-sitting summary. Two or three issues covered in depth beats five mentioned briefly.
Worked example: A twelve-mark ethics question
Question: Discuss the ethical issues raised by the use of AI in recruitment decisions. (12 marks)
Model answer opening: AI recruitment tools, such as Amazon's experimental hiring algorithm, use machine learning to score CVs and shortlist candidates. While they offer faster shortlisting and can reduce human bias on factors like name or address, they raise serious concerns. Bias in training data is the biggest issue: Amazon's tool learned from ten years of CVs from a male-dominated industry and began down-ranking applications containing the word "women's" (for example, "women's chess club captain"). Privacy is also a concern, as applicants may not know their CV is being processed by an algorithm.
On the other hand, AI can screen large applicant pools faster than any human team and apply consistent criteria across every candidate. Overall, AI in recruitment is acceptable only when the training data is carefully audited, the decision-making is transparent, and a human reviews every shortlisting decision. Without those safeguards, the bias risk outweighs the efficiency gain.
Key facts to memorise for the exam
- AI is named in OCR H446 only in section 1.5.2, Moral and ethical issues
- Machine learning and neural networks are not named on the spec, useful background only
- Narrow AI performs one task; general AI is hypothetical and does not exist yet
- Tree traversal (2.3.1(e)) covers breadth-first and depth-first traversal
- Standard search algorithms (2.3.1(f)) are Dijkstra's, A*, binary search, and linear search
- A* uses a heuristic to estimate the cheapest path to the goal
- Ethics: Bias, accountability, privacy, and employment are the four main issues
- Use one real-world example per ethical issue for top-band marks
