Computer Science · Big Data · Topic 5 of 5

Every Big Data question in this unit ends the same way: who benefits, and who is put at risk?

The 2026 specification names this directly: "big data considerations (privacy, ethics, and data governance)." This is where the highest-level answers (Merit/Excellence: "discuss", "critically analyse") are actually won or lost — you need more than "it's bad for privacy".

AS91908 · Big Data·Final topic in this unit

Photo: Frankie Cordoba / Unsplash

Executive summary

Goal: distinguish privacy, ethics, and governance precisely, name a real privacy-engineering technique, and weigh a named use case's benefits against its risks with a justified conclusion.

Section	What it covers
Three distinct concepts	Privacy, ethics, and governance — related but not the same thing
Privacy in practice	Google's real anonymisation and consent techniques
Ethical concerns	Profiling, discrimination, manipulation, informed consent at scale
Data governance	Ownership, access, retention, accountability
Weighing benefits against harms	Six real use cases and the highest-value question in the paper

Three distinct concepts

Students often blur these together. They're related but not the same thing, and naming the right one precisely is worth marks.

Concept	What it actually means
Privacy	An individual's right to control what personal information about them is collected and who can see it
Ethics	Broader questions of right and wrong in how data is used — even when it's technically legal and the person consented
Governance	The rules, policies and accountability structures an organisation puts in place to manage data responsibly — who owns it, who can access it, how long it's kept

Privacy in practice: anonymisation and consent

Case file: Google's privacy engineering

Google's 2020 Community Mobility Report is a genuinely useful worked example of real privacy engineering, not just a policy statement. Its published "Preserving privacy" section describes several concrete techniques together:

• Opt-in consent — the underlying Location History setting is off by default; only users who actively turn it on contribute data, and they can turn it off or delete their history at any time
• Aggregation — reports are built from combined data across many users, not individual movements
• Differential privacy — the reports describe adding artificial statistical noise to the dataset, which allows overall trends to remain accurate while making it much harder to identify any single individual's actual behaviour from the published numbers
• Suppressing absolute numbers — the report deliberately only publishes percentage change, not raw visit counts, which further limits what could be inferred about specific people or places

(Source: 91908-res-2021.pdf, Resource G.)

This is worth remembering as a concrete answer to "how can privacy risks be addressed" — rather than a vague "they should anonymise the data", you can name the actual mechanism (aggregation + differential privacy + suppressing raw counts) and explain what each part protects against.

Ethical concerns of predictions and recommendations

Recommendation and prediction systems (see Topic 4) raise ethical questions that go beyond privacy:

• Profiling — building a detailed behavioural profile of a person from patterns in their data, even without knowing their name
• Discrimination — a model trained on biased historical data can produce systematically unfair predictions for particular groups
• Manipulation / engagement optimisation — a system designed to maximise time-on-platform may prioritise content that keeps someone scrolling over content that's actually good for them
• Informed consent at scale — most users don't realistically read or understand what they're agreeing to when they accept a platform's terms of service

Past-paper practice

What are the potential ethical concerns that a platform might need to consider when using big data to make predictions or recommendations? (91908-cat-A-2023.pdf / 91908-cat-B-2023.pdf, Question Two (c).)

Reveal how to answer well

Pick two or three of the concerns above and explain the mechanism that causes each one — e.g. "a recommendation model trained on what users already engage with can create a feedback loop that narrows the range of content someone sees over time, which raises a concern about manipulation rather than informed choice."

Data governance

Governance is the organisational layer that's supposed to prevent privacy and ethical harms in the first place. Key governance questions include:

• Ownership — who owns the data: the user, the platform, or is it shared?
• Access control — who inside the organisation can see raw vs. aggregated data?
• Retention — how long is data kept, and is it deleted when no longer needed?
• Accountability — who is responsible if the data is misused or breached?

Weighing benefits against harms

Both 2023 papers ask you to conclude with the current and future implications of big data, giving named real-world use cases as prompts. Across both papers, these prompts cover:

Positive use case	Risk it also raises
Fraud detection in banking	False positives can wrongly block a legitimate customer's transactions
Personalised marketing in retail	Relies on detailed behavioural profiling, which some customers may not have knowingly consented to
Traffic and transportation management in smart cities	Requires continuous location tracking of residents
Predictive maintenance in manufacturing	Can be used to monitor and evaluate individual workers, not just machinery
Healthcare	Highly sensitive personal data; a wrong prediction can directly affect patient treatment
Public services	Risk of unequal service if the underlying data underrepresents some communities (see Topic 3)

Past-paper practice

What are the current and future implications of big data? Discuss the positive and negative implications, and consider the potential effects on people. (91908-cat-A-2023.pdf / 91908-cat-B-2023.pdf, Question Two (g); a similar version appears in 91908-exm-2021.pdf, Question Three (d)–(e).)

This is typically the highest-value question in the paper.

Reveal how to answer well

1. Choose one named use case rather than trying to cover all of them
2. Explain the mechanism that makes it positive (e.g. how pattern detection genuinely helps fraud detection)
3. Explain a specific, connected negative implication — not a generic "privacy is a concern", but the actual trade-off this use case creates
4. Reach a conclusion — can the benefit and the harm be balanced, and what would that require? (e.g. governance rules, transparency, opt-out options)

Exit question

Pick one use case from the table above and write a one-paragraph answer that follows all four steps in the "how to answer well" notes.

Key terms

• privacy
• ethics
• governance
• anonymisation
• aggregation
• differential privacy
• profiling

1. 1
2. 2
3. 3
4. 4
5. 5

← Back to Topic 4: Tools, Technologies & AI

Back to the Big Data overview →