Data Privacy in Data Ethics#

📦 Data Preparation ⚖️ Bias & Data Ethics Lesson 014

Important

✨ AI-generated content. This page was written with the assistance of an AI language model and is provided as a learning aid. Despite careful review, it may still contain mistakes, omissions, or out-of-date information. Whether you are new to the topic, a team lead, or a senior practitioner, treat it as a starting point rather than an authoritative reference: read it critically and independently verify anything you act on (code, commands, figures, and factual claims) against official documentation and primary sources before relying on it.

Privacy as protection#

Data privacy is the aspect of data ethics concerned with protecting a person’s information — the right to control how their personal data is collected, used, shared, and retained. Where ownership and consent establish whose data it is and whether it may be used, privacy governs how it is protected once entrusted, and it is often the most legally regulated part of an analyst’s obligations.

What privacy requires#

Several concrete duties follow from the privacy principle:

Protect personal data — secure it against unauthorised access, with the access controls, encryption, and careful handling the data-security lesson addresses. A breach is a privacy failure regardless of intent.
Limit use to purpose — use personal data only for the purpose it was collected and consented for. Repurposing data for a new use people never agreed to violates privacy even when the data is held securely.
Minimise — collect and retain only the personal data actually needed, and no longer than needed. Data you do not hold cannot be breached or misused; the life cycle’s destroy stage is a privacy control.
Guard identity — recognise that combining datasets or retaining detail can make supposedly anonymous data re-identifiable.

Personally identifiable information (PII)#

The heart of privacy is PII — information that can identify a specific individual, directly (name, government ID, email) or indirectly (a combination of attributes that together single someone out). Analysts handle PII with particular care: accessing it only when necessary, and preferring to work with data from which identifiers have been removed.

Anonymisation (removing identifying information so individuals cannot be recognised) is the standard protective technique — but it is harder than it looks. The recurring lesson from real re-identification cases is that stripping obvious identifiers is often insufficient: a combination of “anonymous” attributes — a postal code, a birth date, a gender — can uniquely identify a person even with names removed. Genuine anonymisation must consider what the remaining fields, and any dataset they could be joined with, reveal in combination.

Privacy in the analyst’s workflow#

Practically: access the minimum PII the task requires; work with anonymised or aggregated data by default; never join datasets in ways that re-identify people without authorisation; and treat “could this analysis expose an individual?” as a question asked before running it. Aggregation — reporting groups rather than individuals — is a workhorse privacy technique, though even aggregates can leak when a group is small enough to point at one person.

The caveat#

Privacy exists in tension with analytical usefulness: the same detail that makes data powerful can make it privacy-invasive, and stronger protection often means coarser data. There is no universal setting — the right balance depends on the data’s sensitivity, the purpose, applicable law, and what people consented to. The obligation is to weigh usefulness against protection deliberately, default toward protecting people, and know the legal requirements that apply. The next lesson turns to the opposite impulse: data that should be open.

Hint