4.1 Ethical and Social Issues Around Data Collection

1. Personal data is a privacy risk

Every time someone uses a computer, an app or a website, personal data can be collected: names, locations, contacts, search history, health or school records. Once this data is stored on a computer system, the person’s privacy is at risk 4.1.A.1. Stored data can be stolen in a data breach, shared with other companies, or combined with other data to reveal more than the user ever agreed to share.

Programmers make choices that decide how big that risk is. When you develop a new program, you should try to safeguard the personal privacy of the user 4.1.A.1:

• Collect only what you need. A fitness app that counts steps does not need the user’s contact list.
• Protect what you store. Keep personal data away from places where other users or programs can read it.
• Keep it only as long as needed. Old data that is deleted cannot leak.
• Be honest with users. Tell them what is collected and why.

Watch out: “The data is only stored, not shared” does not make it safe. Storage itself is the risk.

2. Algorithmic bias

Algorithmic bias describes systemic and repeated errors in a program that create unfair outcomes for a specific group of users 4.1.B.1. “Systemic” means the error comes from how the program or its data is built, so it happens again and again to the same group.

Example: a face-unlock feature is trained mostly on photos of adults. It works well for adults but often fails for children. The program has no single “bug” line, yet it keeps treating one group of users unfairly. The cause is the data used to build it.

Feature	One random error	Algorithmic bias
How often?	Once, by chance	Repeated
Who is affected?	Anyone	A specific group of users
Where it comes from	A slip in one run	The design or the data (systemic)

3. How was the data collected?

Before you use a data set to extrapolate new information or draw conclusions, you should know how the data was collected and whether that method could bring bias 4.1.B.2.

A school wants to know how students travel to school. Compare two collection methods:

Collection method	Possible bias
An online form shared only in the cycling club group chat	Cyclists are over-represented, so “most students cycle” would be a biased conclusion.
A short question asked to every student at the start of first lesson	Everyone has the same chance to answer, so the result is much more representative.

Every answer in the first data set may be true, but the method still produces a biased picture.

4. Incomplete and inaccurate data

Some data sets are incomplete (values are missing) or inaccurate (values are wrong). If a program is developed or used with such data, it can work incorrectly or inefficiently 4.1.B.3.

In this example, a weather station failed on one day and stored -999.0 as a “missing” marker. The program averages the readings without knowing that.

double day1 = 21.5;
double day2 = 23.0;
double day3 = -999.0;   // sensor failed; this is not a real temperature
double day4 = 22.5;

double average = (day1 + day2 + day3 + day4) / 4;
System.out.println("Average temperature: " + average);

Output:

Average temperature: -233.0

The algorithm is correct, but the inaccurate data makes the answer useless. A program can also become inefficient with bad data, for example by processing thousands of duplicate records.

Watch out: when a question describes missing values, typing errors or duplicate records, the expected answer is usually that the program may produce incorrect results.

5. Choosing the right data set

The contents of a data set are usually related to a specific question or topic. A data set collected for one question might not give correct answers, or support extrapolation, for a different question 4.1.C.1.

Question: “Which hour of the day is busiest at a city library?”

Data set	Appropriate?	Why
Library entrance gate counts, by hour, for the last year	Yes	It measures visits per hour at that library.
Number of books borrowed per month	No	Monthly totals cannot show hours, and borrowing is not the same as visiting.
Gate counts, by hour, from a different city’s library	No	It is about a different library, so it cannot answer the question for this one.
A survey asking residents their favorite book	No	It is about a different topic.

To choose, check that the data set matches the question’s topic, place (population) and time period, and that it measures the thing the question asks about.