TOPIC 4.1 · STUDY

Ethical and Social Issues Around Data Collection

Data choices have privacy and fairness costs. Collect only what the program needs, watch for bias and gaps in who or what is measured, and match each data set to the question you actually asked.

4.1.A

Collect only what the program needs

This topic is about data choices, not Java class syntax. When a program stores personal details, those details can identify someone, contact them, or embarrass them if the store leaks or is reused for a new purpose. Programmers help decide which fields are required and which are extra risk.

A useful rule is to collect only what the program needs for its stated job. A school theater kiosk that prints a seat stub needs a seat letter and a show title. Storing a full legal name, a street address, and a health note for that stub creates a privacy problem: the ticket does not need those fields, yet they remain available if the file is shared or stolen.

A safeguard keeps the useful data and drops the rest. For the seat stub, keep the seat letter and show title; discard or never request the address and health note. The same idea applies to loyalty stamps, bike claim tags, or a library hold code: store an identifier that fulfills the task, not a full personal profile.

Stored field Needed for a seat stub? Privacy note
Seat letter Yes Identifies the chair without naming the person.
Show title Yes Tells which performance the stub belongs to.
Street address No Can locate a household if the file leaks.
Health note No Sensitive detail unrelated to printing a seat.
All collected fields enter a safeguard box. Useful fields continue to Keep useful; extras branch to Drop extras.

A privacy safeguard keeps fields the program needs and drops personal extras the task does not require.

Key points

  • Personal data can identify, contact, or embarrass someone if it leaks or is reused.
  • Collect only fields the stated program job requires.
  • Extra sensitive fields on a simple printout are a privacy problem.
  • A safeguard keeps useful data and drops or never stores the rest.
  • Programmers share responsibility for those collection choices.

Needing a seat letter does not justify storing a home address or a health note beside it.

4.1.B

Bias and gaps in data sets

A data set can be biased when it over-represents some groups and leaves others out. Programs trained or decisions made on that data then favor the groups that were measured. Bias can appear even when no one typed an insulting label into the code.

How you collect the sample matters. A survey that asks only students who walk through the cafeteria during one short lunch window does not represent every student. Morning-only visitors, late lunch periods, and people who pack lunch never appear. The dinner-plan vote then leans toward whoever happened to be in that narrow window.

Incomplete or inaccurate records also harm quality. If a garden moisture log skips three rainy days and stores them as zero, a watering program may soak soil that is already wet. Missing values treated as ordinary zeros are still wrong inputs for the decision.

Problem What goes wrong Everyday picture
Over-representation Some groups dominate the sample. Almost all training rows come from two zip codes.
Collection bias The method excludes people who matter. Only one short lunch window is surveyed.
Incomplete data Gaps are filled or ignored incorrectly. Missed rain days stored as 0 mm.

Key points

  • Bias means some groups are over-represented and others are left out.
  • A narrow collection window can miss the people the question cares about.
  • Unfair outcomes can follow from skewed training or survey data without hostile intent.
  • Incomplete or inaccurate values can drive the wrong action.
  • Improving who is included and how gaps are handled improves data quality.

A large file is not automatically fair. Who is missing from the rows can matter as much as how many rows you have.

4.1.C

Match the data set to the question

Before you analyze anything, name the question in plain words. Then ask whether each available data set records the quantities that answer it. Related-sounding files from the same place are not automatically useful.

Rainfall totals by day answer a rainfall question such as which day was wettest. Door counts by hour answer a busiest-hour question. A snack-menu price list does not answer either one, even if the snack cart sits next to the same garden gate or gym door.

Choose the set whose fields line up with the ask. If you want the busiest gym hour, use entries that store an hour and how many people entered. If you want which Saturday showtime sold the most tickets, use ticket counts by showtime, not a popcorn-flavor poll from last year.

Question Fitting data Poor fit
Which day had the most rain? Daily rainfall totals Snack flavors sold that week
Which hour is the door busiest? Entries counted by hour Menu prices or pool temperature
Which showtime sold most? Tickets by showtime An unrelated flavor survey

Key points

  • State the question first, then judge each data set against it.
  • Fitting data records the quantities the question names.
  • Same-building or same-theme files can still be a poor fit.
  • Door counts by hour fit busiest-hour questions; daily rain fits rainfall questions.
  • A convenient nearby sample is not automatically the right sample.

A convenient sample is not the same as the question you asked.

Practice this topic