Not sure which lesson to do first? Find your level and get a plan →

CS Thinking · Systems, Networks & Impact · Grade 9-12

Ethics of Computing

The study of moral issues and responsibilities that arise from the development and use of computing technology.


Quick Answer

The study of moral issues and responsibilities that arise from the development and use of computing technology. You use it whenever a technology decision affects people beyond the person making it: an app that collects location data, an algorithm that screens job applicants, a camera that recognizes faces. The method is to list everyone affected, weigh what each gains and loses, check the data for bias and consent, and then recommend whether to build the system, change it, or limit it with safeguards. A product can be legal and technically excellent and still be the wrong thing to ship.

Why This Matters

Computing systems make the same decision millions of times, so one unfair rule repeats at scale: a screening model that marks down one group's applications does it to every applicant from that group, every day. Ethics gives you a structured way to catch that before launch and to argue for a fix with evidence instead of a feeling.

Intuitive Explanation

Imagine a ride app that raises prices when demand jumps. During a snowstorm, fares triple. The code works perfectly: it measured demand and raised the price exactly as designed. Whether it should do this is a different question, and that question has several sides. Drivers earn more, so more of them come out; riders who can pay get a car sooner; a nurse trying to reach a hospital shift may be priced out. Ethics of computing is the habit of asking the second question as seriously as the first.

A good analysis is concrete. Name the stakeholders, including people who never chose the technology, like the pedestrian near a self-driving car or the friend tagged in your photo. State what each one gains or loses. Check the data: who collected it, whether people agreed, and whether it records past unfairness that the system will copy. Then decide: build it, change it, or limit it with safeguards such as human review, opt-outs and regular audits.

Two things look like ethics but are not the whole of it. A legal question asks what a law or a license allows, and it has a factual answer; an ethical question asks what is right, and a perfectly legal action can still be unfair. A personal opinion such as "I'd never use that app" is not an ethical argument either, until it says who is affected and how. You can usually tell an ethics question by the word "should" and by several groups whose interests pull in different directions.

Think of it like this. An ethics review works like an environmental impact study for software. Before a highway is built, someone studies who gains the shortcut, whose homes get the noise, and which wetland floods. A new app deserves the same study of who gains and who pays. Unlike a highway, software keeps changing after launch, so the review never really ends.

Core idea. Computing ethics covers bias in algorithms, digital divide, environmental impact, automation and jobs, misinformation, and responsible AI development.


When to Use

Use it when a question asks whether a technology should be built or used, or how its effects on different people compare: words like fair, harm, bias, consent and responsibility. It fits when the answer depends on weighing several groups' interests. If the question is only what a law or license permits, use Intellectual Property; if it is about control of one person's data, start with Privacy; if it is whether disabled users can operate a product, use Accessibility.

Tip. Ask: Is the question whether a technology should be built or used, given who gains and who is harmed, not only whether it works?

How to Recognize It

Before you treat it as an ethics question, list the people the technology touches and what each of them stands to gain or lose.

  1. Does the decision affect people who did not choose it, such as job applicants, neighbors or future users?

    Yes is the core of an ethics question. If only the user is affected and the issue is getting the tool to work, it is a design or Debugging question.

  2. Is the action allowed by law or by the terms of service, yet still possibly unfair?

    Yes keeps it in ethics, which asks more than the law does. If the question is only what a license or copyright allows, it points to Intellectual Property.

  3. Is the main issue someone's control over their own personal data?

    Yes points to Privacy, one part of computing ethics with its own laws. Ethics weighs privacy against other values, such as safety or fairness.

  4. Is someone attacking a website, network or device on purpose?

    Yes points to Cybersecurity, which defends against the attack. Ethics comes in when deciding, for example, whether to report a flaw you found or keep it quiet.

Ethics of Computing vs Privacy vs Intellectual Property vs Cybersecurity

Privacy, ownership and security are topics inside computing ethics, and each also has its own laws and tools. Ask whether you are applying a specific rule or weighing what is right where the rules run out.

Ethics of Computing

Meaning
Ethics of computing weighs the benefits and harms of a technology for everyone it affects and asks what people should do.
Key test
No single rule settles it; you must weigh competing interests.
Example
Deciding whether a hospital should let an algorithm rank which patients get scans first.

Privacy

Meaning
Privacy concerns a person's control over the collection, storage and sharing of their personal data.
Key test
One specific right is at stake, often backed by laws such as the GDPR.
Example
A teenager asking a social app to delete everything it has stored about them.

Intellectual Property

Meaning
Intellectual property law sets who may copy, sell or adapt creative works and inventions.
Key test
A license or law gives a definite answer about permission.
Example
An open-source license lets anyone reuse code as long as they keep the author's copyright notice.

Cybersecurity

Meaning
Cybersecurity protects systems and data from deliberate attacks on their secrecy, accuracy or availability.
Key test
There is an attacker to stop, and technical defenses answer it.
Example
Patching a school server after finding a flaw that lets outsiders read its files.

Formula & Notation

Worked Examples

Example 1 — Stakeholder analysis: scanning student messages

Easy

A school district tests software that scans students' school-account messages for words like "kill" and alerts the principal. In one month it sends 120 alerts: 4 are real safety concerns, and 116 are about video games or schoolwork ("kill the process", "this test will kill me"). Should the district keep it, and on what terms?
Show the steps and the answer
  1. List the stakeholders: students in danger, every student whose messages are read, the counselors who handle alerts, parents, and the company that stores the messages.

    Each group gains or loses something different, and the company is easy to forget even though it holds the data.

  2. Measure the benefit and the cost: 4 real concerns found, but 116/120≈97% of alerts were false.

    Each false alert means a student called in over a joke or a complaint about homework. Numbers turn "it keeps kids safe" and "it's creepy" into claims you can weigh.

  3. Check consent and data use: do students and parents know messages are scanned, and how long are they kept?

    Monitoring that people don't know about can't be questioned, and stored messages can later be used for purposes nobody agreed to.

  4. Check for bias: compare which groups of students receive the false alerts.

    If slang used more by one group triggers more alerts, those students carry more of the cost.

  5. Recommend terms: a counselor reviews every alert before any student is contacted, families are told in writing, messages are deleted after 30 days, and false alerts are audited by group each term.

    These safeguards keep the 4 real catches while cutting the harm of the 116 false ones.

Answer: Keep it only with safeguards: human review before contact, written notice to families, 30-day deletion and a regular bias audit. Without them, a 97% false-alert rate makes the harm outweigh the benefit.

Takeaway: An ethics answer names every stakeholder, puts numbers on benefits and harms where it can, and ends in a decision with its conditions.

Example 2 — Legal is not the same as right

Standard

PhotoPals' terms of service, which users accept at sign-up, let the company use uploaded photos "to improve our services". It then uses 50 million user photos to train a face-recognition tool that it sells to stores to spot shoplifters. Sort the questions this raises into Intellectual Property and Ethics of Computing, and answer the ethics question.
Show the steps and the answer
  1. Intellectual Property question: users own the copyright in their photos, so does the license in the terms cover this use?

    That has a factual answer in the contract and the law, which a lawyer could settle. Suppose the answer is yes.

  2. Ethics question: even if the license allows it, should PhotoPals do it?

    Being allowed and being right are separate standards, so a yes to the legal question leaves this one open.

  3. Weigh the stakeholders: users expected better photo filters, not store surveillance; friends in group photos never signed anything; shoppers are scanned by a system trained on strangers' faces.

    The people pictured who never joined PhotoPals had no chance to agree at all, and a license signed by someone else can't give consent for them.

  4. Recommend: ask users to opt in, remove the faces of non-users, and require stores to tell shoppers they are being scanned.

    Each safeguard targets one harm found above: surprise use, missing consent and hidden surveillance.

Answer: The license question is Intellectual Property and may well be answered yes. The ethics question is answered no as things stand, because people in the photos never consented and users never expected this use; opt-in consent would change that.

Takeaway: "What does the contract or law allow?" is a legal question; "should we, given who is affected?" is an ethics question, and a legal yes does not settle it.

Example 3 — Find the mistake: "the algorithm can't be biased"

Application

A bank's loan model uses income, credit score and ZIP code, but never race. Jordan writes: "The model never sees race, so its decisions are objective and can't be biased." In a test, applicants with the same $50,000 income and the same 700 credit score are approved 72% of the time in ZIP code A and 31% of the time in ZIP code B, where most residents are Black. Find the mistake and correct it.
Show the steps and the answer
  1. Compare outcomes, not inputs: identical income and credit score, yet 72% versus 31% approval.

    The only difference between these applicants is their ZIP code, so ZIP code alone is producing a gap of 41 percentage points.

  2. Ask why ZIP code would matter: the model learned from past loan decisions, and for decades banks refused loans in mostly Black neighborhoods, a practice called redlining.

    The training data records that history, so ZIP code can carry race into the model without race ever being an input. An input that stands in for a missing one is called a proxy.

  3. Name Jordan's error: leaving a column out does not make a model objective.

    A model copies the patterns in human-made data, unfair ones included, through any input linked to the missing one.

  4. Fix: test approval rates for equally qualified applicants across groups, drop ZIP code or show it predicts repayment for a fair reason, and let rejected applicants appeal to a person.

    Fairness has to be measured in outcomes; a test like this one is exactly how the gap was found.

Answer: The model can be biased without seeing race: ZIP code acts as a proxy, giving 72% versus 31% approval for identical finances. Audit outcomes by group and drop or justify the proxy.

Takeaway: Judge an algorithm by what it does to equally qualified people, not by which columns it reads.

Where You'll See It

  • A hiring algorithm that favors certain applicants

    If the system was trained on biased past hiring data, it can unfairly screen out qualified people — an ethics question about who is harmed, even if the tool is technically working as built.

  • A 'free' app that sells your data

    Collecting and selling user data may be legal under the fine print, but ethics asks whether it's fair to users who didn't really understand what they agreed to.

  • Self-checkout machines replacing cashier jobs

    Automation can lower costs and speed up lines, but ethics weighs that benefit against the workers whose jobs are displaced — asking who gains and who loses.

Common Mistakes

  1. The slip: If something is legal, it's automatically ethical.

    Why it's tempting: Laws feel like society's official line for right and wrong.

    Instead: Legality and ethics are different standards; some legal actions are still unfair or harmful, and ethics asks who benefits and who is harmed beyond what the law requires.

  2. The slip: Automated systems are inherently objective and fair.

    Why it's tempting: A machine has no feelings, so it seems incapable of prejudice.

    Instead: Algorithms learn from human-made data and choices, so they can encode and even amplify bias — fairness must be actively tested, not assumed.

  3. The slip: Ethics is only the developer's responsibility.

    Why it's tempting: The developer builds the system, so the blame seems to land entirely on them.

    Instead: Users, companies, and policymakers all share ethical responsibility for how computing technology is built, deployed, and used.


Mini Practice

Try these on your own. Tap Reveal when you want to check.

  1. A school plans an app that tracks students' phone locations during school hours to take attendance. Name two benefits, two harms and one safeguard.

    Hint: Think about students, parents, teachers, and students who have no phone.

  2. A hiring algorithm screened 200 men and 200 women with similar qualifications. It passed 80 men and 40 women. What is each group's pass rate, and what does the gap suggest?

    Hint: Divide the number passed by the number screened for each group.

  3. Under the "four-fifths rule", a group's selection rate below 80% of the highest group's rate is a warning sign. A loan tool approves 60% of group A and 45% of group B. Is there a warning sign?

    Hint: Divide the lower rate by the higher rate and compare with 0.8.

  4. Training one large AI model used about 1,300 megawatt-hours of electricity. A typical home uses about 10 megawatt-hours a year. How many home-years of electricity is that, and why might it matter ethically?

    Hint: Divide the training energy by one home's yearly use.

  5. Question A: should social media companies design feeds that keep teens scrolling for hours? Question B: can I stop this app from reading my contacts? Which is a privacy question, and which is the broader ethics of computing?

    Hint: One is about a single person's data; the other weighs a design choice for many people.

  6. A company says: "Our terms of service let us sell users' data, and that is legal in our country, so there is no ethical problem." What is wrong with this reasoning?

    Hint: Are "legal" and "right" always the same thing?

Want the full set?

50 practice questions for this concept — free to try, every one with a complete worked solution showing the why, not just the answer.

Frequently Asked Questions

If a computer makes the decision, who is responsible when it goes wrong?

People are. Someone chose the data, built the model, decided to use it and set how closely a human checks its output. When an automated system wrongly denies a loan or flags an innocent person, responsibility is shared by the builders, the organization that deployed it and the officials who approved it. Newer rules, such as the EU AI Act, require human oversight of high-risk systems.

Is it ethical to report a security flaw I found in my school's website?

Reporting it privately to the people who run the site is the responsible choice, and many organizations welcome it. Do not test further, copy data or tell others first, because probing a system without permission can break rules or laws even with good intentions. Security researchers call this responsible disclosure: tell the owner, give them time to fix it, and only then talk about it.

Why do people worry about the environmental cost of computing?

Data centers use a lot of electricity, roughly 1 to 2 percent of the world's total in recent years, and training large AI models adds to it. Many also use water for cooling. Ethics asks who bears those costs, such as communities near power plants, and whether a given use is worth it. Companies respond by building where power is clean and by making chips and code more efficient.

Are ethics just personal opinions?

Not entirely. People do disagree, but ethical arguments use reasons others can check: who is harmed, whether consent was given, whether similar people are treated alike, and what would happen if everyone did the same. Professional groups such as the ACM publish codes of ethics that computing professionals are expected to follow. Two people can reach different answers and still both show careful reasoning.

Learning Path

Ethics of Computing

You are here

Next →

Nothing after this yet
You need Privacy (control over personal data), Intellectual Property (who owns creative work) and Accessibility (who can use a design) first; each is one lens on a narrower question. Ethics of Computing combines them into a full stakeholder analysis that ends in a recommendation. From here you can write an impact assessment for a real proposal, such as facial recognition at a school entrance, or take a side in a debate on AI rules and back it with evidence.

See Also