Examhaus — Achieve with confidence.

IB Computer Science A4.4: Ethical considerations

New IB Computer Science syllabus · first exams May 2027 · Theme A · SL & HL

A4.4 asks whether machine learning should be used, and how, not just whether it works. Both statements use Discuss, so you need balanced, well-argued answers with a conclusion.

StatementTopicCommand term
A4.4.1Ethical implications of ML in real-world scenariosDiscuss
A4.4.2Ethics of integrating technology into daily lifeDiscuss
⚠️ Command term alert: a strong "Discuss" answer has six parts:
  1. Define the issue.
  2. Give an argument for, with an example.
  3. Give an argument against, with an example.
  4. Identify the stakeholders.
  5. Weigh the trade-offs.
  6. Reach a reasoned conclusion.

Key points

A4.4.1 Ethics of machine learning

  • Issues named in the guide:
    • accountability
    • algorithmic fairness
    • bias
    • consent
    • environmental impact
    • privacy
    • security
    • societal impact
    • transparency
  • Bias in training data:
    • Historical bias: the model repeats past unfairness.
    • Representation bias: some groups are under-represented.
    • Bias can also come from how data is measured, labelled or deployed.
    • Models can amplify the bias in their data. A real example is facial recognition being less accurate for darker-skinned women.
  • ML in online communication:
    • misinformation and deepfakes
    • biased automated moderation
    • online harassment
    • re-identifying "anonymous" users
    • privacy risks from recommendation engines
  • Mitigations: diverse data, testing fairness across groups, audits, keeping a human in the loop, and regulation (such as GDPR and the EU AI Act).

A4.4.2 Technology in daily life

Ethical guidelines need continual reassessment as technology advances. Consider the effect of quantum computing, augmented and virtual reality, and the spread of AI on society, individual rights, privacy and equity, including the digital divide.

Exam-style question

Discuss the challenges that biases in training data pose for machine-learning systems. [6]

Mark scheme

  • Definition [1]: bias in the data leads the model to learn unfair or inaccurate patterns.
  • Sources [2]: any two named sources of bias, such as historical, representation, measurement, label or deployment bias.
  • Real example [1]: for example, facial recognition with lower accuracy for darker-skinned women, or a biased healthcare risk-prediction tool.
  • Amplification [1]: training can exaggerate biases already present in the data.
  • Mitigations [1]: diverse data, fairness metrics by group, audits.

Common mistakes that cost marks

  • One-sided answers. "Discuss" needs arguments on both sides and a conclusion.
  • Vague claims such as "AI is biased" without saying where the bias comes from or who is affected.
  • Ending with "it's complicated". Give a reasoned position.

Frequently asked questions

Is A4.4 studied at SL?
Yes. A4.4 is studied at both SL and HL.

Which emerging technologies are named?
Quantum computing, augmented reality, virtual reality and the pervasive use of AI.

Go further