COMP90087 · The Ethics of Artificial Intelligence
Deontology
Week 4 covers duty-based ethics: rightness comes from duties, rules and rights, not consequences. The exam explicitly tests Kant’s Formula of Humanity — always treat rational beings as ends in themselves, never merely as means — as a near-verbatim MCQ, and W. D. Ross’s prima facie duties recur in the bias week. Being able to run a deontological argument (name the duty, show the act violates or honours it, resolve conflicts) is a core essay skill and a reliable contrast to utilitarianism.
What this chapter covers
- 01Deontology = duty/rule/rights-based ethics: some acts are wrong in themselves even if they produce good outcomes
- 02Kant’s good will — acting from duty, not mere inclination — as the only thing good without qualification
- 03The categorical imperative and its formulations: Formula of Universal Law (universalisability) and Formula of the Kingdom of Ends
- 04Formula of Humanity (exam item): treat rational beings as ends in themselves, never merely as means
- 05W. D. Ross’s prima facie duties — fidelity, reparation, gratitude, justice (plus beneficence, non-maleficence, self-improvement) that can outweigh one another in context
- 06Applying deontology to AI: consent, prohibition of deception, limits on surveillance and opaque automated decisions
- 07Standard objections: rigid rules can conflict or give counter-intuitive results; Ross gives no fixed ranking of duties; can seem to ignore overall welfare
- 08Kant vs Aristotle contrast: for Kant feeling/inclination is inessential; the right motive is following duty and reason
Apply Kant’s Formula of Humanity to non-consensual data reuse
- +1State the test. Kant’s Formula of Humanity: always treat rational beings (yourself and others) as ends in themselves, never merely as means. This is the near-verbatim exam answer — give it precisely.
- +1Apply it. By harvesting users’ notes without consent to serve a commercial goal, the app uses people merely as means (as data-sources) and bypasses their rational agency — impermissible on Kantian grounds regardless of how much aggregate benefit the model brings.
- +1Name a salient duty (Ross). Fidelity (honesty / keeping faith with users about how their data is used) is breached; you could also cite non-maleficence. If duties conflict you would weigh the prima facie duties for the situation, since Ross gives no fixed ranking.
- +1Show the divergence. A utilitarian might permit the reuse if the expected benefit to millions outweighs the privacy harm — so utilitarianism and deontology can reach opposite verdicts on the same facts. That divergence is the examinable meta-point.
Key terms
- Deontology
- Duty-based ethics: rightness is fixed by duties, rules and rights rather than consequences, so some acts (lying, killing the innocent) are wrong in themselves even when they produce good outcomes.
- Categorical imperative
- Kant’s unconditional command of reason. Its Formula of Universal Law tests whether you could will your maxim to become a universal law; the Formula of the Kingdom of Ends asks you to act as a legislating member of a community of rational ends.
- Formula of Humanity
- Kant’s formulation, and the exam’s key item: always treat rational beings as ends in themselves, never merely as means. It grounds AI limits on deception, non-consensual data use and uncontestable automated decisions.
- Good will
- For Kant, the only thing good without qualification — acting from duty rather than from inclination or expected reward. The moral worth of an act lies in its motive.
- Prima facie duties (Ross)
- W. D. Ross’s pluralist deontology: several conditional duties — fidelity, reparation, gratitude, justice, plus beneficence, non-maleficence and self-improvement — that can outweigh one another depending on the situation, with no fixed ranking.
- Fidelity
- Ross’s duty to keep promises, honour agreements and tell the truth. In AI cases it is breached by misleading users about how their data or an automated decision works.
Deontology FAQ
What is the exact wording examiners want for Kant’s Formula of Humanity?
“Always treat rational beings (humanity, in yourself and others) as ends in themselves, never merely as means.” This appears as a near-verbatim MCQ, so learn it word-for-word, and be ready to apply “merely as means” to an AI case (deception, non-consensual data use, uncontestable automated decisions).
How is Ross’s deontology different from Kant’s?
Kant offers a single unconditional test (the categorical imperative); Ross offers several prima facie duties — fidelity, reparation, gratitude, justice and more — that hold conditionally and can be outweighed by one another in context. Ross deliberately gives no fixed ranking, which is both his flexibility and a standard criticism. Ross’s duties resurface in the algorithmic-bias week.
When should I pick deontology over utilitarianism in an essay?
Choose deontology when the case turns on consent, honesty, rights or using people as mere means — situations where a good outcome does not settle the matter. It is the natural lens for non-consensual surveillance, deception or opaque automated decisions. A powerful essay move is to run the case under deontology and then show a utilitarian reaching the opposite verdict on the same facts.
Can AI help me practise deontological arguments for COMP90087?
Yes. Sia can test your recall of the Formula of Humanity, set fresh cases where you must name the salient duty and resolve a conflict, and check that your Kantian and utilitarian verdicts genuinely diverge — explaining each step. It mirrors how the University of Melbourne assesses this and never writes your submitted essay; the subject’s academic-integrity and GenAI rules apply.
Exam move
Memorise the Formula of Humanity verbatim — it is the single most-tested deontology item — and keep Ross’s four core duties (fidelity, reparation, gratitude, justice) on the same flashcard so you can also answer attribution MCQs. Practise the four-move argument: state your stance, name the variant (Kantian vs Ross), run the duty reasoning, and rebut one objection. Rehearse the same case under both deontology and utilitarianism so you can show, and not just assert, that the theories diverge — the examinable meta-point. Because the exam is a closed-book hurdle, make the Formula and the duty list automatic recall rather than SWOTVAC cramming.
Working through Deontology in COMP90087? Sia is AskSia’s AI AI Ethics tutor — ask any COMP90087 Deontology question and get a clear, step-by-step explanation grounded in how COMP90087 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.
