eduKateSG · STEM CONNECTIONS · BOUNDARIES
STEM can reveal what is possible, estimate consequences and build powerful systems. It cannot by itself decide every question about what humans should permit, value or owe one another.
Medicine, law, ethics, history, culture, governance and professional authority enter when technical capability affects bodies, rights, public power and unequal human consequences.
Technical possibility → evidence → affected receiver → authority → consent → law and ethics → decision → monitored consequence → correction.
This article protects the boundary of STEM rather than weakening it.
How STEM Meets Its Boundaries: The Short Answer
STEM meets its boundary whenever a technically correct answer is not enough to decide the human action.
A medical device may measure accurately, yet a clinician must decide whether the measurement applies to a particular patient. A facial-recognition system may reach a benchmark threshold, yet law and public authority must decide whether its use is permitted. A dam may generate energy, yet communities, ecosystems, heritage and distribution of benefit remain part of the decision.
Feasibility Is Not Permission
Science can show that an intervention may work. Mathematics can estimate effects. Engineering can design it. Technology can deploy it. None of those steps automatically creates moral, legal or political authority.
“Can” is a technical question. “Should,” “who decides,” “who consents,” and “who bears the harm” require additional owners.
Medicine: From Scientific Knowledge to Authorised Care
Medicine uses Science, data, instruments and Technology, but patient care is not reducible to STEM. Diagnosis and treatment require clinical judgement, professional standards, informed consent, confidentiality, attention to comorbidity and responsibility for follow-up.
A model can estimate risk for a population while being inappropriate for one patient. A test can detect a marker without deciding what the person should do next. A technically successful intervention can still be medically contraindicated.
The correct route is: scientific evidence → clinical interpretation → patient-specific context → authorised decision → monitoring and revision.
Veterinary Science: Similar Machinery, Different Clinical Owner
Animal health shares Biology, epidemiology, imaging, laboratory methods and pharmaceuticals with human Medicine. But species, physiology, welfare, legal ownership, ecological impact and public-health context change the judgement.
Use Veterinary World for the protected animal-health route.
Ethics: What Should Count as an Acceptable Outcome?
Optimisation requires an objective. Ethics asks whether the objective is acceptable, whose values it represents and which harms were excluded. A system can minimise average waiting time while making access worse for a vulnerable group. It can increase productivity while removing dignity, privacy or meaningful consent.
Ethical analysis should not be a final decorative paragraph. It changes requirements before design begins.
Law: Which Actions Are Authorised and Contestable?
Law allocates rights, duties, powers, procedures and remedies. Technical evidence can inform law, but it does not replace legal interpretation or due process.
| Technical question | Additional legal question |
|---|---|
| Can the system identify a person? | Under which authority, for what purpose, with what appeal? |
| Can data improve a model? | Was collection lawful, necessary and proportionate? |
| Can a robot act autonomously? | Who is responsible for harm or failure? |
| Can an algorithm rank applicants? | Can the decision be explained, challenged and corrected? |
| Can infrastructure be built here? | Which land, environmental and consultation rules apply? |
Governance: Who Owns the Decision After Deployment?
A pilot often has an enthusiastic project team. A permanent system needs accountable ownership. Governance defines decision rights, monitoring, escalation, incident response, audit, public reporting and retirement.
Without governance, responsibility can disappear across vendors, agencies, technical teams and users. Everyone controls one component while nobody owns the complete human outcome.
History and Culture: Why the Same Technology Lands Differently
Technologies enter societies with memory. Communities may interpret surveillance, medical research, land use, automation or public data through earlier experiences of trust, exclusion or harm.
A technically identical system can therefore produce different adoption and legitimacy outcomes in different places. Culture is not an irrational obstacle to be engineered away. It can contain information about values, identity and historical risk that the technical model omitted.
Accessibility: A Function Is Not Public If the Receiver Cannot Reach It
A digital service may pass functional tests while excluding users who cannot see, hear, type, navigate the interface or afford the required device. A transport system may operate safely while a wheelchair user cannot complete the route because one lift, kerb or gap breaks the chain.
Accessibility belongs in requirements, testing and procurement—not as post-deployment charity.
A Boundary Ledger for STEM Projects
- Technical owner: Which STEM field owns the mechanism?
- Human receiver: Who is affected directly and indirectly?
- Authority: Who is permitted to decide and act?
- Consent: What choice does the receiver genuinely have?
- Rights: Which legal protections and remedies apply?
- Distribution: Who receives benefit, cost, delay and risk?
- Accessibility: Can different users reach and control the system?
- Historical context: Which earlier experiences shape trust?
- Governance: Who monitors, audits, corrects and retires it?
- World Return: What changed in human life beyond the technical metric?
Common Boundary Failures
- Ethics after design: values enter only after the architecture is fixed.
- Consent theatre: the user technically agrees but has no realistic alternative.
- Authority confusion: a technical team makes a clinical, legal or political decision.
- Population-to-person leap: group-level evidence is treated as individual certainty.
- Accessibility omission: the assumed user becomes the only user the system can serve.
- Vendor diffusion: responsibility disappears across contracts and components.
- Metric substitution: a measurable proxy replaces the actual human outcome.
- Irreversibility blindness: deployment creates data, dependency or physical change that cannot easily be undone.
STEM Is Stronger When Its Boundary Is Honest
STEM does not lose authority by naming what it cannot decide alone. It gains precision. The technical evidence remains visible, while law, ethics, Medicine, history, culture and governance contribute their own standards.
The aim is not to slow every project indefinitely. It is to prevent speed from hiding the missing owner until harm makes the omission obvious.
Evidence Base and Further Reading
- NIST — AI Risk Management Framework
- NIST — AI RMF Playbook
- National Academies — Integrated STEM Education
- UNESCO — STEM
Return to STEM, How X Works or World & Knowledge.
Keep the owner clear, preserve the handoff, and return to the world to see whether the capability actually worked.