Why Science? | Singapore-MIT Alliance for Research and Technology

A shared mission gives different disciplines a reason to connect. Each team then states the part it can test, the evidence it produces and the decisions that depend on it.

Interdisciplinary research succeeds when each field keeps its standards while making interfaces explicit. Shared words, data formats and decision rights matter as much as exciting ideas. The practical capability is problem-framing literacy. Keep the question, method, data, assumptions, uncertainty, review and next decision in one dated evidence receipt. SMART, MIT, NRF and participating institutions control current programmes and released results; ethics, safety, clinical, data and product authorities control applicable studies and deployment.

Put this into practice by turning the idea into a small, reviewable action: Build a one-page evidence card for “Interdisciplinary work begins with a shared problem” with the question, source or system, method, observation, limitation and next check. Then ask what the resulting note or artefact actually demonstrates. A useful record names context, source or instruction, action, observation, interpretation and uncertainty. It also preserves this limit: A large mission is not itself a testable hypothesis.

An evidence check keeps confidence proportional to what the work can show. Separate the original observation or official statement from the learner’s explanation. Record feedback and any revision. If information is missing, write “not shown” or “not yet known” rather than guessing. That honest gap can become the next question.

Contents · Next chapter

Terms such as performance, safety, intelligence or quality can mean different things across fields. Teams define how each quantity or category will be observed and measured.

A research group can connect biology, engineering, computation and social systems, yet every claim still needs the evidence appropriate to its own scale. The practical capability is cross-disciplinary vocabulary. Keep the question, method, data, assumptions, uncertainty, review and next decision in one dated evidence receipt. SMART, MIT, NRF and participating institutions control current programmes and released results; ethics, safety, clinical, data and product authorities control applicable studies and deployment.

Put this into practice by turning the idea into a small, reviewable action: Build a one-page evidence card for “Words need operational definitions” with the question, source or system, method, observation, limitation and next check. Then ask what the resulting note or artefact actually demonstrates. A useful record names context, source or instruction, action, observation, interpretation and uncertainty. It also preserves this limit: Agreement on a word does not guarantee agreement on its measurement.

An evidence check keeps confidence proportional to what the work can show. Separate the original observation or official statement from the learner’s explanation. Record feedback and any revision. If information is missing, write “not shown” or “not yet known” rather than guessing. That honest gap can become the next question.

Contents · Previous chapter · Next chapter

Samples, sensors, files, models and human handoffs connect work packages. Naming units, formats, tolerances, ownership and acceptance criteria prevents invisible assumptions from accumulating.

Translation is a sequence of bounded tests. A prototype, a pilot, a validated process and an adopted solution answer different questions. The practical capability is systems-interface reasoning. Keep the question, method, data, assumptions, uncertainty, review and next decision in one dated evidence receipt. SMART, MIT, NRF and participating institutions control current programmes and released results; ethics, safety, clinical, data and product authorities control applicable studies and deployment.

Put this into practice by turning the idea into a small, reviewable action: Build a one-page evidence card for “Interfaces deserve their own design” with the question, source or system, method, observation, limitation and next check. Then ask what the resulting note or artefact actually demonstrates. A useful record names context, source or instruction, action, observation, interpretation and uncertainty. It also preserves this limit: A strong component can still fail at a weak interface.

An evidence check keeps confidence proportional to what the work can show. Separate the original observation or official statement from the learner’s explanation. Record feedback and any revision. If information is missing, write “not shown” or “not yet known” rather than guessing. That honest gap can become the next question.

Contents · Previous chapter · Next chapter

Shared protocols, reference materials, code versions and inter-laboratory comparisons test whether a result survives different people, instruments and settings.

International collaboration adds expertise and comparison, while provenance, authorship, safety and local context keep the work accountable. The practical capability is reproducibility planning. Keep the question, method, data, assumptions, uncertainty, review and next decision in one dated evidence receipt. SMART, MIT, NRF and participating institutions control current programmes and released results; ethics, safety, clinical, data and product authorities control applicable studies and deployment.

Put this into practice by turning the idea into a small, reviewable action: Build a one-page evidence card for “Reproducibility crosses laboratories” with the question, source or system, method, observation, limitation and next check. Then ask what the resulting note or artefact actually demonstrates. A useful record names context, source or instruction, action, observation, interpretation and uncertainty. It also preserves this limit: Replication requires comparable conditions, not vague similarity.

An evidence check keeps confidence proportional to what the work can show. Separate the original observation or official statement from the learner’s explanation. Record feedback and any revision. If information is missing, write “not shown” or “not yet known” rather than guessing. That honest gap can become the next question.

Contents · Previous chapter · Next chapter

Resistance research can connect microbial evolution, drug action, diagnostics, treatment systems and public-health behaviour. Each scale needs distinct samples, controls and outcome measures.

Interdisciplinary research succeeds when each field keeps its standards while making interfaces explicit. Shared words, data formats and decision rights matter as much as exciting ideas. The practical capability is AMR systems literacy. Keep the question, method, data, assumptions, uncertainty, review and next decision in one dated evidence receipt. SMART, MIT, NRF and participating institutions control current programmes and released results; ethics, safety, clinical, data and product authorities control applicable studies and deployment.

Put this into practice by turning the idea into a small, reviewable action: Build a one-page evidence card for “Antimicrobial resistance joins mechanism with use” with the question, source or system, method, observation, limitation and next check. Then ask what the resulting note or artefact actually demonstrates. A useful record names context, source or instruction, action, observation, interpretation and uncertainty. It also preserves this limit: Laboratory susceptibility does not by itself prescribe patient care.

An evidence check keeps confidence proportional to what the work can show. Separate the original observation or official statement from the learner’s explanation. Record feedback and any revision. If information is missing, write “not shown” or “not yet known” rather than guessing. That honest gap can become the next question.

Contents · Previous chapter · Next chapter

Cell therapies vary because cells grow and respond. Manufacturing research measures identity, potency, purity, consistency and process conditions while preserving biological function.

A research group can connect biology, engineering, computation and social systems, yet every claim still needs the evidence appropriate to its own scale. The practical capability is biomanufacturing reasoning. Keep the question, method, data, assumptions, uncertainty, review and next decision in one dated evidence receipt. SMART, MIT, NRF and participating institutions control current programmes and released results; ethics, safety, clinical, data and product authorities control applicable studies and deployment.

Put this into practice by turning the idea into a small, reviewable action: Build a one-page evidence card for “Cell-therapy manufacturing treats living cells as products” with the question, source or system, method, observation, limitation and next check. Then ask what the resulting note or artefact actually demonstrates. A useful record names context, source or instruction, action, observation, interpretation and uncertainty. It also preserves this limit: A laboratory cell result is not an approved therapy.

An evidence check keeps confidence proportional to what the work can show. Separate the original observation or official statement from the learner’s explanation. Record feedback and any revision. If information is missing, write “not shown” or “not yet known” rather than guessing. That honest gap can become the next question.

Contents · Previous chapter · Next chapter

Sensors, molecular tools and models can track plant state, nutrients and stress. Field heterogeneity and seasons test whether a controlled-system result remains useful in production.

Translation is a sequence of bounded tests. A prototype, a pilot, a validated process and an adopted solution answer different questions. The practical capability is agricultural-sensing literacy. Keep the question, method, data, assumptions, uncertainty, review and next decision in one dated evidence receipt. SMART, MIT, NRF and participating institutions control current programmes and released results; ethics, safety, clinical, data and product authorities control applicable studies and deployment.

Put this into practice by turning the idea into a small, reviewable action: Build a one-page evidence card for “Precision agriculture measures plants and environments” with the question, source or system, method, observation, limitation and next check. Then ask what the resulting note or artefact actually demonstrates. A useful record names context, source or instruction, action, observation, interpretation and uncertainty. It also preserves this limit: A greenhouse signal may not predict field yield.

An evidence check keeps confidence proportional to what the work can show. Separate the original observation or official statement from the learner’s explanation. Record feedback and any revision. If information is missing, write “not shown” or “not yet known” rather than guessing. That honest gap can become the next question.

Contents · Previous chapter · Next chapter

Useful automation depends on interface design, skills, work allocation, trust and organisational rules as well as algorithm performance. Research should measure effects on people and tasks.

International collaboration adds expertise and comparison, while provenance, authorship, safety and local context keep the work accountable. The practical capability is human-machine systems reasoning. Keep the question, method, data, assumptions, uncertainty, review and next decision in one dated evidence receipt. SMART, MIT, NRF and participating institutions control current programmes and released results; ethics, safety, clinical, data and product authorities control applicable studies and deployment.

Put this into practice by turning the idea into a small, reviewable action: Build a one-page evidence card for “Human–machine research includes institutions” with the question, source or system, method, observation, limitation and next check. Then ask what the resulting note or artefact actually demonstrates. A useful record names context, source or instruction, action, observation, interpretation and uncertainty. It also preserves this limit: Technical accuracy alone does not establish beneficial adoption.

An evidence check keeps confidence proportional to what the work can show. Separate the original observation or official statement from the learner’s explanation. Record feedback and any revision. If information is missing, write “not shown” or “not yet known” rather than guessing. That honest gap can become the next question.

Contents · Previous chapter · Next chapter

Nanostructured optical surfaces, emitters, detectors and semiconductor electronics can be combined for compact sensing. Fabrication tolerances and calibration connect device physics to usable 3D measurements.

Interdisciplinary research succeeds when each field keeps its standards while making interfaces explicit. Shared words, data formats and decision rights matter as much as exciting ideas. The practical capability is integrated-sensing literacy. Keep the question, method, data, assumptions, uncertainty, review and next decision in one dated evidence receipt. SMART, MIT, NRF and participating institutions control current programmes and released results; ethics, safety, clinical, data and product authorities control applicable studies and deployment.

Put this into practice by turning the idea into a small, reviewable action: Build a one-page evidence card for “Metasurface sensing integrates optics and electronics” with the question, source or system, method, observation, limitation and next check. Then ask what the resulting note or artefact actually demonstrates. A useful record names context, source or instruction, action, observation, interpretation and uncertainty. It also preserves this limit: A simulated optical response is not a manufactured sensor.

An evidence check keeps confidence proportional to what the work can show. Separate the original observation or official statement from the learner’s explanation. Record feedback and any revision. If information is missing, write “not shown” or “not yet known” rather than guessing. That honest gap can become the next question.

Contents · Previous chapter · Next chapter

Soft materials, transducers, electronics and algorithms may support extended imaging outside conventional sessions. Motion, contact, power, comfort and interpretation create new validation needs.

A research group can connect biology, engineering, computation and social systems, yet every claim still needs the evidence appropriate to its own scale. The practical capability is wearable-imaging reasoning. Keep the question, method, data, assumptions, uncertainty, review and next decision in one dated evidence receipt. SMART, MIT, NRF and participating institutions control current programmes and released results; ethics, safety, clinical, data and product authorities control applicable studies and deployment.

Put this into practice by turning the idea into a small, reviewable action: Build a one-page evidence card for “Wearable ultrasound changes the imaging context” with the question, source or system, method, observation, limitation and next check. Then ask what the resulting note or artefact actually demonstrates. A useful record names context, source or instruction, action, observation, interpretation and uncertainty. It also preserves this limit: A wearable research platform is not a clinical diagnosis.

An evidence check keeps confidence proportional to what the work can show. Separate the original observation or official statement from the learner’s explanation. Record feedback and any revision. If information is missing, write “not shown” or “not yet known” rather than guessing. That honest gap can become the next question.

Contents · Previous chapter · Next chapter

A prototype shows that selected functions can work under stated conditions. It should expose assumptions, failure modes and the next performance threshold rather than presenting completion.

Translation is a sequence of bounded tests. A prototype, a pilot, a validated process and an adopted solution answer different questions. The practical capability is prototype-evaluation literacy. Keep the question, method, data, assumptions, uncertainty, review and next decision in one dated evidence receipt. SMART, MIT, NRF and participating institutions control current programmes and released results; ethics, safety, clinical, data and product authorities control applicable studies and deployment.

Put this into practice by turning the idea into a small, reviewable action: Build a one-page evidence card for “A prototype answers a feasibility question” with the question, source or system, method, observation, limitation and next check. Then ask what the resulting note or artefact actually demonstrates. A useful record names context, source or instruction, action, observation, interpretation and uncertainty. It also preserves this limit: Feasibility is distinct from reliability and adoption.

An evidence check keeps confidence proportional to what the work can show. Separate the original observation or official statement from the learner’s explanation. Record feedback and any revision. If information is missing, write “not shown” or “not yet known” rather than guessing. That honest gap can become the next question.

Contents · Previous chapter · Next chapter

Teams choose comparators, datasets and metrics based on the use case. A better average score may hide failures in important subgroups or operating conditions.

International collaboration adds expertise and comparison, while provenance, authorship, safety and local context keep the work accountable. The practical capability is benchmark-design reasoning. Keep the question, method, data, assumptions, uncertainty, review and next decision in one dated evidence receipt. SMART, MIT, NRF and participating institutions control current programmes and released results; ethics, safety, clinical, data and product authorities control applicable studies and deployment.

Put this into practice by turning the idea into a small, reviewable action: Build a one-page evidence card for “Benchmarks should match the decision” with the question, source or system, method, observation, limitation and next check. Then ask what the resulting note or artefact actually demonstrates. A useful record names context, source or instruction, action, observation, interpretation and uncertainty. It also preserves this limit: One metric cannot represent every user need.

An evidence check keeps confidence proportional to what the work can show. Separate the original observation or official statement from the learner’s explanation. Record feedback and any revision. If information is missing, write “not shown” or “not yet known” rather than guessing. That honest gap can become the next question.

Contents · Previous chapter · Next chapter

Consent, access, licensing, privacy, retention and provenance should survive data movement between institutions. Shared repositories need clear versions and responsibilities.

Interdisciplinary research succeeds when each field keeps its standards while making interfaces explicit. Shared words, data formats and decision rights matter as much as exciting ideas. The practical capability is collaborative-data stewardship. Keep the question, method, data, assumptions, uncertainty, review and next decision in one dated evidence receipt. SMART, MIT, NRF and participating institutions control current programmes and released results; ethics, safety, clinical, data and product authorities control applicable studies and deployment.

Put this into practice by turning the idea into a small, reviewable action: Build a one-page evidence card for “Data governance travels with collaboration” with the question, source or system, method, observation, limitation and next check. Then ask what the resulting note or artefact actually demonstrates. A useful record names context, source or instruction, action, observation, interpretation and uncertainty. It also preserves this limit: Research access does not imply unlimited reuse.

An evidence check keeps confidence proportional to what the work can show. Separate the original observation or official statement from the learner’s explanation. Record feedback and any revision. If information is missing, write “not shown” or “not yet known” rather than guessing. That honest gap can become the next question.

Contents · Previous chapter · Next chapter

Biological, clinical, cyber, workplace and social risks can interact when systems are combined. Review and mitigation should be planned at the relevant institutional level.

A research group can connect biology, engineering, computation and social systems, yet every claim still needs the evidence appropriate to its own scale. The practical capability is integrated-risk reasoning. Keep the question, method, data, assumptions, uncertainty, review and next decision in one dated evidence receipt. SMART, MIT, NRF and participating institutions control current programmes and released results; ethics, safety, clinical, data and product authorities control applicable studies and deployment.

Put this into practice by turning the idea into a small, reviewable action: Build a one-page evidence card for “Safety and ethics begin before integration” with the question, source or system, method, observation, limitation and next check. Then ask what the resulting note or artefact actually demonstrates. A useful record names context, source or instruction, action, observation, interpretation and uncertainty. It also preserves this limit: Interdisciplinary novelty does not bypass oversight.

An evidence check keeps confidence proportional to what the work can show. Separate the original observation or official statement from the learner’s explanation. Record feedback and any revision. If information is missing, write “not shown” or “not yet known” rather than guessing. That honest gap can become the next question.

Contents · Previous chapter · Next chapter

Authorship, intellectual property, materials, code and decisions should be attributed precisely. A contribution statement shows how the result depends on the team.

Translation is a sequence of bounded tests. A prototype, a pilot, a validated process and an adopted solution answer different questions. The practical capability is research-attribution literacy. Keep the question, method, data, assumptions, uncertainty, review and next decision in one dated evidence receipt. SMART, MIT, NRF and participating institutions control current programmes and released results; ethics, safety, clinical, data and product authorities control applicable studies and deployment.

Put this into practice by turning the idea into a small, reviewable action: Build a one-page evidence card for “Contribution records protect trust” with the question, source or system, method, observation, limitation and next check. Then ask what the resulting note or artefact actually demonstrates. A useful record names context, source or instruction, action, observation, interpretation and uncertainty. It also preserves this limit: Collaboration does not erase individual responsibility.

An evidence check keeps confidence proportional to what the work can show. Separate the original observation or official statement from the learner’s explanation. Record feedback and any revision. If information is missing, write “not shown” or “not yet known” rather than guessing. That honest gap can become the next question.

Contents · Previous chapter · Next chapter

Translation teams test who has the problem, what current alternatives achieve and whether the new approach improves a meaningful outcome. Manufacturing and regulatory paths shape design early.

International collaboration adds expertise and comparison, while provenance, authorship, safety and local context keep the work accountable. The practical capability is responsible-innovation literacy. Keep the question, method, data, assumptions, uncertainty, review and next decision in one dated evidence receipt. SMART, MIT, NRF and participating institutions control current programmes and released results; ethics, safety, clinical, data and product authorities control applicable studies and deployment.

Put this into practice by turning the idea into a small, reviewable action: Build a one-page evidence card for “Innovation needs evidence of a real need” with the question, source or system, method, observation, limitation and next check. Then ask what the resulting note or artefact actually demonstrates. A useful record names context, source or instruction, action, observation, interpretation and uncertainty. It also preserves this limit: A market story cannot substitute for technical validation.

An evidence check keeps confidence proportional to what the work can show. Separate the original observation or official statement from the learner’s explanation. Record feedback and any revision. If information is missing, write “not shown” or “not yet known” rather than guessing. That honest gap can become the next question.

Contents · Previous chapter · Next chapter

Archive the mission, subquestions, disciplines, interfaces, protocols, raw data, code, controls, validation, contribution, ethics, translation stage, uncertainty and next decision.

Interdisciplinary research succeeds when each field keeps its standards while making interfaces explicit. Shared words, data formats and decision rights matter as much as exciting ideas. The practical capability is interdisciplinary-research career preparation. Keep the question, method, data, assumptions, uncertainty, review and next decision in one dated evidence receipt. SMART, MIT, NRF and participating institutions control current programmes and released results; ethics, safety, clinical, data and product authorities control applicable studies and deployment.

Put this into practice by turning the idea into a small, reviewable action: Build a one-page evidence card for “Keep an interdisciplinary research receipt” with the question, source or system, method, observation, limitation and next check. Then ask what the resulting note or artefact actually demonstrates. A useful record names context, source or instruction, action, observation, interpretation and uncertainty. It also preserves this limit: Only responsible institutions and authorities can approve studies, clinical use or deployment.

An evidence check keeps confidence proportional to what the work can show. Separate the original observation or official statement from the learner’s explanation. Record feedback and any revision. If information is missing, write “not shown” or “not yet known” rather than guessing. That honest gap can become the next question.

Contents · Previous chapter