Research Ethics, Safety, Sustainability, and Governance

Introduction

Mechatronics research combines mechanical systems, electronics, control engineering, software, sensors, robotics, artificial intelligence, and increasingly connected digital infrastructure. This combination creates substantial benefits, but it also creates distinctive responsibilities. A prototype robot may injure a person; a poorly secured controller may become an entry point into an industrial network; an algorithm may discriminate against users; and an apparently harmless experiment may generate hazardous waste or consume excessive energy.

Research ethics and governance provide the principles and processes through which research is conducted responsibly. They do not merely restrict innovation. They improve the reliability, legitimacy, safety, and social value of engineering knowledge. Ethical research protects participants, researchers, communities, animals, data, the environment, and the reputation of institutions. It also helps ensure that published findings can be trusted and reproduced.

In this chapter, ethics is treated as an engineering responsibility. Decisions about experimental design, data collection, code, authorship, safety, procurement, and deployment are all part of responsible research practice.

Learning Objectives

By the end of this chapter, students should be able to:

1. Explain research integrity and distinguish honest error from misconduct.
2. Apply principles of responsible authorship, citation, data management, and reproducibility.
3. Identify ethical, safety, privacy, cybersecurity, and environmental risks in mechatronics projects.
4. Design proportionate controls for laboratory, robotic, human-participant, and animal research.
5. Explain responsible AI, data governance, regulatory compliance, and institutional oversight.
6. Evaluate a research project using an integrated ethics-and-risk framework.

1. Research Integrity and the Social Contract of Engineering

Research integrity is the commitment to honesty, accuracy, transparency, accountability, and respect throughout the research lifecycle. It applies to planning, funding, experimentation, analysis, publication, supervision, and dissemination.

Three serious forms of research misconduct are commonly recognized:

• Fabrication: inventing data, observations, participants, experiments, or results.
• Falsification: manipulating materials, equipment, processes, images, code, or data so that the research record is misrepresented.
• Plagiarism: presenting another person’s words, ideas, designs, code, or results without appropriate acknowledgment.

These categories do not include every irresponsible practice. Questionable research practices include selective reporting, inappropriate exclusion of inconvenient measurements, p-hacking, undisclosed conflicts of interest, duplicate publication, inadequate supervision, and failing to preserve research records. Honest mistakes are not automatically misconduct, but they must be corrected promptly when discovered.

A sound research record should allow another competent researcher to understand what was intended, what was done, what changed, and how conclusions were reached. Laboratory notebooks, version-controlled code, calibration records, raw data, analysis scripts, photographs, and deviation reports are therefore evidence of both scientific quality and ethical conduct.

Conflicts of Interest

A conflict of interest exists when personal, financial, professional, or institutional interests could improperly influence research judgment. A conflict is not necessarily wrongdoing. It becomes a governance problem when it is concealed or unmanaged. Examples include evaluating a component manufactured by a company in which the researcher owns shares, accepting sponsorship with publication restrictions, or supervising a student whose performance affects a personal relationship.

Appropriate responses include disclosure, independent review, separation of duties, data-access restrictions, or recusal from a decision. Researchers should distinguish a sponsor’s legitimate interest in confidentiality or intellectual property from an improper demand to suppress unfavorable results.

2. Authorship, Contribution, and Responsible Publication

Authorship should reflect substantial intellectual or practical contribution and responsibility for the published work. In a mechatronics project, contributions may include conceptual design, mechanical design, circuit development, firmware, experimental work, statistical analysis, safety review, or writing. Supplying funding, laboratory space, routine supervision, or administrative support alone does not normally justify authorship.

Before a project becomes difficult to divide, the team should discuss:

• expected contributions;
• provisional author order and its possible revision;
• responsibility for data, code, and documentation;
• intellectual-property and confidentiality constraints;
• procedures for resolving disputes.

Author order has different conventions across disciplines, so the team should follow the target journal or conference policy. Every listed author should approve the final manuscript and be able to identify the parts for which they are accountable. Guest authorship gives credit without meaningful contribution; ghost authorship omits a substantial contributor. Both distort the research record.

Good publication practice includes accurate citation, disclosure of limitations, correction of errors, acknowledgment of technical assistance, and avoidance of duplicate publication. Images and plots should not be altered in ways that conceal relevant evidence. If an image is enhanced for readability, the procedure should be documented and applied consistently.

3. Data Governance, Privacy, and Reproducibility

Data governance is the framework for deciding how data are collected, classified, accessed, stored, shared, retained, and destroyed. A practical data-management plan should answer six questions:

1. What data will be collected, and for what purpose?
2. What legal, ethical, contractual, or safety restrictions apply?
3. Who may access the data, and under what controls?
4. How will data quality, provenance, and integrity be maintained?
5. How will data be preserved and shared after the project?
6. When and how will data be deleted or securely destroyed?

Privacy and Personal Data

A robot’s camera, microphone, wearable sensor, access badge, or location tracker may collect personal data even when the research objective is technical. Privacy protection should follow principles such as purpose limitation, data minimization, transparency, accuracy, security, retention limitation, and accountability. Researchers should collect only data necessary for the stated purpose and should not assume that publicly observable behavior is ethically unrestricted for recording and analysis.

Useful safeguards include informed consent, pseudonymization, anonymization where feasible, encryption, role-based access, audit logs, retention schedules, and secure deletion. Pseudonymization replaces direct identifiers with codes but does not make data anonymous if a re-identification key exists. Anonymization aims to prevent identification, although absolute anonymity can be difficult when datasets contain unique movement, biometric, or location patterns.

Consent must be understandable, voluntary, specific to the intended use, and capable of withdrawal where applicable. Consent is not a substitute for good security or proportional data collection. Researchers must also consider bystanders, who may be recorded by an autonomous system without actively agreeing to participate.

Data Quality and Provenance

Each dataset should have a clear provenance: its source, collection conditions, sensor configuration, units, calibration status, preprocessing steps, exclusions, and transformations. Sensor data should retain timestamps, sampling rates, coordinate conventions, and uncertainty information. A missing-value code must not be confused with a measured zero.

Data integrity can be supported by checksums, read-only raw-data storage, access logs, version control, and documented analysis pipelines. Raw data should normally be preserved separately from processed data. If records are corrected, the original should remain recoverable and the correction documented.

Reproducibility

A result is reproducible when the same data, procedures, software, and conditions produce substantially the same result. Replicability is often used for obtaining consistent findings through a new study or independent implementation. Reproducibility is strengthened by:

• preregistering hypotheses or analysis plans when appropriate;
• reporting failed trials and deviations;
• sharing code, parameter settings, hardware specifications, and synthetic or permitted datasets;
• recording software and library versions;
• using automated tests and calibration procedures;
• distinguishing exploratory from confirmatory analysis.

Reproducibility does not require releasing confidential, unsafe, proprietary, or personally identifiable material. In such cases, researchers can provide controlled access, a data dictionary, simulated data, executable workflows, or a detailed protocol.

4. Human-Participant and Animal Research

Human-participant research includes more than medical experimentation. It may include usability tests, operator studies, workplace observation, wearable sensing, interviews, video recording, or experiments involving physical interaction with a robot. The central principles are respect for persons, beneficence, non-maleficence, justice, and informed consent.

An ethics review committee or institutional review board should assess risks before recruitment begins. The review should consider physical injury, psychological discomfort, privacy, confidentiality, coercion, unequal power relationships, compensation, accessibility, and whether vulnerable groups are involved. Students or employees may feel pressured to participate when the researcher is also their teacher or manager; independent recruitment and consent procedures can reduce this risk.

A valid protocol explains the purpose, procedures, foreseeable risks, benefits, data use, withdrawal process, and contact details for concerns. Participants should not be exposed to avoidable risk merely because a safer method is less convenient. In human–robot experiments, an emergency stop, trained operator, speed and force limits, protective barriers, and a clear abort procedure are essential.

Animal research requires additional justification, specialist oversight, veterinary involvement, humane handling, and compliance with applicable law. The principles of the 3Rs are:

• Replacement: use non-animal alternatives where scientifically possible.
• Reduction: use the minimum number of animals needed for valid evidence.
• Refinement: minimize pain, distress, and lasting harm.

These principles apply particularly to robotic systems that interact with, monitor, restrain, or assist animals.

5. Laboratory, Robotic, and Embedded-System Safety

Safety is a design property, not merely a warning label. A suitable process begins with hazard identification, risk estimation, control selection, verification, and continuous review.

A simple qualitative risk-priority model is:

where R is a risk score, S is the severity of harm, and L is the likelihood of occurrence. The scales must be defined by the institution, such as 1–5 for each factor. The product is a screening aid, not a physical law; it should not be used to hide uncertainty or to compare unrelated hazards without a common scale. A low-probability catastrophic event may require control even if its numerical score is moderate.

The hierarchy of controls ranks interventions from strongest to weakest:

1. Elimination of the hazard.
2. Substitution with a less hazardous material, process, or motion.
3. Engineering controls such as guards, interlocks, current limiting, isolation, and emergency stops.
4. Administrative controls such as training, procedures, permits, and supervision.
5. Personal protective equipment.

For a mobile or industrial robot, a safety case should address unexpected motion, loss of communication, sensor failure, software faults, power restoration, entanglement, crushing, sharp edges, stored energy, electromagnetic interference, and unauthorized commands. Safety functions should fail to a safe state where feasible. An emergency stop should remove or control hazardous energy, but it is not a substitute for guarding or risk assessment.

Electrical work requires isolation, verification of de-energization, suitable fusing, grounding, insulation, and protection from exposed conductors. Batteries require controls for short circuits, overcharge, thermal runaway, mechanical damage, and inappropriate disposal. Pneumatic and hydraulic systems require attention to stored pressure, hose failure, unexpected movement, and safe depressurization.

Software safety includes input validation, watchdog timers, bounded operating modes, fault detection, logging, safe defaults, and testing under abnormal conditions. A prototype must not be treated as safe merely because it worked during a demonstration.

6. Cybersecurity as a Research-Ethics Obligation

Connected mechatronic systems can affect physical processes, so cybersecurity failures may cause bodily, environmental, or economic harm. Responsible security research requires authorization, minimum necessary access, controlled testing, and coordinated disclosure of vulnerabilities.

A basic threat analysis considers assets, adversaries, attack surfaces, vulnerabilities, impacts, and controls. Assets may include firmware, calibration values, safety parameters, personal data, intellectual property, and physical actuators. Controls include secure authentication, least privilege, network segmentation, encrypted communication, signed firmware, patch management, secure boot, backup and recovery, logging, and incident response.

Researchers should never scan or exploit systems without permission. Demonstrations should use an isolated laboratory network or a purpose-built testbed. If a vulnerability is found in a third-party product, the researcher should preserve evidence, avoid unnecessary disclosure, notify the responsible party through an appropriate channel, allow reasonable time for mitigation, and coordinate public communication. Safety-critical details should not be published in a form that enables immediate harm without corresponding defensive information.

7. Environmental Sustainability and Life-Cycle Thinking

Sustainability involves more than reducing electricity during an experiment. A life-cycle perspective considers extraction of materials, manufacturing, transport, operation, maintenance, upgrade, reuse, and end-of-life disposal.

Researchers should assess:

• energy consumption and carbon implications;
• material criticality and supply-chain conditions;
• toxic substances and laboratory waste;
• battery and electronic-waste management;
• repairability, modularity, and reuse;
• noise, vibration, emissions, and ecological disturbance;
• whether the claimed benefit justifies environmental cost.

A useful operational measure is energy consumed during an experiment:

where E is energy in joules, P is power in watts, and t is time in seconds, assuming approximately constant power. If power varies with time, the general expression is:

where P(t) is instantaneous power and T is the total operating time. Unit consistency is essential: watt-seconds equal joules. For example, a 120 W test running for 30 minutes consumes E = 120 × 1800 = 216000 J, or 216 kJ, equivalent to 0.06 kWh. This calculation excludes embodied energy unless that is separately assessed.

Sustainable design may involve lower-power computation, duty cycling, efficient motion planning, longer component life, standardized interfaces, repairable enclosures, recyclable materials, and transparent environmental reporting. Sustainability claims should identify the boundary of analysis; claiming that a device is “green” from operating energy alone is misleading if manufacture and disposal dominate its impact.

8. Responsible Artificial Intelligence in Mechatronics

AI systems in mechatronics may classify objects, estimate human intent, detect faults, optimize motion, or make decisions affecting safety and access. Responsible AI requires attention to validity, fairness, transparency, privacy, security, human oversight, and accountability.

Important questions include:

• Is the training and test data representative of operating conditions?
• Are labels reliable, and are errors distributed unevenly across groups or environments?
• Does performance degrade under lighting, temperature, sensor drift, adversarial input, or unusual objects?
• Can an operator understand the system’s limits and override it?
• Is the model being used outside the conditions for which it was validated?

Accuracy alone is insufficient. A model may achieve high average accuracy while failing dangerously for a minority condition. Researchers should report confusion matrices or other suitable performance measures, confidence limitations, false-positive and false-negative consequences, and out-of-distribution behavior. Data and models should be versioned, and changes should trigger regression testing and renewed risk assessment.

Human oversight must be meaningful. An operator cannot provide effective oversight if the system is too fast to interrupt, hides uncertainty, or creates automation bias. For safety-critical action, a conservative fallback mode and independent safety mechanism are preferable to relying solely on a learned model.

Generative AI also creates research risks, including fabricated citations, leakage of confidential data, unclear ownership of generated material, and unverified code. Researchers remain responsible for checking AI-assisted outputs, documenting material use where required, and ensuring that confidential or personal information is not entered into an unsuitable service.

9. Regulatory Compliance and Institutional Governance

Compliance requirements depend on jurisdiction and application. Relevant domains may include occupational health and safety, electrical and machinery safety, radio and telecommunications regulation, privacy and data protection, export controls, environmental law, intellectual property, medical-device regulation, and animal or human research law. Standards may provide technical consensus, but a standard is not automatically a legal approval, and compliance with one standard does not eliminate the need for risk assessment.

Governance allocates authority and accountability. A mature research organization normally defines:

• who approves protocols and risk assessments;
• who controls access to laboratories and data;
• who may modify safety-critical software;
• who reports incidents and near misses;
• who verifies corrective actions;
• who can stop a project when unacceptable risk emerges.

A useful governance structure separates development from independent review where practical. It also maintains traceability from research objectives to requirements, tests, approvals, changes, incidents, and final claims. Documentation should be proportionate to risk but sufficient for later investigation.

10. An Integrated Ethics-and-Risk Workflow

Ethical review should begin at project conception rather than at publication. The following workflow integrates technical and ethical judgment:

1. Define purpose and stakeholders. Identify intended benefits, affected people, communities, operators, maintainers, and the environment.
2. Map the system boundary. Include hardware, software, data flows, supply chains, operators, bystanders, and end-of-life processes.
3. Identify hazards and ethical concerns. Consider physical, digital, social, legal, privacy, environmental, and epistemic risks.
4. Assess proportionality. Ask whether the expected knowledge or benefit justifies residual risk.
5. Select controls. Prefer elimination and engineering controls over warnings and training alone.
6. Obtain approvals. Secure institutional, safety, human-participant, animal, data, or regulatory approval before relevant work begins.
7. Test and document. Test normal, abnormal, boundary, failure, and recovery conditions; preserve evidence.
8. Monitor and report. Record incidents, near misses, deviations, complaints, and unexpected outcomes.
9. Review before deployment and publication. Confirm that conclusions match evidence and that disclosure does not create unreasonable harm.
10. Learn and improve. Update procedures, designs, training, and governance after experience.

Illustrative Risk Register

11. Case Study: Autonomous Warehouse Robot Prototype

This is an illustrative case based on a hypothetical university research project.

Background

A student team develops a small autonomous mobile robot to transport components between laboratory workstations. The robot uses cameras, ultrasonic sensors, a wireless controller, and a machine-learning model for obstacle classification. The team plans to test it in a shared workshop.

Problem

The initial demonstration succeeds, but the project proposal does not address bystanders, wireless security, battery failure, data retention, or the possibility that the classifier performs poorly under low light. A sponsor requests raw video and asks to approve any publication before release.

Analysis

The system presents interconnected risks. A collision could injure a person; wireless compromise could cause unauthorized motion; video could identify students; poor low-light performance could create unsafe behavior; and unrestricted sponsor control could undermine research independence. The team also discovers that several video clips were copied from an online source without clear licensing or documentation.

Approach and Solution

The supervisor requires a revised protocol. Testing is moved to a controlled area with physical barriers and a trained safety operator. A hardware emergency stop, conservative speed limit, collision-monitoring layer, and independent power isolation are installed. The robot is tested with loss of wireless communication, sensor obstruction, low battery, controller restart, and unexpected obstacles.

Video collection is minimized. The team uses a defined retention period, access control, masking of faces, and clear notices. Where possible, development uses synthetic or staged scenes. The wireless network is isolated, credentials are unique, and command authorization is implemented. The machine-learning model is evaluated under varied lighting and with separate test data. The team replaces the questionable online clips or obtains permission and records provenance.

The sponsor may review confidential intellectual property for a limited period, but cannot veto accurate reporting of results. Authorship and contribution records are agreed before the final paper. Battery charging and electronic-waste procedures are added to the laboratory risk assessment.

Outcome

The revised testing reveals that the classifier has a substantially higher false-negative rate in low light. The team disables autonomous operation below a defined illumination condition and reports the limitation. No participant data are retained beyond the approved period. The project takes longer, but the final prototype is safer, more reproducible, and more credible.

Learning Points

1. Ethical risks often arise from interactions among technical subsystems.
2. A successful demonstration is not evidence of safe deployment.
3. Privacy, cybersecurity, and sustainability should be designed into the system.
4. Independent review protects both participants and researchers.
5. Transparent limitations strengthen, rather than weaken, a research claim.

Challenges and Limitations

Ethical decision-making rarely produces a single numerical answer. Risk estimates may be uncertain, stakeholders may disagree, and legal requirements vary by jurisdiction. Privacy protection can reduce data utility; open science can conflict with confidentiality or cybersecurity; safety controls can reduce speed or experimental flexibility; and sustainability improvements may increase initial cost.

Risk matrices are useful communication tools but can create false precision. Compliance checklists may encourage superficial box-ticking if researchers do not understand the underlying hazards. Institutional review cannot replace professional judgment, and an approved protocol may become unsafe when equipment, personnel, software, or context changes. Continuous review is therefore essential.

Best-Practice Checklist

Before beginning a mechatronics research project, confirm that:

• the purpose, beneficiaries, and affected stakeholders are identified;
• contributions, authorship, conflicts, and intellectual property are discussed;
• hazards, privacy concerns, cyber threats, and environmental impacts are assessed;
• required ethics, safety, data, and regulatory approvals are obtained;
• raw data, code, hardware versions, and deviations are documented;
• human and animal welfare protections are proportionate and effective;
• safety-critical functions are independently tested;
• data access, retention, sharing, and deletion are controlled;
• AI limitations, bias, uncertainty, and human oversight are evaluated;
• incidents and near misses are reported and used for improvement;
• conclusions and publications accurately represent the evidence.

Chapter Summary

Research ethics, safety, sustainability, and governance are integrated dimensions of professional mechatronics practice. Research integrity requires honest records, accurate reporting, responsible authorship, disclosure of conflicts, and correction of errors. Data governance protects privacy and preserves provenance, security, quality, and reproducibility. Human and animal research requires justified procedures, informed consent or specialist oversight, and minimization of harm.

Laboratory and robotic safety depend on hazard identification, the hierarchy of controls, engineered safeguards, failure testing, and competent supervision. Cybersecurity is an ethical obligation because connected devices can produce physical consequences. Sustainability requires life-cycle thinking about energy, materials, waste, repair, and disposal. Responsible AI requires valid data, performance analysis beyond average accuracy, transparency about limitations, security, and meaningful human oversight.

Effective governance assigns authority, documents decisions, obtains appropriate approvals, and supports incident reporting and continuous improvement. The central professional principle is that innovation should be beneficial, defensible, safe, and accountable—not merely technically impressive.

Key Terms

Research integrity: Honest, transparent, accountable, and responsible conduct of research.

Fabrication: Inventing data, observations, or results.

Falsification: Manipulating research materials, processes, or records to misrepresent findings.

Plagiarism: Using another person’s work without appropriate acknowledgment.

Conflict of interest: An interest that could improperly influence professional judgment.

Authorship: Credit and responsibility assigned to substantial contributors to a scholarly work.

Data governance: Rules and processes for managing data throughout its lifecycle.

Pseudonymization: Replacing identifiers with codes while retaining a possible re-identification link.

Anonymization: Processing intended to prevent identification of individuals.

Data provenance: A documented history of data origin, processing, and transformation.

Reproducibility: Obtaining consistent results using the same data, procedures, and tools.

Informed consent: Voluntary agreement based on understandable information about participation.

Institutional review board: A body that reviews research involving human participants.

Hierarchy of controls: Ranking of hazard controls from elimination to personal protective equipment.

Risk assessment: Systematic identification, evaluation, and control of hazards.

Emergency stop: A control intended to rapidly stop or control hazardous operation.

Threat model: Structured analysis of assets, adversaries, vulnerabilities, and impacts.

Responsible AI: Development and use of AI that addresses validity, fairness, transparency, safety, privacy, and accountability.

Life-cycle thinking: Evaluation of impacts from resource extraction through end-of-life.

Governance: Allocation of authority, accountability, decision rights, and oversight.

Review Questions

Short Answer Questions

1. Distinguish fabrication, falsification, and plagiarism.
2. Why is pseudonymization not equivalent to anonymization?
3. State the five levels in the hierarchy of controls.
4. What are the three principles represented by the 3Rs in animal research?
5. Why should raw data be preserved separately from processed data?

Long Answer Questions

1. Explain how research integrity, authorship, and reproducibility are connected in a mechatronics project.
2. Discuss the ethical and technical controls needed when a robot records video in a shared workspace.
3. Explain why cybersecurity should be treated as a safety and ethics issue in connected robotic systems.
4. Compare operational energy assessment with life-cycle sustainability assessment.
5. Describe an integrated governance workflow for a human–robot interaction study.

Analytical / Application Questions

1. A mobile robot occasionally continues moving after wireless communication is lost. Identify the hazards, propose controls using the hierarchy of controls, and specify tests that should precede human-accessible operation.
2. A machine-learning inspection system has 95% overall accuracy but performs poorly on products made under a different lighting condition. Analyze the ethical and technical implications and propose a validation plan.
3. Design a compact risk register for a wearable-sensor study involving students, including physical, privacy, cybersecurity, data-governance, and environmental concerns.

Multiple Choice Questions with Answers

Question 1

Which action is an example of falsification?

A. Inventing measurements that were never collected

B. Changing recorded values to hide inconvenient results

C. Using a colleague’s paragraph without citation

D. Making an honest calculation error and correcting it

Correct Answer: B. Changing recorded values to hide inconvenient results

Question 2

What is the main limitation of pseudonymization?

A. It prevents all data analysis

B. It always removes the need for consent

C. A person may still be re-identified if the key or linked information exists

D. It can only be used for numerical data

Correct Answer: C. A person may still be re-identified if the key or linked information exists

Question 3

Which is the strongest control in the hierarchy of controls?

A. Personal protective equipment

B. Warning signs

C. Operator training

D. Elimination of the hazard

Correct Answer: D. Elimination of the hazard

Question 4

A system consumes 120 W for 30 minutes. What energy does it use?

A. 0.006 kWh

B. 0.06 kWh

C. 0.6 kWh

D. 6 kWh

Correct Answer: B. 0.06 kWh

Question 5

Which principle requires using the minimum number of animals needed for valid evidence?

A. Replacement

B. Reduction

C. Refinement

D. Reciprocity

Correct Answer: B. Reduction

Question 6

Which practice best supports reproducibility?

A. Reporting only successful trials

B. Changing analysis settings without records

C. Preserving code, parameter settings, and software versions

D. Deleting raw data after producing a graph

Correct Answer: C. Preserving code, parameter settings, and software versions

Question 7

Why is a risk matrix not a physical law?

A. Risk cannot involve likelihood

B. Its score depends on chosen scales and does not replace judgment

C. Severity is never relevant

D. It can only be used for financial risks

Correct Answer: B. Its score depends on chosen scales and does not replace judgment

Question 8

Which is an appropriate cybersecurity practice for a robotic research testbed?

A. Testing exploits on any nearby network

B. Using shared default passwords

C. Isolating the testbed and controlling authorized access

D. Publishing credentials with the experimental procedure

Correct Answer: C. Isolating the testbed and controlling authorized access

Question 9

Which statement best describes responsible AI evaluation?

A. Average accuracy is always sufficient

B. Only training performance needs to be reported

C. The system should be tested under relevant conditions and its limitations reported

D. Human oversight is unnecessary when a model is accurate

Correct Answer: C. The system should be tested under relevant conditions and its limitations reported

Question 10

What is the primary purpose of a research governance structure?

A. To eliminate all experimental uncertainty

B. To allocate authority, accountability, approvals, and oversight

C. To guarantee that every project receives funding

D. To replace technical testing with paperwork

Correct Answer: B. To allocate authority, accountability, approvals, and oversight