1. Foundations of Research Ethics and Responsible Inquiry

Introduction: Why Research Ethics Is a Foundation of Doctoral Work

Research is a social practice as well as an intellectual activity. It produces knowledge, but it also affects people, institutions, communities, environments, public policy, and technological systems. A doctoral researcher therefore does more than design a technically sound study. The researcher makes judgments about what questions may be asked, whose interests are represented, what risks are acceptable, how evidence is handled, and how findings are communicated.

Research ethics provides the principles and practices through which those judgments are made responsibly. It is not limited to obtaining approval from an ethics committee or completing an online training module. Ethics begins before data collection and continues through analysis, publication, preservation, reuse, and correction of the scholarly record. In information technology, ethical responsibility is especially significant because research may involve personal data, algorithmic systems, cybersecurity vulnerabilities, artificial intelligence, online communities, or technologies capable of operating at large scale.

A technically impressive study can still be ethically defective. A dataset may be statistically valuable but obtained without meaningful consent. An algorithm may achieve high predictive accuracy while reproducing discriminatory patterns. A researcher may not fabricate results yet still mislead readers through selective reporting, undisclosed conflicts of interest, or inappropriate authorship. Responsible inquiry requires attention to both the integrity of the research process and the consequences of its products.

At doctoral level, research ethics must be understood as a form of disciplined reasoning. Rules and codes are essential, but they cannot anticipate every situation. Researchers must interpret principles in context, identify competing values, justify decisions, and remain accountable to affected parties and the wider research community.

Learning Objectives

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

1. Explain the relationship between research ethics, research integrity, and responsible conduct of research.

2. Distinguish research misconduct from questionable research practices and honest error.

3. Apply major ethical principles and ethical reasoning frameworks to information technology research.

4. Analyse the professional responsibilities of doctoral researchers, supervisors, institutions, and collaborators.

5. Identify ethical risks associated with data, software, artificial intelligence, cybersecurity, and digital research environments.

6. Evaluate an ethically complex research situation and formulate a defensible response.

7. Recognize current challenges involving reproducibility, open science, automated systems, and responsible innovation.

The Ethical Character of Inquiry

Research ethics, integrity, and responsible conduct

Research ethics concerns the moral principles governing research involving people, data, institutions, animals, technology, and the environment. It asks whether a research activity is permissible, fair, respectful, and appropriately justified. Research integrity refers more specifically to honesty, rigor, transparency, reliability, and accountability in proposing, conducting, reviewing, and reporting research. Responsible conduct of research combines these dimensions: it requires respect for persons and communities as well as trustworthy production and communication of knowledge.

The concepts overlap but are not identical. A study may have strong methodological integrity but violate privacy. Conversely, a study may have obtained consent but use weak methods that waste participants’ contributions and produce unreliable conclusions. Ethical research must address both the treatment of those affected by research and the quality of the knowledge claimed by the researcher.

Integrity is therefore not merely a personal virtue. It is also an institutional and professional condition. Supervisors, research teams, universities, funders, publishers, professional associations, and regulators all influence whether responsible conduct is possible. Incentive systems that reward publication quantity, rapid results, or commercial novelty without valuing careful verification can create pressure toward questionable practices.

The social contract of research

Research operates on a form of social contract. Participants often provide information, time, labor, or access because they trust that researchers will use these contributions responsibly. Public and private organizations provide funding, infrastructure, and access because they expect credible results. Readers and other researchers rely on published claims to make further decisions. When researchers deceive, conceal relevant limitations, or misuse data, they damage not only a particular project but also confidence in scholarship.

This social contract is particularly important in computing. Software libraries, datasets, benchmarks, and models are frequently reused beyond the original project. A coding decision made in a research laboratory may later influence a health service, hiring platform, educational system, or security tool. The potential scale and persistence of digital outputs expand the researcher’s duty to anticipate foreseeable misuse and downstream effects.

Principles for Ethical Reasoning

No single principle resolves every ethical problem. The following principles provide a foundation for analysis, but they may conflict and must be interpreted together.

Respect for persons and autonomy

Respect for persons recognizes individuals as capable of making decisions about their participation and personal information. In research involving human participants, this normally requires informed consent. Consent should be based on understandable information about the purpose of the study, procedures, risks, benefits, data use, confidentiality, voluntary nature of participation, and withdrawal options.

Consent is not simply a signature. It is a continuing process of communication. In digital research, researchers must consider whether a person who posts information publicly expects that information to be collected, linked, analysed, or reproduced in a scholarly publication. Public availability does not automatically eliminate ethical obligations. Special care is required when participants have limited autonomy, are vulnerable to coercion, or may experience employment, educational, legal, or social consequences from participation.

Beneficence and non-maleficence

Beneficence concerns the promotion of benefits, while non-maleficence concerns avoiding or minimizing harm. Harm may be physical, psychological, social, economic, reputational, legal, or informational. A cybersecurity study, for example, may generate valuable knowledge while exposing a vulnerable system or revealing a security weakness. Ethical design requires risk assessment, safeguards, responsible disclosure, and a plan for responding to unexpected harm.

Risk-benefit analysis should be realistic rather than formalistic. Researchers should identify who bears the risks, who receives the benefits, whether risks are necessary for the research question, and whether a less harmful method could produce comparable knowledge. A technically convenient method is not ethically justified merely because it is inexpensive or efficient.

Justice and fairness

Justice requires fair distribution of the burdens and benefits of research. Participant selection should not exploit groups that are accessible, powerless, or unlikely to challenge the researcher. Nor should potentially beneficial research systematically exclude populations whose data are difficult to obtain.

In information technology, justice also applies to datasets and automated systems. If a dataset underrepresents certain communities, a model trained on it may perform poorly for those groups. Fairness cannot be reduced to a single mathematical metric because different definitions may conflict. Researchers should examine the social context of classification, the consequences of errors, the distribution of decision-making power, and the possibility that the system itself should not be deployed.

Fidelity, honesty, and accountability

Researchers owe duties of fidelity to participants, collaborators, institutions, funders, and readers. Honesty requires accurate representation of methods, data, contributions, uncertainties, and results. Accountability means that researchers can explain and defend their decisions and can correct the record when errors are discovered.

Transparency is an important instrument of accountability. Where appropriate, researchers should document protocols, analytical decisions, code, data provenance, limitations, conflicts of interest, and deviations from preregistered plans. Transparency does not mean that every dataset must be publicly released. Privacy, security, contractual obligations, and participant consent may require controlled access or non-disclosure. Ethical transparency is therefore calibrated transparency: enough information to permit evaluation without creating unjustifiable risk.

Ethical Reasoning Beyond Rule Following

A practical reasoning sequence

Ethical decision-making is strengthened by a structured process. A researcher can begin by asking:

1. What is the decision or proposed action, and who may be affected?

2. What facts are known, uncertain, or missing?

3. Which principles, policies, laws, professional standards, and agreements apply?

4. What options are available, including redesign, delay, consultation, or refusal?

5. What benefits, harms, rights, and responsibilities attach to each option?

6. Who has decision-making power, and whose interests may be overlooked?

7. Can the decision be explained publicly and defended to an independent reviewer?

8. What monitoring, mitigation, documentation, and correction mechanisms are needed?

This sequence prevents a narrow focus on whether an action is technically possible. It also encourages consultation with supervisors, ethics committees, data protection officers, security specialists, community representatives, or legal advisers when appropriate.

Deontological, consequentialist, and virtue perspectives

Deontological reasoning evaluates duties and rights. It may conclude that deception, unauthorized disclosure, or fabrication is wrong even if the researcher expects a beneficial result. Consequentialist reasoning evaluates likely outcomes, comparing benefits and harms. It is useful for risk assessment but can become dangerous if it treats the interests of a minority as expendable. Virtue ethics focuses on the character and habits of the researcher, asking what a person committed to honesty, intellectual humility, courage, fairness, and care would do.

These perspectives are complementary. A responsible researcher may reject fabrication because it violates a duty, recognize that fabricated results can cause serious harm, and understand that honesty is part of scholarly character. In difficult cases, ethical justification is strongest when it identifies the relevant duties, evaluates consequences, and considers what professional virtues require.

Procedural and relational approaches

Procedural ethics emphasizes formal review, consent documentation, data management, and compliance. These processes are necessary for accountability, but they may not capture changing relationships in fieldwork or digital communities. Relational ethics emphasizes trust, context, power, vulnerability, and ongoing responsibility. For example, a researcher studying an online support group may meet formal requirements yet still harm the community by quoting identifiable posts without considering members’ expectations.

Doctoral researchers should use both approaches. Procedures establish minimum safeguards; relational judgment addresses situations in which the written rule does not fully express the ethical issue.

Research Misconduct and Questionable Research Practices

Fabrication, falsification, and plagiarism

Research misconduct is commonly discussed through three central forms:

• Fabrication is making up data, observations, participants, experiments, or results and reporting them as real.
• Falsification is manipulating research materials, equipment, processes, images, code, data, or results so that the research record is misrepresented.
• Plagiarism is presenting another person’s ideas, words, code, images, data, or creative work as one’s own without appropriate acknowledgment.

These acts are serious because they directly compromise the reliability of the scholarly record. Falsification in computing may include altering benchmark results, removing failed trials without disclosure, manipulating visualizations, or changing source code while claiming to have used a different version. Plagiarism includes unattributed copying of program code, dataset documentation, model architectures, or text. Intent and due process matter in institutional investigations, but researchers must not treat misconduct as a minor technical breach.

Honest error and responsible correction

Not every inaccurate result is misconduct. Research involves uncertainty, faulty instruments, programming bugs, misunderstood methods, and unexpected variation. An honest error becomes an ethical problem when the researcher refuses to correct it, conceals a material limitation, or knowingly continues to rely on a discredited result.

Responsible researchers preserve records, report deviations, respond to questions, issue corrections when necessary, and distinguish what is known from what is inferred. Retractions and corrections are mechanisms for protecting the research record, not necessarily evidence of personal failure.

Questionable research practices

Questionable research practices may not meet the formal threshold for misconduct but can still undermine knowledge. Examples include selective reporting of favorable outcomes, p-hacking, undisclosed changes to hypotheses, inappropriate exclusion of observations, honorary or ghost authorship, duplicate publication, inadequate record keeping, coercive citation, and failure to disclose conflicts of interest.

In computational research, additional concerns include data leakage between training and test sets, changing evaluation datasets until a desired result appears, reporting only the best random seed, comparing a new method with weak baselines, undocumented hyperparameter tuning, and releasing code that cannot reproduce the published claims. Each practice may appear small, but cumulative distortion can produce exaggerated conclusions.

The distinction between misconduct and questionable practice should not be used to excuse the latter. Ethical evaluation concerns the effect on trustworthiness, not only whether a formal label applies.

Doctoral Researchers as Responsible Professionals

Doctoral status confers authority and responsibility. A doctoral researcher is simultaneously a learner, investigator, collaborator, author, and emerging member of a profession. The researcher must develop independence while recognizing the limits of personal expertise.

Responsibilities to the research record

Doctoral researchers should maintain complete and dated records of research decisions, versions of software, data transformations, code dependencies, analytical scripts, approvals, consent materials, and deviations from plans. Reproducibility is not always fully achievable, particularly where data are confidential or systems are proprietary, but researchers should make the work as auditable as circumstances permit.

Researchers should also report negative, null, and ambiguous findings when they are relevant. A dissertation is not a marketing document. Its value lies in the quality and honesty of its reasoning, including recognition of limitations.

Responsibilities to participants and communities

Researchers must avoid treating participants as sources of data detached from their social identities. They should use accessible information sheets, minimize collection, protect confidentiality, and establish realistic expectations about future use. Community engagement may be necessary when research concerns collective identities, local knowledge, or populations historically subject to extractive research.

In digital studies, de-identification is not a guarantee of anonymity. Combining seemingly harmless attributes can permit re-identification. Researchers should assess linkage risks, limit access, remove unnecessary identifiers, and define retention and deletion policies.

Responsibilities within research teams

Authorship should reflect substantial intellectual or scholarly contribution and should be discussed early. Supervisors must not appropriate students’ work, and students should not claim contributions they did not make. Teams should agree on data ownership, publication procedures, intellectual property, code maintenance, dispute resolution, and responsibility for responding to errors.

A power imbalance can make it difficult for students or junior staff to challenge questionable decisions. Institutions should provide confidential reporting routes and protect good-faith whistleblowers from retaliation. Researchers also have a duty to raise concerns when silence would permit serious harm or deception.

Conflicts of interest

A conflict of interest exists when personal, financial, institutional, professional, or relational interests could reasonably be perceived to influence research judgment. A conflict is not automatically wrongdoing. The ethical requirement is disclosure and management. Possible controls include independent analysis, separation of roles, recusal, monitoring, preregistration, and transparent reporting.

Information Technology Research: Distinctive Ethical Questions

Data governance and privacy

IT research commonly involves logs, location records, social media content, biometric information, educational records, health information, or inferred attributes. Ethical data governance requires attention to purpose limitation, data minimization, access control, retention, provenance, and lawful as well as ethically legitimate use. Researchers should distinguish data that are technically accessible from data whose use is reasonably expected by the people concerned.

Anonymization, pseudonymization, encryption, and aggregation reduce risk but do not eliminate it. Security controls must address both accidental disclosure and insider misuse. Data management plans should specify who can access data, how access is logged, how backups are protected, and what happens at the end of the project.

Artificial intelligence and algorithmic research

AI research introduces concerns about training-data provenance, consent, copyright, representational harms, bias, explainability, environmental cost, and dual use. A model may be reproducible yet ethically unacceptable if it systematically harms a vulnerable group. Researchers should report dataset composition, known limitations, evaluation populations, uncertainty, and intended and prohibited uses.

Generative AI also affects the research process. It may assist with coding, translation, literature organization, or drafting, but researchers remain responsible for accuracy, attribution, confidentiality, and intellectual contribution. Sensitive unpublished manuscripts, identifiable data, credentials, and proprietary code should not be entered into external systems without authorization. AI-assisted outputs require verification because fluent language is not evidence of correctness.

Cybersecurity and dual-use research

Security research often involves a tension between discovering vulnerabilities and preventing exploitation. Ethical practice includes authorization, controlled testing, data minimization, secure storage of exploit details, coordination with affected parties, and responsible disclosure. Researchers should define scope before testing and avoid scanning or attacking systems without permission.

Dual-use research can support beneficial and harmful applications. The possibility of misuse does not automatically prohibit research, but it requires proportional risk assessment, access controls, staged release, monitoring, and careful communication. Claims about capabilities should be precise rather than sensational.

Open science and reproducibility

Open methods, code, materials, and data can improve scrutiny and accelerate discovery. However, openness must be balanced against privacy, security, contractual restrictions, and the rights of communities. A responsible open-science plan may use synthetic data, de-identified samples, secure enclaves, controlled repositories, executable documentation, or access committees rather than unrestricted release.

Reproducibility also depends on software environments, hardware, random seeds, dependency versions, and adequate documentation. Releasing code without explaining how data were cleaned or models were selected does not by itself make a study reproducible.

Applied Case: The Campus Safety Prediction Project

Background

The following is an illustrative case. A doctoral researcher, Mira, is developing a machine-learning system intended to predict locations where campus security incidents may occur. The university provides historical incident reports, emergency-call metadata, building access logs, and anonymized student movement data. A private security company funds part of the project and hopes to commercialize the system.

Situation and problem

The initial model shows high overall accuracy. However, the data contain far more reports from campuses and neighborhoods with heavier security presence. The model therefore recommends increased patrols in those areas. Mira discovers that the movement data were collected under a broad institutional privacy notice that did not clearly explain research reuse. The company asks her to omit subgroup performance results because they may delay deployment. Her supervisor argues that the data are already anonymized and that the project has received institutional approval.

Analysis

Several ethical issues interact. First, the model may learn patterns of surveillance and reporting rather than underlying safety risks. Increased patrols could generate more recorded incidents in the same areas, reinforcing the model’s assumptions. Second, high aggregate accuracy may conceal unequal error rates. Third, the original notice may not constitute meaningful consent for the new research purpose. Fourth, commercial funding creates a potential conflict of interest. Finally, public deployment could affect students and communities who had no opportunity to understand or challenge the system.

Formal approval is important but not conclusive. The ethical question is not simply whether the data are legally accessible or whether an algorithm produces a useful prediction. It includes proportionality, fairness, legitimacy of data reuse, contestability, and foreseeable consequences.

Responsible approach

Mira should pause deployment and document the concerns. She should consult the ethics committee, privacy experts, affected community representatives, and an independent statistician. The team should conduct subgroup evaluations, examine data provenance and reporting bias, test alternative models, and consider whether the intended target can be measured validly. The university should clarify the funding relationship and require disclosure of the sponsor’s role.

Possible safeguards include using aggregated rather than individual movement data, excluding unnecessary attributes, limiting outputs to resource planning rather than individual suspicion, publishing limitations, creating an appeal and review mechanism, and establishing an independent monitoring process. If meaningful consent or a defensible public-interest basis for reuse cannot be established, the dataset may need to be narrowed or the project redesigned.

Outcome

In this illustrative case, the team reports that predictive performance falls substantially when evaluated across different campus groups and time periods. The university decides not to deploy the original model. Instead, it funds a smaller study on emergency-response travel times using aggregated data and a participatory design process. The company withdraws from the project but permits publication of the methodological findings.

Learning points

The case demonstrates that ethical responsibility includes questioning the research question, not only protecting the dataset. Aggregate accuracy is not sufficient evidence of fairness. Institutional approval does not transfer responsibility away from the researcher. Commercial sponsorship must be disclosed and managed. A decision to redesign or discontinue a project can be an ethical success rather than a research failure.

Common Failure Modes and Institutional Safeguards

Ethical problems often arise through ordinary organizational pressures rather than dramatic fraud. Common warning signs include rushed approvals, unclear ownership of data, undocumented code changes, excessive dependence on a supervisor, unrealistic publication targets, incentives based solely on positive results, and the assumption that a public dataset is risk-free.

Researchers can reduce these risks by creating a data management plan, agreeing on authorship and roles at the beginning, using version control, maintaining an audit trail, preregistering hypotheses where appropriate, separating exploratory from confirmatory analysis, conducting internal replication, and inviting critical review from people outside the immediate project.

Institutions also have obligations. They should provide accessible ethics review, training tailored to disciplines, secure research infrastructure, fair investigation procedures, conflict-of-interest management, mentorship, whistleblower protection, and mechanisms for correcting the scholarly record. Ethics committees should avoid becoming purely bureaucratic checkpoints; they should support proportionate, context-sensitive reasoning.

Current Developments and Emerging Concerns

Several developments are reshaping research ethics. Large language models and automated analysis systems make it easier to generate text, code, synthetic data, and research hypotheses, but they complicate attribution, verification, and accountability. Data-intensive research increasingly relies on infrastructures operated by commercial platforms, raising questions about access, dependency, and long-term preservation. Open and collaborative research expands participation while creating new challenges concerning consent, moderation, intellectual property, and governance.

There is also growing attention to research justice, which asks who defines research priorities, who benefits from innovation, and whose knowledge is treated as authoritative. Responsible innovation encourages researchers to anticipate social consequences, reflect on alternatives, include affected stakeholders, and remain responsive as evidence changes. These approaches do not replace established principles; they extend them to systems whose effects may be distributed, adaptive, and difficult to reverse.

Best-Practice Principles for Doctoral Inquiry

A doctoral researcher should cultivate the following habits:

• Treat ethics as a continuing part of research design, not a one-time approval event.
• Collect only data necessary for a justified purpose and protect it throughout its life cycle.
• Keep accurate, secure, and reviewable records of methods, code, decisions, and changes.
• Report uncertainty, negative findings, limitations, conflicts, and deviations honestly.
• Discuss authorship, ownership, access, and responsibilities before disputes arise.
• Test claims across relevant populations and conditions rather than relying on a single favorable metric.
• Seek independent advice when power, privacy, security, or dual-use risks are substantial.
• Prefer the least harmful method capable of answering the research question.
• Correct the record promptly when an error or misleading claim is discovered.
• Ask not only whether a study can be conducted, but whether it ought to be conducted in its proposed form.

Chapter Summary

Research ethics is the disciplined examination of how research should be designed, conducted, reported, and used. Research integrity emphasizes honesty, rigor, transparency, reliability, and accountability, while responsible conduct also addresses respect for persons, non-maleficence, beneficence, justice, and the social consequences of knowledge. Ethical reasoning requires more than rule compliance: it involves identifying affected parties, examining duties and consequences, considering power and context, and selecting proportionate safeguards.

Research misconduct includes fabrication, falsification, and plagiarism. Questionable practices such as selective reporting, p-hacking, inadequate documentation, inappropriate authorship, and data leakage may also damage the research record even when they do not meet a formal misconduct definition. Doctoral researchers have distinctive obligations to participants, collaborators, institutions, disciplines, and future users of their work.

In information technology, privacy, data governance, artificial intelligence, cybersecurity, dual use, open science, and reproducibility create distinctive ethical challenges. Digital systems can scale harms and preserve them over time, so responsible researchers must evaluate downstream effects and design for accountability. Ethical research is therefore not an obstacle added to technical work; it is part of what makes research credible, socially legitimate, and professionally valuable.

Key Terms

Research ethics: Moral principles and practices governing the design, conduct, reporting, and use of research.

Research integrity: Honesty, rigor, transparency, reliability, and accountability in scholarly work.

Responsible conduct of research: Ethical and professional behavior across the entire research life cycle.

Informed consent: Voluntary agreement based on adequate understanding of research information and risks.

Beneficence: The obligation to promote or maximize legitimate benefits.

Non-maleficence: The obligation to avoid or minimize foreseeable harm.

Justice: Fair distribution of research burdens, benefits, opportunities, and risks.

Autonomy: A person’s capacity and right to make informed decisions without undue pressure.

Fabrication: Making up data, observations, participants, or results.

Falsification: Manipulating research materials, processes, data, or results to misrepresent the record.

Plagiarism: Presenting another person’s words, ideas, code, data, or creative work as one’s own.

Questionable research practice: A conduct that may fall short of formal misconduct but undermines research reliability or fairness.

Conflict of interest: An interest that could influence, or appear to influence, research judgment.

Data provenance: Documented information about the origin, ownership, transformation, and use of data.

Reproducibility: The ability to obtain consistent findings using the same data, methods, and documented procedures.

Responsible disclosure: A process for reporting vulnerabilities while reducing the risk of exploitation.

Dual-use research: Research that can support beneficial purposes but may also enable harmful applications.

Research misconduct: Serious violation of research standards, especially fabrication, falsification, or plagiarism.

Calibrated transparency: Providing enough information for evaluation while protecting privacy, security, and legitimate confidentiality.

Research justice: Attention to who defines research priorities, bears research burdens, and receives its benefits.

Review Questions

Short Answer Questions

1. How do research ethics and research integrity differ, and why must both be addressed?

2. Define fabrication, falsification, and plagiarism.

3. Why is informed consent considered an ongoing process rather than merely a signed form?

4. What is a questionable research practice? Give two examples relevant to computing.

5. Why does public availability of digital information not automatically make its research use ethically unproblematic?

Long Answer Questions

1. Explain the principles of autonomy, beneficence, non-maleficence, justice, and accountability, using information technology examples.

2. Compare deontological, consequentialist, virtue-ethical, and relational approaches to research ethics.

3. Discuss the distinctive ethical responsibilities of a doctoral researcher within a supervisory research team.

4. Analyse the ethical challenges of artificial intelligence research, including data provenance, fairness, transparency, and dual use.

5. Explain how open science can improve research integrity while also creating privacy and security risks.

Analytical / Application Questions

1. A researcher removes several unsuccessful experimental runs from a machine-learning evaluation because they are described as “technical anomalies,” but does not report the exclusions. Analyse whether this is misconduct, a questionable practice, or an honest methodological decision. What documentation and disclosure would be required?

2. You discover that a publicly available dataset contains indirectly identifiable records about a vulnerable community. Design an ethically defensible plan for deciding whether and how the dataset may be used.

3. A technology sponsor asks a doctoral researcher to publish only the subgroup in which a predictive model performs best. Apply the ethical reasoning sequence from this chapter and recommend a course of action.

22. Multiple Choice Questions with Answers

Question 1

Which statement best distinguishes research integrity from research ethics?

A. Research integrity concerns only legal compliance, whereas research ethics concerns only publication.

B. Research integrity emphasizes trustworthy scholarly practice, while research ethics also addresses the treatment and interests of those affected by research.

C. Research integrity applies only to quantitative studies, whereas research ethics applies only to qualitative studies.

D. Research integrity is required only after data collection, while research ethics ends after approval.

Correct Answer: B. Research integrity emphasizes trustworthy scholarly practice, while research ethics also addresses the treatment and interests of those affected by research.

Question 2

A researcher reports only the most favorable random seed from many model runs without disclosing the selection process. This is best classified as:

A. A questionable research practice that may distort the evidence.

B. Fabrication because every reported result is necessarily invented.

C. Informed consent because the researcher chose a result voluntarily.

D. Responsible disclosure because the result is publicly reported.

Correct Answer: A. A questionable research practice that may distort the evidence.

Question 3

Why is institutional ethics approval not sufficient to transfer all ethical responsibility away from a doctoral researcher?

A. Approval guarantees that every future use of data is ethical.

B. Approval replaces the need for professional judgment during analysis and publication.

C. Approval reviews a proposed activity, while researchers remain responsible for changing risks, honest reporting, and unforeseen consequences.

D. Approval is relevant only to funding and has no connection with research conduct.

Correct Answer: C. Approval reviews a proposed activity, while researchers remain responsible for changing risks, honest reporting, and unforeseen consequences.

Question 4

Which action most directly reflects the principle of justice in an AI research project?

A. Maximizing overall accuracy without examining subgroup performance.

B. Selecting participants solely because they are easiest to access.

C. Examining whether research burdens and system errors are unfairly distributed across affected groups.

D. Keeping the model proprietary so that no one can criticize it.

Correct Answer: C. Examining whether research burdens and system errors are unfairly distributed across affected groups.

Question 5

A cybersecurity researcher discovers a vulnerability in a company’s system. Which response is most consistent with responsible disclosure?

A. Immediately publishing exploit code without contacting anyone.

B. Testing unrelated systems to determine whether the vulnerability is widespread.

C. Defining authorized scope, protecting exploit details, and coordinating notification with the affected organization.

D. Suppressing all evidence permanently, regardless of public risk.

Correct Answer: C. Defining authorized scope, protecting exploit details, and coordinating notification with the affected organization.

Question 6

Which statement about anonymization is most accurate?

A. Anonymization always makes research use ethically risk-free.

B. Anonymization may reduce risk, but combinations of attributes can still permit re-identification.

C. Anonymization eliminates the need for data access controls.

D. Anonymization means that consent is never relevant.

Correct Answer: B. Anonymization may reduce risk, but combinations of attributes can still permit re-identification.

Question 7

A researcher discovers that a published conclusion depends on a programming error. What is the most responsible first response?

A. Delete the underlying records so the error cannot be investigated.

B. Continue citing the conclusion until another researcher notices the problem.

C. Document the error, assess its effect, inform relevant parties, and correct the research record where necessary.

D. Reclassify the error as fabrication automatically.

Correct Answer: C. Document the error, assess its effect, inform relevant parties, and correct the research record where necessary.

Question 8

Which scenario illustrates a conflict of interest rather than automatically proving misconduct?

A. A researcher receives funding from a company whose technology is being evaluated and discloses and manages that relationship.

B. A researcher invents participants who never existed.

C. A researcher copies another author’s code without attribution.

D. A researcher alters observations to obtain a desired result.

Correct Answer: A. A researcher receives funding from a company whose technology is being evaluated and discloses and manages that relationship.

Question 9

Why may high aggregate accuracy be ethically insufficient for a predictive system?

A. Accuracy is never relevant to ethical evaluation.

B. Aggregate accuracy can conceal unequal error rates or harmful effects on particular groups.

C. High accuracy proves that the data were collected with valid consent.

D. Aggregate accuracy automatically demonstrates fairness.

Correct Answer: B. Aggregate accuracy can conceal unequal error rates or harmful effects on particular groups.

Question 10

Which practice best supports computational reproducibility?

A. Releasing only a final graph without code or methodological records.

B. Changing evaluation data until the desired result appears.

C. Maintaining versioned code, documenting data transformations and dependencies, and reporting deviations.

D. Keeping all analytical decisions undocumented to preserve flexibility.

Correct Answer: C. Maintaining versioned code, documenting data transformations and dependencies, and reporting deviations.