Research Ethics, Privacy and Data Governance

Introduction

Marketing research produces value by converting information about people, markets, and organizations into decisions. The same process can also create significant risks. A survey may reveal health conditions, political preferences, or financial vulnerability; a digital trace may be collected without meaningful awareness; a segmentation model may exclude a group from an offer; and an apparently harmless dataset may identify individuals when combined with other sources.

Research ethics therefore is not an administrative hurdle added after the research design. It is a condition of methodological quality. Participants who feel deceived, coerced, or exposed may give distorted answers or withdraw. Poor governance can invalidate findings, damage brand trust, attract regulatory penalties, and harm people who never consented to become research subjects.

This chapter develops a practical and critical framework for ethical marketing research. It addresses informed consent, transparency, confidentiality, fairness, harm minimization, privacy, intellectual property, vulnerable participants, retention, security, regulatory requirements, professional codes, and the governance of contemporary data and artificial intelligence.

Learning objectives

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

1. Distinguish ethics, privacy, confidentiality, anonymity, and data governance.
2. Design proportionate informed-consent and transparency procedures.
3. Identify and mitigate physical, psychological, social, economic, legal, and informational harms.
4. Manage sensitive data, vulnerable participants, intellectual property, retention, and security.
5. Evaluate research against applicable laws, institutional rules, and professional codes.
6. Assess fairness and accountability in digital, platform, and AI-enabled research.
7. Defend an ethical research decision using a documented risk-based process.

1. Ethics as a condition of valid marketing knowledge

Research ethics concerns what researchers ought to do when their choices affect participants, communities, clients, employees, the public, and future users of data. Privacy concerns a person’s control over information about them and the appropriate conditions under which information is accessed or used. Confidentiality is a professional and organizational promise to restrict disclosure of information supplied in confidence. Anonymity means that identity is not known or cannot reasonably be linked to the response; it is stronger than merely removing names from a file.

Data governance is the system of decision rights, responsibilities, standards, controls, and procedures through which data are collected, classified, accessed, used, shared, retained, and destroyed. Governance connects ethical principles to operational practice. A consent statement without access controls is incomplete governance; encryption without a lawful and fair purpose is not ethical research.

Three classic ethical principles remain useful:

• Respect for persons: treat participants as autonomous decision-makers and provide additional protection where autonomy is limited.
• Beneficence and non-maleficence: pursue worthwhile benefits while minimizing foreseeable harm.
• Justice: distribute the burdens and benefits of research fairly, avoiding exploitation and unjust exclusion.

These principles are reflected in the Belmont Report, the Declaration of Helsinki, professional standards, and many institutional review systems. Marketing research also has distinctive features: commercial objectives, secondary data, covert observation, incentives, client pressure, cross-border processing, and the possibility that research findings will influence prices, credit, employment, advertising, or access to services.

2. A risk-based ethical framework

A defensible research protocol should answer five questions before data collection:

1. Purpose: What decision or knowledge claim justifies collecting the data?
2. Necessity: Is each variable needed, or is collection driven by convenience or curiosity?
3. Proportionality: Are the expected benefits sufficient in relation to the risks?
4. Control: Can participants exercise meaningful choice, withdrawal, access, correction, or complaint rights?
5. Accountability: Can the researcher demonstrate who decided what, when, and on what evidence?

Risk assessment should consider both the probability and severity of harm. A rare disclosure of a highly sensitive dataset may be more serious than frequent disclosure of a low-sensitivity preference. Relevant harms include:

• physical or health-related harm;
• psychological distress, embarrassment, or stigma;
• social harm, discrimination, or reputational damage;
• economic loss, manipulation, or denial of opportunity;
• legal or immigration consequences;
• informational harm caused by surveillance, re-identification, or inappropriate inference.

Risk is not limited to the direct participant. It may affect household members, employees, communities, non-participants captured by observation, or groups represented in a model. A sound protocol identifies foreseeable misuse after the research is delivered to a client.

A practical risk register

Risk registers should be reviewed when the sample, instrument, vendor, technology, or intended use changes. A risk assessment is not a one-time form.

3. Informed consent and transparency

Informed consent requires more than a tick box. Participants should understand, in language appropriate to their context:

• who is conducting or sponsoring the research;
• the purpose and expected duration;
• what participation involves and what data will be collected;
• foreseeable risks and benefits;
• whether participation is voluntary and how to stop;
• incentives and any conditions attached to them;
• who will receive results and whether data will be shared;
• retention, deletion, and cross-border processing arrangements;
• contact details for questions, complaints, or data-rights requests.

Consent must be informed, voluntary, specific, and capable of being withdrawn where applicable. A participant cannot meaningfully consent when essential information is hidden in legalistic text, when refusal threatens employment, or when a platform bundles research consent with unrelated marketing permission. Consent should be renewed if the purpose materially changes.

Transparency is broader than consent. In observational, digital, or secondary-data research, researchers may be unable to obtain individual consent in the conventional manner. They should then provide an appropriate notice where feasible, justify the alternative lawful basis or ethical rationale, minimize data, assess expectations of privacy, and provide an objection or opt-out mechanism when possible. Public availability does not automatically make data ethically free to use. A person may post information for a social purpose without expecting it to be scraped, profiled, and sold.

Consent quality checklist

A consent process is stronger when it is:

• Comprehensible: short sentences, translations, examples, and comprehension checks.
• Granular: separate choices for research, future use, recording, data linkage, and commercial contact.
• Unpressured: incentives are not so large that they undermine free choice.
• Documented: date, version, method, and scope are recorded.
• Reversible: withdrawal instructions are practical and visible.

Withdrawal does not necessarily require erasing information already irreversibly incorporated into aggregated results, but the limits must be explained. Researchers should not promise deletion that their systems cannot perform.

4. Privacy, confidentiality, and data minimization

Privacy by design means embedding privacy protections in the research architecture rather than retrofitting them after collection. Data minimization asks whether the same research objective can be achieved with fewer variables, lower precision, shorter retention, or less intrusive collection. For example, a study of delivery satisfaction may need a service region, not a precise home address; an age band may suffice instead of date of birth.

Identifiers should be classified. Direct identifiers include name, email address, and government identification number. Quasi-identifiers include combinations such as age, postcode, occupation, and rare purchase behavior. Sensitive data may include health, biometric, political, religious, racial or ethnic, sexual, financial, or criminal information, depending on jurisdiction. Classification should reflect context and re-identification risk, not merely the field label.

Pseudonymization replaces identifiers with codes while retaining a re-linking key separately. It reduces risk but is not anonymity. Anonymization aims to make identification reasonably impossible, considering other accessible data, technological change, and the effort required. Small cells, unique combinations, free-text responses, voice recordings, and precise location can defeat superficial anonymization.

Confidentiality controls include:

• collecting only necessary data;
• separating contact details from response data;
• using role-based access and least privilege;
• encrypting data in transit and at rest;
• keeping keys separately from encrypted files;
• suppressing small cells in reports;
• removing identifying quotations or obtaining explicit permission to use them;
• training staff and binding vendors contractually;
• maintaining access and disclosure logs.

Confidentiality is also a reporting obligation. A client may want a compelling customer story, but a vivid quotation can identify a participant in a small community. Researchers should obtain permission, paraphrase, aggregate, or omit the material.

5. Fairness, inclusion, and harm minimization

Fairness has at least three dimensions. Procedural fairness concerns whether people are treated respectfully and given a meaningful opportunity to participate or challenge decisions. Distributive fairness concerns who bears research burdens and who benefits. Representational fairness concerns whether categories, language, and interpretations portray groups accurately and respectfully.

Sampling can be unfair through systematic exclusion, such as conducting an online-only survey about public services when some target groups have limited internet access. Inclusion does not mean forcing participation; it means reducing avoidable barriers through accessible modes, translation, disability accommodations, culturally appropriate recruitment, and careful incentives.

Researchers should distinguish demographic differences in responses from evidence of bias. A model can have similar overall accuracy across groups while producing materially different false-positive or false-negative rates. Consequently, fairness assessment should be connected to the decision and harm at stake, not reduced to a single metric. If a research output will guide eligibility, pricing, or targeting, the client should define prohibited uses, require human review, and provide correction or appeal mechanisms.

Harm minimization may require excluding a question, changing the order of questions, using a neutral interviewer, offering a skip option, avoiding deceptive stimuli, or providing support information. Deception is ethically exceptional: it requires strong justification, minimal risk, absence of a feasible non-deceptive alternative, and prompt debriefing where appropriate.

6. Vulnerable participants and power relationships

Vulnerability is contextual rather than a permanent label. Children, people with cognitive impairment, prisoners, refugees, people experiencing illness or poverty, and employees in subordinate relationships may face special risks. A participant can also be vulnerable because the topic is stigmatizing or because the researcher controls access to a valued service.

Researchers should assess capacity, voluntariness, comprehension, and the consequences of refusal. Children commonly require age-appropriate assent and permission from a parent or authorized guardian, subject to local law and institutional policy. Guardian permission does not replace the child’s developing autonomy. For employees, recruitment should avoid supervisors knowing who declined; participation should not affect appraisal, scheduling, or employment. For people with limited literacy, oral explanation, visual aids, interpreter support, and a comprehension check may be necessary.

Compensation should reimburse time and inconvenience without becoming coercive. Researchers must plan safeguarding procedures before fieldwork, including what happens if a participant discloses imminent danger, abuse, or a serious risk. Promises of absolute confidentiality may be unethical or unlawful if a defined reporting duty applies; limits should be disclosed in advance.

7. Intellectual property and ownership of research materials

Research governance includes intellectual property (IP), not only personal data. The parties should clarify ownership and permitted use of:

• questionnaires, interview guides, stimuli, scales, code, and dashboards;
• raw data, cleaned datasets, transcripts, and derived variables;
• recordings, photographs, and participant-generated content;
• reports, presentations, models, and training data;
• third-party databases, software, images, and copyrighted instruments.

Copyright, database rights, trade secrets, trademarks, patents, and contractual licenses differ by jurisdiction. Purchasing access to a dataset does not necessarily transfer ownership or permit onward sharing. A researcher should document source, license, attribution requirements, restrictions on commercial use, and whether modification is allowed. Academic integrity also requires accurate attribution and prohibits presenting client or participant work as the researcher’s own.

Contracts should resolve ownership, publication rights, confidentiality, audit rights, breach notification, subcontracting, deletion or return at project end, and use of data to train AI systems. Where participant-generated content is intended for public display, consent should cover that specific use rather than relying on general research consent.

8. Retention, deletion, and security lifecycle

Retention should be tied to purpose, legal requirements, reproducibility needs, and risk. A retention schedule specifies the data category, owner, storage location, retention period, review date, and disposal method. Keeping data indefinitely “in case it becomes useful” conflicts with minimization and increases breach exposure.

Deletion should address production systems, backups, exports, notebooks, local devices, vendor copies, and derived files. Where immediate deletion is technically impossible, access should be restricted and deletion completed according to a documented cycle. Legal holds, audit requirements, or a participant’s request may require a reasoned exception.

A proportionate security program uses administrative, technical, and physical controls:

• governance: policies, training, vendor due diligence, incident roles;
• technical: multifactor authentication, encryption, patching, endpoint protection, segmentation, backups, logging, and vulnerability testing;
• physical: controlled facilities, secure disposal, screen protection, and protection against theft.

An incident response plan should identify, contain, investigate, document, and learn from a breach. Notification duties vary by law and by the nature and jurisdiction of affected persons. The project team should escalate promptly rather than waiting for certainty.

9. Regulatory and professional governance

No single global law governs all marketing research. Applicability depends on jurisdiction, sector, data type, controller–processor relationships, location of participants, and intended use. Important reference points include:

• GDPR and related European data-protection law: principles include lawfulness, fairness and transparency, purpose limitation, data minimization, accuracy, storage limitation, integrity and confidentiality, and accountability. It may apply extraterritorially. Rights and lawful bases require legal analysis; consent is not always the correct basis.
• CCPA/CPRA and other U.S. state privacy laws: may provide notice, access, deletion, correction, opt-out, and limits concerning sale or sharing, subject to scope and exemptions.
• Sectoral and national rules: health, children’s data, financial information, telecommunications, biometrics, employment, and cross-border transfers may have special requirements.
• Human-subjects and institutional review rules: universities and some organizations may require ethics review even where a commercial project is not legally classified as human-subject research.
• Professional codes: ICC/ESOMAR standards, Insights Association guidance, and comparable national codes emphasize participant protection, transparency, professional competence, and responsible reporting.
• Information-security standards: ISO/IEC 27001 supports an information-security management system; ISO/IEC 27701 extends privacy-information management practices. Certification is not a substitute for ethical judgment or legal compliance.

A useful compliance matrix maps each project activity to the relevant rule, owner, evidence, and decision. Legal compliance is a floor, not the complete ethical standard. A practice may be technically lawful yet violate reasonable expectations, create unjust discrimination, or undermine public trust.

10. Governance of digital and AI-enabled research

Digital research introduces passive measurement, cookies, device identifiers, social listening, geolocation, facial or voice analysis, automated transcription, synthetic respondents, and generative AI. These tools can increase scale while reducing participant visibility and increasing inferential risk.

Before using an AI system, researchers should determine:

• what data the provider stores and whether it uses inputs for model training;
• where processing occurs and which subcontractors have access;
• whether sensitive or identifiable data can be excluded;
• how outputs are validated, logged, and corrected;
• whether the system fabricates quotations, demographic claims, or citations;
• whether automated decisions affect participants or groups;
• how bias, prompt leakage, re-identification, and model drift are managed.

Synthetic data are not automatically anonymous. If generated from small or distinctive records, they may reproduce identifiable patterns or sensitive features. AI-generated summaries should be treated as analyst assistance, not evidence; researchers remain responsible for checking source fidelity, uncertainty, and omissions. Participants should be told when interaction is with an automated agent if that fact affects their choice or expectations.

11. An operational governance model

A mature project assigns roles rather than assuming that “the researcher” is accountable for everything. Typical roles include principal investigator or research lead, data controller or client sponsor, data steward, security lead, fieldwork vendor, analyst, and ethics or privacy adviser. A responsibility matrix should identify who approves collection, handles requests, authorizes access, responds to incidents, and signs off dissemination.

The lifecycle can be managed as follows:

1. Define purpose, population, expected use, and prohibited use.
2. Map data flows from recruitment through destruction.
3. Classify sensitivity and assess risks to participants and groups.
4. Select a lawful and ethically defensible basis; design consent and notice.
5. Minimize variables, test instruments, and plan accessibility.
6. Contract and audit vendors; control access and transfers.
7. Monitor fieldwork, complaints, deviations, and adverse events.
8. Validate analysis for privacy, fairness, and unsupported inference.
9. Report limitations and protect identities in dissemination.
10. Review retention, delete or archive appropriately, and document lessons.

Documentation is an accountability mechanism. Keep protocol versions, consent materials, approvals, data maps, risk assessments, training records, access logs, incident records, and deletion certificates. The aim is not paperwork for its own sake; it is the ability to reconstruct and defend a decision.

Applied case: The loyalty-app insight project

The following is an illustrative case.

Background

A grocery chain proposes to combine loyalty-card purchases, mobile-app location data, customer-service transcripts, and a survey to identify households likely to reduce spending. The client wants personalized offers and asks the research agency to retain raw data indefinitely for future modeling.

Problem

The proposed combination may reveal health conditions, household composition, financial stress, and movement patterns. Customers joined the loyalty program to receive discounts, not necessarily to permit unrestricted research or automated vulnerability profiling. The agency also plans to send transcripts to a cloud AI vendor whose contract permits service improvement using submitted content.

Analysis

The research purpose is not sufficiently specific, and the requested variables exceed what is necessary to evaluate offer preferences. Location data and free-text transcripts create high re-identification risk. The proposed “consent” is buried in general terms, and indefinite retention conflicts with purpose limitation and storage minimization. A model that targets financially stressed households with urgency-based promotions could exploit vulnerability. The AI contract creates an uncontrolled secondary use.

Approach and solution

The team separates operational loyalty administration from the research dataset, obtains a documented purpose-specific notice, and offers a meaningful opt-out. It removes precise location, replaces dates of birth with age bands, excludes health inferences, and uses a short voluntary survey for stated preferences. Contact details are held separately; the analysis file is pseudonymized, access is role-based, and small segments are suppressed in reporting. A vendor contract prohibits model training and onward use, requires deletion, specifies breach notification, and permits audit. The model is used only to test aggregate offer strategies, not to make individual eligibility decisions. Fairness testing examines error patterns across relevant groups, and a human review process is required for consequential actions. Data are retained for a documented period, then securely deleted unless a justified, approved archive is needed.

Outcome

The client receives less granular data but a more defensible insight program. Participants receive clearer choices, the risk of discriminatory targeting is reduced, and the agency can explain its methods to regulators and the public. The project also reveals that some desired predictions were not necessary for the business decision.

Learning points

Ethical governance can improve research design rather than merely constrain it. Purpose limitation, minimization, vendor control, fairness review, and retention discipline should be designed together. The most valuable question is often not “Can we collect this?” but “What decision requires it, and what is the least intrusive evidence that can support that decision?”

Challenges and limitations

Ethical principles can conflict. Greater transparency may reveal a study purpose and change behavior; strict anonymization may reduce analytical value; retention may support reproducibility while increasing exposure; incentives may improve inclusion while exerting pressure; and fairness metrics may conflict when base rates differ. No checklist resolves these tensions automatically.

Legal rules also differ across jurisdictions and change over time. Researchers should obtain qualified legal and ethics advice for high-risk projects rather than treating this chapter as legal advice. Professional codes may be voluntary, and certification can create false confidence if controls are not implemented. Finally, governance can become performative: lengthy notices, formal approvals, and policy documents do not compensate for inaccessible consent, weak security, or organizational incentives that reward aggressive data extraction.

Best-practice principles

A graduate researcher should be able to defend the following commitments:

• collect only what the purpose requires;
• make participation and data use understandable;
• protect people who face greater risk or reduced autonomy;
• separate identity from analysis wherever possible;
• anticipate downstream and group harms, not only direct harms;
• give vendors no broader rights than the project requires;
• test analytical outputs for privacy, validity, and fairness;
• retain data for a reason and delete it responsibly;
• document decisions and assign accountable owners;
• treat law, professional standards, and public expectations as complementary sources of responsibility.

Chapter Summary

Ethical marketing research is both a moral responsibility and a condition of credible evidence. Respect for persons, beneficence, non-maleficence, and justice provide foundational principles. A risk-based approach translates them into decisions about purpose, necessity, proportionality, control, and accountability.

Informed consent must be understandable, voluntary, specific, documented, and capable of withdrawal. Transparency remains important in observational and secondary-data research, where public availability does not automatically remove privacy expectations. Privacy by design requires minimization, careful classification, pseudonymization, access control, secure processing, and responsible reporting. Confidentiality is a promise supported by technical, contractual, and organizational safeguards.

Fairness requires attention to recruitment, representation, analytical error, downstream decisions, and group harms. Vulnerable participants need additional protection, while intellectual-property rights and vendor contracts must define permissible use. Retention and deletion should follow a documented schedule. Regulatory regimes, institutional review, professional codes, and information-security standards provide overlapping governance layers, but compliance is only a minimum. Digital and AI-enabled research demands particular scrutiny of inference, automated decisions, model training, bias, and data transfers. Strong governance assigns roles, maps the data lifecycle, records decisions, and continually reviews changing risks.

Key Terms

Ethics: Principles governing what ought to be done when research affects people and society.

Privacy: Appropriate control over information about a person and its use.

Confidentiality: A commitment to restrict disclosure of information supplied in confidence.

Anonymity: A condition in which identity cannot reasonably be linked to data.

Pseudonymization: Replacing identifiers with codes while retaining a separate re-linking key.

Informed consent: Voluntary agreement based on adequate, understandable information.

Transparency: Openness about research purposes, methods, data use, and rights.

Data minimization: Collecting and retaining no more personal data than necessary.

Purpose limitation: Using data for specified, legitimate, and compatible purposes.

Vulnerable participant: A person whose autonomy or ability to avoid harm is reduced in context.

Fairness: Just treatment in participation, analysis, representation, and outcomes.

Data governance: Decision rights, controls, responsibilities, and standards for managing data.

Data steward: The person responsible for operational data quality, access, and lifecycle controls.

Direct identifier: A data element that directly identifies a person.

Quasi-identifier: A data element or combination that may enable re-identification.

Retention schedule: A documented rule specifying how long data are kept and when they are reviewed or destroyed.

Intellectual property: Legally protected creations, information, code, content, and related rights.

Pseudonymous data: Data linked to a person through a separately held code or key.

Privacy by design: Integrating privacy protections into systems and processes from the beginning.

Review Questions

Short Answer Questions

1. Distinguish privacy, confidentiality, and anonymity.
2. What information should a marketing-research consent statement contain?
3. Why is pseudonymization not the same as anonymization?
4. Identify three possible harms arising from digital marketing research.
5. Why should retention periods be linked to research purpose?

Long Answer Questions

1. Explain how the principles of respect for persons, beneficence, and justice apply to a customer-segmentation study.
2. Compare the ethical challenges of primary survey research with those of secondary social-media data research.
3. Discuss how a research agency can protect vulnerable participants who are also employees of the sponsoring organization.
4. Explain why legal compliance is necessary but not sufficient for ethical marketing research.
5. Evaluate the governance controls needed when a cloud-based AI tool processes research transcripts.

Analytical / Application Questions

1. A client asks for precise location histories and health-related purchase inferences to personalize offers. Construct a risk assessment and propose a less intrusive design.
2. A survey vendor accidentally exposes a file containing pseudonymous responses and a separate contact key. Analyze the incident and specify immediate and longer-term actions.
3. An AI model recommends excluding a demographic group from a promotional campaign because its historical response rate is low. Design an investigation addressing validity, fairness, participant expectations, and governance before the recommendation is implemented.

Multiple Choice Questions with Answers

Question 1

Which statement best distinguishes pseudonymization from anonymization?

A. Pseudonymization removes all privacy risk, whereas anonymization does not.

B. Pseudonymization replaces identifiers but retains a possible re-linking key; anonymization aims to make re-identification reasonably impossible.

C. Pseudonymization applies only to paper records, whereas anonymization applies only to digital records.

D. Pseudonymization requires consent, whereas anonymization never requires ethical review.

Correct Answer: B. Pseudonymization replaces identifiers but retains a possible re-linking key; anonymization aims to make re-identification reasonably impossible.

Question 2

Which practice most directly demonstrates data minimization?

A. Collecting precise home addresses for every respondent in case they are useful later

B. Keeping all raw data indefinitely to support possible future studies

C. Collecting a service region instead of a precise address when the region is sufficient for the research purpose

D. Combining survey answers with every available commercial database

Correct Answer: C. Collecting a service region instead of a precise address when the region is sufficient for the research purpose

Question 3

Which element is essential to informed consent?

A. A guarantee that participation will produce personal benefits

B. A clear explanation of purpose, procedures, risks, and voluntary participation

C. A requirement that participants waive all future privacy rights

D. A promise that data will never be used in any aggregate analysis

Correct Answer: B. A clear explanation of purpose, procedures, risks, and voluntary participation

Question 4

Why is public availability of social-media content not automatically sufficient ethical justification for research use?

A. Public data can never be used for research.

B. Public visibility may not match the person’s reasonable expectation about scraping, profiling, or commercial reuse.

C. Social-media data are always inaccurate.

D. Researchers are prohibited from analyzing text data.

Correct Answer: B. Public visibility may not match the person’s reasonable expectation about scraping, profiling, or commercial reuse.

Question 5

Which control most directly supports confidentiality?

A. Publishing small subgroup results without checking re-identification risk

B. Giving every project employee unrestricted access

C. Separating contact details from response data and applying role-based access

D. Retaining identifiable data permanently

Correct Answer: C. Separating contact details from response data and applying role-based access

Question 6

What is the most appropriate ethical response when a research design involves vulnerable participants?

A. Exclude all vulnerable people automatically from every study

B. Assess capacity, voluntariness, risks, safeguards, and appropriate permission or assent

C. Offer an extremely large incentive so participation is guaranteed

D. Allow supervisors to recruit and identify employees who refuse

Correct Answer: B. Assess capacity, voluntariness, risks, safeguards, and appropriate permission or assent

Question 7

Which statement about professional codes and privacy laws is most accurate?

A. Following a professional code always removes the need to comply with law.

B. Legal compliance is a useful baseline, but ethical responsibility may require more than the law.

C. Professional codes apply only to academic researchers.

D. Privacy laws make institutional ethics review unnecessary.

Correct Answer: B. Legal compliance is a useful baseline, but ethical responsibility may require more than the law.

Question 8

A research team wants to send identifiable transcripts to an AI vendor. Which first step is most important?

A. Assume the vendor will delete the data because it is reputable

B. Check processing terms, training use, access, location, security, retention, and contractual restrictions

C. Publish the transcripts before sending them

D. Remove punctuation from the transcripts and send them without review

Correct Answer: B. Check processing terms, training use, access, location, security, retention, and contractual restrictions

Question 9

Why should a retention schedule include a review or deletion date?

A. To ensure data remain available forever

B. To link continued storage to purpose, legal requirements, reproducibility, and risk

C. To avoid documenting data ownership

D. To make encryption unnecessary

Correct Answer: B. To link continued storage to purpose, legal requirements, reproducibility, and risk

Question 10

Which action best reflects fairness in an online-only public-service survey?

A. Assume non-internet users are irrelevant to the target population

B. Use accessibility accommodations and alternative modes where exclusion would distort representation

C. Replace nonrespondents with fabricated responses

D. Offer participation only to the easiest-to-reach group

Correct Answer: B. Use accessibility accommodations and alternative modes where exclusion would distort representation