Introduction to Social Research Methodology

Research ethics

Ben Stanley

Department of Social Sciences, SWPS University

November 3, 2026

Course outline

Course outline: session 4 (today) Course outline: 15 stations. 1 — Nature and purpose of social research (done); 2 — Research goals and research questions (done); 3 — Hypotheses and research design (done); 4 — Research ethics (today); 5 — Conducting a literature review (to come); 6 — Secondary data analysis (to come); 7 — Qualitative approaches (to come); 8 — Quantitative approaches and measurement (to come); 9 — Mixed methods research (to come); 10 — Sampling and participant recruitment (to come); 11 — Questionnaire design and survey fieldwork (to come); 12 — Interviews and focus groups (to come); 13 — Quantitative data analysis (to come); 14 — Qualitative data analysis (to come); 15 — Reporting research; course review (to come). 1 Nature and purpose of social research 2 Research goals and research questions 3 Hypotheses and research design 4 Research ethics TODAY 5 Conducting a literature review 6 Secondary data analysis 7 Qualitative approaches 8 Quantitative approaches and measurement 9 Mixed methods research 10 Sampling and participant recruitment 11 Questionnaire design and survey fieldwork 12 Interviews and focus groups 13 Quantitative data analysis 14 Qualitative data analysis 15 Reporting research; course review

Today’s lecture

  • Foundations — why ethics is not a formality but a condition of good research
  • Core principles — voluntary participation, informed consent, no harm, privacy, and the problem of deception
  • Cautionary tales — the studies that broke the rules and built the modern review system
  • Ethics in organisational research — what changes when the researcher is also the manager, the employer, or the colleague
  • By the end, you should be able to look at any proposed study and ask: who could this harm, and what would make it defensible?

Foundations of research ethics

Why research ethics matters

Why research ethics matters: three things it protects A banner: social research studies people, so every study is an intervention in someone’s life, however small. Ethics protects three things at once, shown as three cards. Participants: their rights, dignity and well-being. The research itself: data obtained through fear or deception is usually bad data, because a frightened employee tells you what is safe, not what is true. The reputation of research: every scandal makes participants warier and the next study harder for everyone, because trust is a commons. A bracket joins the three: they stand or fall together. At Meridian: an employee survey run badly can damage trust, morale and careers, not just a dataset. In short: ethics is not a hurdle before the real research begins; it is part of what makes the research real (Babbie, 2017). Social research studies people — so every study is an intervention in someone’s life, however small ETHICS PROTECTS THREE THINGS AT ONCE Participants their rights, dignity and well-being The research itself data obtained through fear or deception is usually bad data: a frightened employee tells you what is safe, not what is true The reputation of research every scandal makes participants warier and the next study harder for everyone: trust is a commons the three stand or fall together AT MERIDIAN An employee survey run badly can damage trust, morale and careers — not just a dataset. IN SHORT Ethics is not a hurdle before the “real” research begins; it is part of what makes the research real (Babbie, 2017).

Key principles of research ethics

Five principles of research ethics, and the five working rules that put them into practice First codified in medicine, through the Nuremberg Code and then the Belmont Report of 1979, and now standard across the social sciences. Five principles, one per row. Respect for persons, the master principle: treat participants as autonomous agents, never as mere material for a study. Autonomy, which flows from respect: people decide for themselves whether to take part. Beneficence: the research should do good. Nonmaleficence: at an absolute minimum, the research must avoid doing harm; a bracket marks beneficence and nonmaleficence as two different demands. Justice: share risks and benefits fairly; never load the risks onto the powerless. An arrow leads to the five working rules, how the abstractions become practice: voluntary participation, informed consent, no harm, privacy, and deception, which is a near-prohibition. First codified in medicine — the Nuremberg Code, then the Belmont Report (1979) — and now standard across the social sciences Respect for persons the master principle treat participants as autonomous agents, never as mere material for a study Autonomy flows from respect people decide for themselves whether to take part Beneficence the research should do good Nonmaleficence at an absolute minimum, the research must avoid doing harm two different demands Justice share risks and benefits fairly; never load the risks onto the powerless FIVE WORKING RULES how the abstractions become practice 1 Voluntary participation 2 Informed consent 3 No harm 4 Privacy 5 Deception a near-prohibition

Core ethical principles

Voluntary participation

Voluntary participation: a free choice at every stage, the test, and soft coercion A free choice at every stage, along a timeline. Before: take part or not, without pressure and without penalty. During: decline any question, because agreeing to a survey is not agreeing to answer everything. At any point: withdraw, without giving a reason and without consequence. The test: not whether we said participation was voluntary, but whether a reasonable participant would feel free to refuse. Soft coercion: no threat is made, yet refusal now has a price. Examples: course credit, a gift voucher, the boss’s cheerful “I hope everyone will take part”. Refusal feels costly, in money, convenience or how one is seen. Hold that thought: in the workplace, voluntary is the principle that comes under the greatest strain, and the final section of the lecture returns to it. A FREE CHOICE — AT EVERY STAGE BEFORE Take part — or not without pressure and without penalty DURING Decline any question agreeing to a survey is not agreeing to answer everything AT ANY POINT Withdraw without giving a reason, and without consequence THE TEST “We said participation was voluntary.” not the test “Would a reasonable participant feel free to refuse?” the test SOFT COERCION no threat is made — yet refusal now has a price Course credit A gift voucher The boss’s cheerful “I hope everyone will take part” Refusal feels costly — in money, convenience or how one is seen HOLD THAT THOUGHT In the workplace, “voluntary” comes under the greatest strain — the final section returns to it.

No harm to participants

No harm to participants: three kinds of harm, and the duty to anticipate them Three kinds of harm to anticipate, rarely physical. Psychological: distress, embarrassment, reawakened trauma, for example being asked in detail about a bereavement or a redundancy. Social: damaged reputation or relationships if answers become known. Economic: consequences for a job, a promotion or benefits. A bracket over social and economic harm: usually caused not by the questions but by a failure of protection afterwards. Anticipate, don’t apologise: before the study, a risk assessment asks what could go wrong for participants; it leads to the mitigation tools, careful question design, debriefing, signposting support and, above all, protecting what participants tell you. At work: asking employees to rate their manager is harmless only if the manager can never trace an answer back to its author. THREE KINDS OF HARM TO ANTICIPATE — RARELY PHYSICAL Psychological distress, embarrassment, reawakened trauma e.g. being asked in detail about a bereavement or a redundancy Social damaged reputation or relationships if answers become known Economic consequences for a job, a promotion or benefits usually caused not by the questions, but by a failure of protection afterwards ANTICIPATE, DON’T APOLOGISE Before the study: a risk assessment — what could go wrong for them? Mitigation tools careful question design debriefing signposting support above all: protecting what participants tell you AT WORK Asking employees to rate their manager is harmless only if the manager can never trace an answer back to its author.

Privacy, anonymity, confidentiality

Privacy, anonymity and confidentiality: three terms that are not interchangeable Three panels. Privacy, about the boundary itself: the participant’s right to control what they reveal about themselves, and when; the pictogram shows a person inside a boundary, revealing only if they choose. Anonymity, a property of the data: not even the researcher can link a response to a person; the line between person and response is broken. Confidentiality, a property of the researcher’s conduct: the researcher could identify respondents, but promises not to reveal identities; the person and the response are linked, but the link is kept locked. A warning: most anonymous surveys are only confidential. Send invitations by personalised e-mail link and you can link responses to people; if you could work out who answered, do not call it anonymous, and never claim anonymity you cannot deliver. Protections in practice: pseudonyms, aggregated reporting, encryption, secure storage, and identifiers separated from responses early. reveals if they choose Privacy about the boundary itself the participant’s right to control what they reveal about themselves, and when no link — not even for the researcher Anonymity a property of the data not even the researcher can link a response to a person linked — but kept locked Confidentiality a property of the researcher’s conduct the researcher could identify respondents, but promises not to reveal identities MOST “ANONYMOUS” SURVEYS ARE ONLY CONFIDENTIAL Send invitations by personalised e-mail link and you can link responses to people. If you could work out who answered, do not call it anonymous — never claim anonymity you cannot deliver. PROTECTIONS IN PRACTICE pseudonyms aggregated reporting encryption secure storage identifiers separated from responses early

Deception

Deception: the case for and against, and the conditions that decide it Definition: deception is misleading participants about the purpose or nature of the research, by lying or by concealment. The case for it: for some questions, disclosure destroys the phenomenon; tell people you are studying honesty and they behave differently. The case against: it directly violates informed consent, because no one can consent to what has been hidden; hence a heavy burden of justification. It may be considered only if all three conditions hold: the question has real value and cannot be answered any other way; the risk of harm is minimal; participants are fully debriefed afterwards. Flip each condition and deception is indefensible: when an honest design would have answered the question; when it hides real risks from participants; when the deceived can never be told. Covert observation, watching people who do not know they are being studied, is deception’s close cousin and is judged by exactly the same test. Deception: misleading participants about the purpose or nature of the research — by lying, or by concealment THE CASE FOR IT For some questions, disclosure destroys the phenomenon: tell people you are studying honesty, and they behave differently. THE CASE AGAINST It directly violates informed consent: no one can consent to what has been hidden. Hence a heavy burden of justification. MAY BE CONSIDERED ONLY IF ALL THREE HOLD The question has real value and cannot be answered any other way The risk of harm is minimal Participants are fully debriefed afterwards INDEFENSIBLE WHEN An honest design would have answered the question It hides real risks from participants The deceived can never be told flip each condition COVERT OBSERVATION Watching people who do not know they are being studied: deception’s close cousin, judged by exactly the same test.

Cautionary tales and the rise of ethical review

Five cautionary tales

Five cautionary tales on a timeline, 1932 to 2014 A timeline from the 1930s to the 2010s: the studies that broke the rules and built the modern review system. Bars show the Tuskegee syphilis study, 1932 to 1972, and the Guatemala syphilis study, 1946 to 1948, unearthed only in 2010; both violated justice, because the powerless bore the risk. Points mark Milgram’s obedience experiment, 1961 (deception and severe distress), Tearoom Trade, 1970 (covert observation and privacy), and the Stanford Prison Experiment, 1971 (consent, and harm left to run). An arrow leads from the end of Tuskegee to the Belmont Report, 1979, and the modern system of independent ethical review. Last, the Facebook emotional contagion study, 2014: consent by checkbox, and the story is not over. The studies that broke the rules — and built the modern review system 1930 1940 1950 1960 1970 1980 1990 2000 2010 Tuskegee syphilis study, 1932–1972 Guatemala, 1946–1948 unearthed only in 2010 justice: the powerless bore the risk Milgram, 1961 deception; severe distress Stanford Prison, 1971 consent; harm left to run Tearoom Trade, 1970 covert observation; privacy Belmont Report, 1979 and the modern system of independent ethical review Facebook, 2014 consent by checkbox the story is not over

Milgram’s obedience experiment (1961)

Milgram’s obedience experiment, 1961: case file Case file: 1961, Stanley Milgram at Yale; aim: explore the limits of obedience to authority. Principles breached: deception and no harm. What was done: participants believed they were giving painful electric shocks to a learner in the next room, in fact an actor; many were visibly distressed, yet were urged to continue. The breach: deception about the study’s true nature, and intense psychological stress inflicted without warning; many learned something about themselves they had never agreed to learn. The consequence: lasting criticism of deceptive practice, one of the founding controversies of modern research ethics. The lesson: protect participants from severe distress even in simulated settings, and never let scientific value excuse the means by default. CASE FILE 1961 Stanley Milgram, Yale Aim: explore the limits of obedience to authority PRINCIPLES BREACHED Voluntary participation Informed consent No harm Privacy Deception Justice WHAT WAS DONE Participants believed they were giving painful electric shocks to a learner in the next room — in fact an actor. Many were visibly distressed, yet were urged to continue: “the experiment requires that you continue”. THE BREACH Deception about the study’s true nature, and intense psychological stress inflicted without warning: many learned something about themselves they had never agreed to learn. No harm Deception THE CONSEQUENCE Lasting criticism of deceptive practice — one of the founding controversies of modern research ethics. THE LESSON Protect participants from severe distress even in simulated settings, and never let scientific value excuse the means by default.

The Stanford Prison Experiment (1971)

The Stanford Prison Experiment, 1971: case file Case file: 1971, Philip Zimbardo at Stanford; aim: understand the psychological effects of perceived power. Principles breached: informed consent and no harm. What was done: students were randomly assigned as guards or prisoners in a mock prison for a planned two weeks; the guards became abusive and the prisoners broke down. The breach: consent to a role-play did not cover risks of this kind, and once the harm was visible the study was allowed to continue; the researcher had, in effect, joined the experiment. The consequence: halted after six days of a planned fourteen, and ended by an outside observer, not the researcher: one origin of the case for independent oversight. The lesson: the duty of care does not pause once the study is running; monitor well-being throughout, and stop the study when harm emerges. CASE FILE 1971 Philip Zimbardo, Stanford Aim: understand the psychological effects of perceived power PRINCIPLES BREACHED Voluntary participation Informed consent No harm Privacy Deception Justice WHAT WAS DONE Students were randomly assigned as guards or prisoners in a mock prison, for a planned two weeks. The guards became abusive; the prisoners broke down. THE BREACH Consent to a role-play did not cover risks of this kind — and once the harm was visible, the study was allowed to continue. The researcher had, in effect, joined the experiment. Informed consent No harm THE CONSEQUENCE Halted after six days of a planned fourteen — and ended by an outside observer, not the researcher: one origin of the case for independent oversight. THE LESSON The duty of care does not pause once the study is running: monitor well-being throughout, and stop the study when harm emerges.

Tuskegee and Guatemala

The Tuskegee and Guatemala syphilis studies: case file Case file: Tuskegee, 1932 to 1972, and Guatemala, 1946 to 1948, both involving the U.S. Public Health Service; aims: observe untreated syphilis, and study penicillin as a treatment. Principles breached: justice, informed consent, no harm and deception. What was done: in Tuskegee, African-American men were never told their diagnosis, only that they were being treated for bad blood, and were denied penicillin once it became the standard cure; in Guatemala, prisoners, soldiers and psychiatric patients were deliberately infected, without consent. The breach: vulnerable, powerless people bore all of the risk for none of the benefit, the precise violation of justice, with no consent, the truth withheld and treatment denied. The consequence: public outrage over Tuskegee led directly to the Belmont Report of 1979 and the modern ethics-review system; Guatemala was unearthed only in 2010, prompting formal U.S. government apologies. The lesson: never compromise on informed consent, above all with vulnerable populations, people chosen because they could not refuse or complain. CASE FILE Tuskegee 1932–1972 Guatemala 1946–1948 U.S. Public Health Service Aim: observe untreated syphilis; study penicillin as a treatment PRINCIPLES BREACHED Voluntary participation Informed consent No harm Privacy Deception Justice WHAT WAS DONE Tuskegee: African-American men were never told their diagnosis — only that they were being treated for “bad blood” — and were denied penicillin once it became the standard cure. Guatemala: prisoners, soldiers and psychiatric patients were deliberately infected, without consent. THE BREACH Vulnerable, powerless people bore all of the risk for none of the benefit — the precise violation of justice — with no consent, the truth withheld, and treatment denied. Informed consent No harm Deception Justice THE CONSEQUENCE Public outrage over Tuskegee led directly to the Belmont Report (1979) and the modern ethics-review system. Guatemala was unearthed only in 2010, prompting formal U.S. government apologies. THE LESSON Never compromise on informed consent, above all with vulnerable populations — people chosen because they could not refuse or complain.

Tearoom Trade (1970)

Tearoom Trade, 1970: case file Case file: 1970, Laud Humphreys; aim: study anonymous sexual encounters between men in public restrooms. Principles breached: deception and privacy. What was done: acting as a covert lookout, he observed the encounters, recorded car number plates, traced the men’s home addresses, and later interviewed them under a false pretext. The breach: deception and covert observation, then the linking of what was observed to identities, a profound violation of privacy. The stakes: when exposure could mean prison, unemployment and social ruin, his files held the means to destroy every man in them; their safety rested on his trustworthiness alone. The lesson: covert observation plus identifiable records compounds one violation with another, turning a passing intrusion into a permanent risk. At work: covert watching plus identifying records is exactly what some employee monitoring amounts to. CASE FILE 1970 Laud Humphreys Aim: study anonymous sexual encounters between men in public restrooms PRINCIPLES BREACHED Voluntary participation Informed consent No harm Privacy Deception Justice WHAT WAS DONE Acting as a covert lookout, he observed the encounters, recorded car number plates, traced the men’s home addresses — and later interviewed them under a false pretext. THE BREACH Deception and covert observation, then the linking of what was observed to identities: a profound violation of privacy. Privacy Deception THE STAKES When exposure could mean prison, unemployment and social ruin, his files held the means to destroy every man in them. Their safety rested on his trustworthiness alone. THE LESSON Covert observation plus identifiable records compounds one violation with another: a passing intrusion becomes a permanent risk. AT WORK Covert watching plus identifying records is exactly what some “employee monitoring” amounts to.

Facebook emotional contagion (2014)

The Facebook emotional contagion study, 2014: case file Case file: 2014, Facebook with researchers from Cornell and UCSF; aim: test whether emotional states spread through social networks. Principle breached: informed consent. What was done: the news feeds of about 700,000 users were deliberately manipulated, some seeing fewer positive posts and others fewer negative ones, and their own posting was measured. The breach: no one was asked and no one was told; the only cover was a terms-of-service clause that mentioned research. The consequence: public outrage, and a global debate about the boundaries of research on platforms. The lesson: digital settings do not suspend the rules; a terms-of-service checkbox is not informed consent. At work: A/B tests on customers and analytics on employees raise the same question, when does routine business practice become research on unwitting subjects? CASE FILE 2014 Facebook, with researchers from Cornell and UCSF Aim: test whether emotional states spread through social networks PRINCIPLES BREACHED Voluntary participation Informed consent No harm Privacy Deception Justice WHAT WAS DONE The news feeds of about 700,000 users were deliberately manipulated — some saw fewer positive posts, others fewer negative ones — and their own posting was measured. THE BREACH No one was asked, and no one was told: the only cover was a terms-of-service clause that mentioned “research”. Informed consent THE CONSEQUENCE Public outrage, and a global debate about the boundaries of research on platforms. THE LESSON Digital settings do not suspend the rules: a terms-of-service checkbox is not informed consent. AT WORK A/B tests on customers and analytics on employees raise the same question: when does routine business practice become research on unwitting subjects?

Research integrity

Research integrity: three cardinal sins, four positive duties, and conflicts of interest Ethics also governs how researchers treat the truth, not only how they treat participants. Three cardinal sins: fabrication, inventing data that were never collected; falsification, distorting data or cherry-picking what to report, where no single number is false yet the picture deceives; plagiarism, presenting other people’s work or words as your own. Four positive duties: preserve raw data, so that claims can be checked; report honestly, including the findings you wished you had not found; correct errors promptly and transparently; disclose conflicts of interest, so that readers can weigh the findings. At work, three conflicts of interest: a consultant paid by a sponsor who wants a particular answer; an internal analyst whose bonus depends on the initiative being evaluated; a manager evaluating their own pet project. The minimum response is to declare them: an undeclared conflict turns research into advocacy in research’s clothes. Ethics also governs how researchers treat the truth — not only how they treat participants THREE CARDINAL SINS Fabrication inventing data that were never collected Falsification distorting data, or cherry-picking what to report: no single number is false, yet the picture deceives Plagiarism presenting other people’s work or words as your own POSITIVE DUTIES Preserve raw data so that claims can be checked Report honestly including the findings you wished you had not found Correct errors promptly and transparently Disclose conflicts of interest so that readers can weigh the findings AT WORK: CONFLICTS OF INTEREST A consultant paid by a sponsor who wants a particular answer An internal analyst whose bonus depends on the initiative being evaluated A manager evaluating their own pet project The minimum response: declare them. An undeclared conflict turns research into advocacy in research’s clothes.

Ethics committees and review

Ethics committees and review: what they check, proportionate scrutiny, and what to do without one A flow: design the study, then independent ethical review by an ethics committee (in the US, an Institutional Review Board or IRB), then collect data, only after review. The committee asks four questions: risks, what are they and to whom; consent, is it informed and voluntary; data protection, how will data be kept safe; vulnerable groups, what extra safeguards. Proportionate scrutiny, shown as a scale from light touch to full review: a low-risk anonymous CAWI survey of adults about their shopping habits at the light end; covert observation, deception designs and research with children at the full end. Scrutiny is matched to the risk, not designed to obstruct. The deeper value: the discipline of justification, because writing the application forces you to confront your own design’s risks, and researchers routinely find their flaws at exactly that moment. No committee? Be the committee: organisations rarely have ethics committees, so the manager-researcher must ask the reviewer’s questions of their own design. Rule of thumb: if you would be uncomfortable defending it to a reviewer, or seeing it in the press, redesign it. Design the study INDEPENDENT ETHICAL REVIEW an ethics committee — in the US, an Institutional Review Board (IRB) Risks what are they, and to whom? Consent informed and voluntary? Data protection how will data be kept safe? Vulnerable groups what extra safeguards? Collect data only after review PROPORTIONATE SCRUTINY light touch full review A low-risk anonymous CAWI survey of adults about their shopping habits Covert observation, deception designs, research with children matched to the risk — not designed to obstruct THE DEEPER VALUE The discipline of justification Writing the application forces you to confront your own design’s risks — researchers routinely find their flaws at exactly that moment. NO COMMITTEE? BE THE COMMITTEE Organisations rarely have ethics committees, so the manager-researcher must ask the reviewer’s questions of their own design. Rule of thumb: uncomfortable defending it to a reviewer, or seeing it in the press? Redesign it.

Ethics in organisational research

The manager as researcher

The manager as researcher: power asymmetry and the principles it contaminates Power asymmetry: a manager above a row of subordinates, who are also the participants, with an arrow labelled with what the manager holds power over: pay, promotion, workload and continued employment. This contaminates every principle at once. Voluntariness: can they really refuse the person who writes their performance review? Honesty of data: will they tell a powerful person the truth, or what is safe to say? No harm: a careless finding can follow a named person through their career. Employees are therefore a vulnerable population, not because they lack the capacity to consent, but because both refusal and candour carry risks they did not choose; the coerced study is also a biased study. At Meridian: the board asks the team to survey staff about why colleagues quit, and the first design question is ethical, not statistical: who will ask, who will know, who could be hurt? POWER ASYMMETRY Manager pay promotion workload continued employment subordinates — and participants Voluntariness can they really refuse the person who writes their performance review? Honesty of data will they tell a powerful person the truth — or what is safe to say? No harm a careless finding can follow a named person through their career EMPLOYEES: A VULNERABLE POPULATION Not because they lack the capacity to consent, but because both refusal and candour carry risks they did not choose. The coerced study is also a biased study. AT MERIDIAN The board asks the team to survey staff about why colleagues quit. The first design question is ethical, not statistical: who will ask, who will know, who could be hurt?

Employee monitoring and covert observation

Employee monitoring and covert observation: may monitor is not may study, and the covert test The employer may monitor, but that does not mean the researcher may study. Lawful monitoring of email, chat and CCTV for security or operations, silently repurposed as research, bypasses consent entirely, because nobody was ever asked: purpose matters. The covert test: covert observation of employees is justifiable only if all four conditions hold: the question is important; no open design would work; harm is minimal; disclosure follows. The second gate is circled: most workplace questions fail here, because communication, morale and teamwork can almost always be studied openly. Remember Humphreys: covert watching plus identifiable records is the worst combination, and workplace monitoring is that combination by default. Test yourself: studying team communication by silently reading the team’s chat history; which principles does it breach, and could any safeguard fix it? THE EMPLOYER MAY MONITOR ≠ THE RESEARCHER MAY STUDY Lawful monitoring email, chat, CCTV for security or operations silently repurposed purpose matters Research bypasses consent entirely: nobody was ever asked THE COVERT TEST: JUSTIFIABLE ONLY IF ALL FOUR HOLD 1 The question is important 2 No open design would work 3 Harm is minimal 4 Disclosure follows Most workplace questions fail here: communication, morale and teamwork can almost always be studied openly. REMEMBER HUMPHREYS covert watching + identifiable records = the worst combination — and workplace monitoring is that combination by default TEST YOURSELF Studying “team communication” by silently reading the team’s chat history: which principles does it breach — and could any safeguard fix it?

Insider research

Insider research: each advantage casts an ethical shadow, and each shadow has a safeguard Insider research means studying the organisation you belong to: the manager, HR analyst or student researching their own workplace. Three columns. Access, what an outsider could never negotiate, casts the shadow of role confusion: is this a chat between colleagues or data collection, and do they know which? Safeguard: make the researcher role explicit and visible, so colleagues always know when the researcher hat is on. Context, an insider’s understanding of what things mean, casts the shadow of prior knowledge: you know things participants never consented to have used as data. Safeguard: draw consent boundaries, so only what is gathered in declared research counts, never the lunchtime confidence. Trust of colleagues who know you casts the shadow of the exit problem: the research ends, but you still work there, holding what you learned. Safeguard: agree what happens to the findings, who sees them and in what form, before collecting them. Underneath: colleagues’ friendliness is not consent; an insider must ask more formally than a stranger, not less. Insider research: studying the organisation you belong to — the manager, HR analyst or student researching their own workplace Access ADVANTAGE what an outsider could never negotiate Context ADVANTAGE an insider’s understanding of what things mean Trust ADVANTAGE of colleagues who know you Role confusion SHADOW is this a chat between colleagues, or data collection — and do they know which? Prior knowledge SHADOW you know things participants never consented to have used as data The exit problem SHADOW the research ends, but you still work there, holding what you learned SAFEGUARD Make the researcher role explicit and visible: colleagues always know when the researcher hat is on SAFEGUARD Draw consent boundaries: only what is gathered in declared research counts — never the lunchtime confidence SAFEGUARD Agree what happens to the findings — who sees them, in what form — before collecting them Colleagues’ friendliness is not consent: an insider must ask more formally than a stranger, not less

Confidentiality in small teams

Confidentiality in small teams: how a promise of anonymity breaks without any names The promise that breaks: responses are anonymous and reported only in aggregate. A manager who knows the team stands above a team of six, two of whom are marked: the two dissatisfied with management are identifiable in fact, with no names needed. Deductive disclosure, identification by combining attributes, drawn as a funnel: all staff, then crossed with department, then with tenure, then with role, leaves one person. At Meridian: the e-commerce unit has seven staff; reporting its engagement score separately identifies no one by name, and everyone in fact. THE PROMISE THAT BREAKS “Responses are anonymous and reported only in aggregate.” Manager knows the team ! ! a team of six: “the two dissatisfied with management” identifiable in fact — no names needed DEDUCTIVE DISCLOSURE identification by combining attributes all staff × department × tenure × role one person no single field identifies anyone — the combination does AT MERIDIAN The e-commerce unit has seven staff. Reporting its engagement score separately identifies no one by name — and everyone in fact.

Small teams: the safeguards

Small teams: three safeguards and the cardinal rule Three safeguards. One, minimum cell size: report only groups above a threshold, commonly n of 10 or more, and suppress or merge anything smaller; a group of ten is reported, a team of six is suppressed or merged. Two, coarsen categories: merge small departments, band tenure, and drop background questions you do not truly need; exact years of service become a few bands. Three, care with open text: never report comments verbatim with group labels attached, because prose style identifies people as surely as names; the group label on a quoted comment is struck out. The cardinal rule: never promise more protection than you can deliver. A broken promise of anonymity does double damage, to the people exposed and to every future study, because the organisation remembers. If you cannot guarantee anonymity, say confidential, and explain exactly who will see what. 1 Minimum cell size Report only groups above a threshold — commonly n ≥ 10 — and suppress or merge anything smaller. report suppress or merge 10 or more a team of six 2 Coarsen categories Merge small departments, band tenure, and drop background questions you do not truly need. exact years of service a few bands 3 Care with open text Never report comments verbatim with group labels attached: prose style identifies people as surely as names. “…” group label THE CARDINAL RULE Never promise more protection than you can deliver. A broken promise of anonymity does double damage — to the people exposed, and to every future study, because the organisation remembers. If you cannot guarantee anonymity, say “confidential”, and explain exactly who will see what.

The ethics of using customer data

The ethics of using customer data: a new purpose, three questions, and where analytics becomes research Customer data, such as transactions, loyalty cards, complaints and browsing: customers consented to a purchase, not to being research subjects, so re-use for research is a new purpose that needs its own justification, purpose limitation on the customer side. Three questions before touching customer data. One: can it run on anonymised or aggregated data? Usually yes, because most managerial questions concern patterns, not persons, and the problem largely dissolves. Two: would customers be surprised? Surprise, in the uncomfortable sense, is a good proxy for a consent boundary crossed. Three: does it create new risks for them, such as profiling that affects what they are offered, discrimination, or exposure of sensitive purchases? A spectrum: at one end, aggregate sales patterns, ordinary business intelligence; at the other, manipulating individual customers’ experience or studying identified individuals, which is human-subjects research, where the Facebook case sits. The label on the activity does not decide its ethics; the relationship to the people in the data does. CUSTOMER DATA transactions loyalty cards complaints browsing Customers consented to a purchase not to being research subjects A NEW PURPOSE Re-use for research needs its own justification: purpose limitation, on the customer side. THREE QUESTIONS BEFORE TOUCHING CUSTOMER DATA 1 Can it run on anonymised or aggregated data? Usually yes: most managerial questions concern patterns, not persons — and the problem largely dissolves. 2 Would customers be surprised? Surprise, in the uncomfortable sense, is a good proxy for a consent boundary crossed. 3 Does it create new risks for them? Profiling that affects what they are offered, discrimination, exposure of sensitive purchases. Aggregate sales patterns ordinary business intelligence Manipulating individual customers’ experience, or studying identified individuals: human-subjects research — the Facebook case sits here The label on the activity does not decide its ethics; the relationship to the people in the data does.

GDPR and organisational research

GDPR and organisational research: four ideas along the life of the data The life of personal data as a pipeline: collect, use, keep, delete. Over all of it, lawful basis: every use of personal data needs one, such as consent or legitimate interest; for employee data, consent is legally doubtful, because the power imbalance means it may not be freely given, so the law agrees with the ethics. Under collect, data minimisation: collect only what the question requires; if you do not need names, do not collect names. Under use, purpose limitation: data collected for one purpose cannot be freely re-used for another. Under keep and delete, storage limitation: keep data only as long as needed, then delete it, and know where it lives and who can access it. Anonymise early: truly anonymised data, with no way back to individuals, falls outside the GDPR altogether. For any study, one sentence each on the lawful basis, the purpose, what is collected, where it is stored and when it is deleted; if you cannot write them, the design is not finished. Collect Use Keep Delete LAWFUL BASIS — EVERY USE OF PERSONAL DATA NEEDS ONE Consent, legitimate interest, … For employee data, consent is legally doubtful: the power imbalance means it may not be freely given. The law agrees with the ethics. Data minimisation Collect only what the question requires: if you do not need names, do not collect names. Purpose limitation Data collected for one purpose cannot be freely re-used for another. Storage limitation Keep data only as long as needed, then delete it — and know where it lives and who can access it. ANONYMISE EARLY Truly anonymised data — with no way back to individuals — falls outside the GDPR altogether. FOR ANY STUDY, ONE SENTENCE EACH lawful basis purpose what is collected where it is stored when it is deleted cannot write them? the design is not finished

Ethics at Meridian: a checklist

Ethics at Meridian: five questions to settle before fieldwork At Meridian, before any fieldwork: the board wants evidence on turnover, up from 14% to 27% in two years, and on falling customer satisfaction. Five questions, each with a check. One: who is the researcher, and what power do they hold over participants? A line manager over their own team means redesign first. Two: is participation genuinely refusable, and invisible when refused? Check the three warning signs. Three: what is promised about anonymity or confidentiality, and can it survive small-team reporting? Count cell sizes before printing the consent text. Four: is existing data, such as HR records, chat logs or customer files, being repurposed, and on what basis? Purpose limitation, ethical and legal. Five: who sees the results, and at what level of aggregation? A finding harmless on the board’s desk can be harmful on a store manager’s. Settle every answer in the design, not improvised after the data arrive. Today’s exercise: five Meridian proposals, and your verdicts. AT MERIDIAN — BEFORE ANY FIELDWORK The board wants evidence on turnover, up from 14% to 27% in two years, and on falling customer satisfaction. 1 Who is the researcher — and what power do they hold over participants? a line manager over their own team? Redesign first 2 Is participation genuinely refusable — and invisible when refused? check the three warning signs 3 What is promised about anonymity or confidentiality — and can it survive small-team reporting? count cell sizes before printing the consent text 4 Is existing data — HR records, chat logs, customer files — being repurposed? On what basis? purpose limitation, ethical and legal 5 Who sees the results — and at what level of aggregation? harmless on the board’s desk, harmful on a store manager’s? Settle every answer in the design — not improvised after the data arrive. TODAY’S EXERCISE five Meridian proposals — your verdicts

Conclusion

Conclusion

  • Ethics protects participants, the quality of the data, and the credibility of research itself — the three stand or fall together
  • The core principles — voluntary participation, informed consent, no harm, privacy, and a near-prohibition on deception — were written in the aftermath of real scandals, from Milgram to Facebook
  • Independent review institutionalised those lessons; where no committee exists, the researcher must be the committee
  • Organisational research adds a distinctive twist: power. When the researcher signs the participants’ performance reviews, voluntariness, candour, and confidentiality are all at risk — and the safeguards must be designed in from the start
  • The habit to build: for every study, ask who could this harm, and what would make it defensible? — and if no safeguard works, be willing not to run the study
  • Questions and discussion are welcome

Exercise

Today’s exercise: Ethics verdicts

QR code linking to the exercise worksheet

bdstanley.netlify.app/social-research-methodology-4-exercise

Model answers: case A, the stockroom camera

Model answer for case A: The stockroom camera Model answer for case A, the stockroom camera: not permissible. At stake: covert observation, informed consent, privacy, no harm, purpose limitation and power asymmetry. Covert watching plus records linked to whoever was on shift is exactly the Humphreys combination; the contract’s line that the premises are monitored is lawful monitoring, not a licence to study, because purpose matters. Instead: treat suspected theft as a security matter, not as research; if the board wants to understand shrinkage, study stock handling openly, with nothing linked to named staff. The covert test fails: the question is arguably important, since shrinkage has doubled, but an open design would work, the harm is not minimal because footage is tied to named staff, and no disclosure is planned. What could go wrong: an assistant who happened to be on shift during a coded irregularity is treated as a suspect, on footage she never knew existed. A The stockroom camera NOT PERMISSIBLE AT STAKE covert observation informed consent privacy no harm purpose limitation power asymmetry Covert watching plus records linked to whoever was on shift: exactly the Humphreys combination. “The contract says the premises are monitored” is lawful monitoring, not a licence to study — purpose matters. WHAT COULD GO WRONG An assistant who happened to be on shift during a coded “irregularity” is treated as a suspect — on footage she never knew existed. INSTEAD Treat suspected theft as a security matter, not as research. If the board wants to understand shrinkage, study stock handling openly, with nothing linked to named staff. THE COVERT TEST: IT FAILS Important question? Arguably — shrinkage has doubled. No open design would work? No. Harm minimal? No: footage is tied to named staff. Disclosure follows? No: none is planned.

Model answers: case B, the chat-log study

Model answer for case B: The chat-log study Model answer for case B, the chat-log study: permissible with safeguards. At stake: informed consent, purpose limitation, privacy, covert observation, data minimisation and, arguably, no harm. It is the lecture’s own test case: messages written to do a job were never offered for research, and reporting patterns, not individuals, protects only the report, because the analyst still reads every message in full. Safeguards: announce the study and ask for consent; export only what the question needs; anonymise and aggregate before analysis, with codes, not names; report team-level patterns only. The cure for covert research is to make it overt: announced, consented to and anonymised before anyone reads it, the study is no longer covert at all. What could go wrong: a cashier’s private complaint about her store manager, read in full by the analyst, finds its way back to the manager. B The chat-log study PERMISSIBLE WITH SAFEGUARDS AT STAKE informed consent purpose limitation privacy covert observation data minimisation no harm (arguably) The lecture’s own test case. Messages written to do a job were never offered for research, and “patterns, not individuals” protects only the report: the analyst still reads every message in full. WHAT COULD GO WRONG A cashier’s private complaint about her store manager, read in full by the analyst, finds its way back to the manager. SAFEGUARDS Announce the study and ask for consent; export only what the question needs; anonymise and aggregate before analysis — codes, not names; report team-level patterns only. THE CURE FOR COVERT RESEARCH Make it overt. Announced, consented to and anonymised before anyone reads it, the study is no longer covert at all.

Model answers: case C, the manager’s own survey

Model answer for case C: The manager’s own survey Model answer for case C, the manager’s own survey: permissible with safeguards, a judgement call. At stake: voluntary participation, power asymmetry, confidentiality, no harm and a conflict of interest. Her 14 subordinates cannot freely refuse the person who writes their reviews, and named answers make candour risky, so the data will be biased as well as coerced; and she wants a particular answer. Safeguards: a neutral party, HR or an external researcher, invites and runs it; no named accounts; refusal invisible and without consequence; she sees only the overall store result. Also defensible: not permissible in this form, if the group argues that no safeguard makes refusal safe while she is both requester and subject. Without the words hope and expect, consent would still not be free: the constraint is the situation, not the wording, because she is still their manager and she reads named answers. What could go wrong: an assistant who rates her fairness low is followed up individually, and her shifts get worse. C The manager’s own survey PERMISSIBLE WITH SAFEGUARDS JUDGEMENT CALL AT STAKE voluntary participation power asymmetry confidentiality no harm conflict of interest Her 14 subordinates cannot freely refuse the person who writes their reviews, and named answers make candour risky: the data will be biased as well as coerced. And she wants a particular answer. WHAT COULD GO WRONG An assistant who rates her fairness low is “followed up individually” — and her shifts get worse. SAFEGUARDS A neutral party (HR or an external researcher) invites and runs it; no named accounts; refusal invisible and without consequence; she sees only the overall store result. ALSO DEFENSIBLE Not permissible in this form, if the group argues that no safeguard makes refusal safe while she is both requester and subject. WITHOUT “HOPE AND EXPECT”? Still not free. The constraint is the situation, not the wording: she is still their manager, and she reads named answers.

Model answers: case D, the recycled exit interviews

Model answer for case D: The recycled exit interviews Model answer for case D, the recycled exit interviews: permissible with safeguards, a judgement call. At stake: purpose limitation, informed consent, confidentiality, data minimisation and no harm. Leavers were told the interviews were confidential and for HR purposes only, so a board report with quotes is a new purpose and a broken promise; names and health disclosures are far more than the question needs. Safeguards: analyse anonymised transcripts, with names and health details removed, and report aggregated reasons for leaving only, with no verbatim quotes. Also defensible: not permissible, if the group holds that the original promise rules out any re-use without fresh consent. The surprise test: would the leavers be surprised to find their words in a board report? If so, a consent boundary is being crossed. What could go wrong: a leaver’s health disclosure, quoted anonymously, is recognised by the manager she criticised. D The recycled exit interviews PERMISSIBLE WITH SAFEGUARDS JUDGEMENT CALL AT STAKE purpose limitation informed consent confidentiality data minimisation no harm Leavers were told “confidential and for HR purposes only”: a board report with quotes is a new purpose and a broken promise. Names and health disclosures are far more than the question needs. WHAT COULD GO WRONG A leaver’s health disclosure, quoted “anonymously”, is recognised by the manager she criticised. SAFEGUARDS Analyse anonymised transcripts, with names and health details removed; report aggregated reasons for leaving only — no verbatim quotes. ALSO DEFENSIBLE Not permissible, if the group holds that the original promise rules out any re-use without fresh consent. THE SURPRISE TEST Would the leavers be surprised to find their words in a board report? If so, a consent boundary is being crossed.

Model answers: case E, the six-person team survey

Model answer for case E: The six-person team survey Model answer for case E, the six-person team survey: permissible with safeguards. At stake: anonymity versus confidentiality, power asymmetry, no harm and data minimisation. In a team of six, role and tenure identify people, so the agency can honestly promise only confidentiality; and the results, with verbatim comments, go to the very supervisor being rated. Safeguards: drop the role question and the tenure breakdown; report the team only merged into a larger unit, the minimum cell size; no verbatim comments. The cover message it needed: your answers are confidential, not anonymous; the agency sees them, no one at Meridian does; results are reported only with a larger group, and comments are never quoted. More damaging: a promise made and broken, which harms those exposed and every future study, because the organisation remembers. What could go wrong: the newest member is alone in her tenure band, and her critical comment lands, verbatim, on her supervisor’s desk. E The six-person team survey PERMISSIBLE WITH SAFEGUARDS AT STAKE anonymity vs confidentiality power asymmetry no harm data minimisation In a team of six, role and tenure identify people: the agency can honestly promise only confidentiality. And the results, with verbatim comments, go to the very supervisor being rated. WHAT COULD GO WRONG The newest member is alone in her tenure band; her critical comment lands, verbatim, on her supervisor’s desk. SAFEGUARDS Drop the role question and the tenure breakdown; report the team only merged into a larger unit (minimum cell size); no verbatim comments. THE COVER MESSAGE IT NEEDED “Your answers are confidential, not anonymous: the agency sees them, no one at Meridian does. Results are reported only with a larger group; comments are never quoted.” MORE DAMAGING? A promise made and broken. It harms those exposed and every future study, because the organisation remembers.

Model answers: the pattern check

Model answer to Task 3: which principles are at stake in which case A matrix of the eight principles from the worksheet against the five cases, one defensible reading; the class poll leaves this question open. Case A, the stockroom camera: informed consent, no harm, privacy, deception and covert observation, purpose limitation, and arguably power asymmetry; not permissible. Case B, the chat-log study: informed consent, privacy, covert observation, purpose limitation, data minimisation, and arguably no harm; permissible with safeguards. Case C, the manager’s own survey: voluntary participation, no harm, privacy and confidentiality, power asymmetry, and arguably data minimisation; permissible with safeguards, a judgement call. Case D, the exit interviews: informed consent, no harm, privacy and confidentiality, purpose limitation, data minimisation; permissible with safeguards, a judgement call. Case E, the six-person team: no harm, privacy and confidentiality, power asymmetry, data minimisation, and arguably informed consent and deception; permissible with safeguards. Totals: privacy, anonymity and confidentiality is at stake in all five cases, and is outlined; no harm in four plus one arguably; informed consent and data minimisation in three plus one; purpose limitation in three; deception and covert observation, and power asymmetry, in two plus one; voluntary participation in one. Voluntary participation Informed consent No harm Privacy, anonymity, confidentiality Deception and covert observation Power asymmetry Purpose limitation Data minimisation cases at stake A Stockroom camera not permissible B Chat-log study with safeguards C Manager’s own survey with safeguards D Exit interviews with safeguards E Six-person team with safeguards 1 3 + 1 4 + 1 5 2 + 1 2 + 1 3 3 + 1 verdict at stake arguably judgement call one defensible reading — the poll leaves this question open

Model answers: plenary

Model answers to two Task 4 plenary prompts Two cards. “It’s already in the contract”: the answer is purpose limitation. Lawful access is not ethical research use: data gathered, or monitoring allowed, for one purpose cannot be freely re-used for another, and no one was ever asked to be a research subject. Cases A, B and D all lean on lawful access: the employment contract in A, the company’s own platform in B, records HR already holds in D; none of them amounts to research use. If Meridian had an ethics committee, which proposals would reach it? A discussion point: each looks like routine business, A like security, B like analytics, C like a manager’s feedback, D like HR housekeeping and E like a staff survey, so probably few, if any. Review is most needed exactly where nobody calls it research; where no committee exists, the researcher must be the committee. “It’s already in the contract” PURPOSE LIMITATION lawful access … A the employment contract B the company’s own platform D records HR already holds ≠ research use Data gathered — or monitoring allowed — for one purpose cannot be freely re-used for another, and no one was ever asked to be a research subject. If Meridian had an ethics committee… DISCUSSION which would reach it? each looks like routine business: A security B analytics C a manager’s feedback D HR housekeeping E a staff survey Probably few, if any. Review is most needed exactly where nobody calls it research: where no committee exists, the researcher must be the committee.

Study guide

Full summary of this session, for revision: Research ethics

QR code linking to the session handout

bdstanley.netlify.app/social-research-methodology-4-handout