Objectivity, Values and Reflexivity

Nobody studies anything at random.

You are reading about sociology because something bothered you — an unfairness you noticed, a pattern you could not explain, a question somebody dismissed. Every research project in the history of the discipline began in something like that.

And then the training starts, and the training says: be objective. Which sounds, on the face of it, like an instruction to stop caring about the thing that made you interested in the first place.

This lesson is about what that instruction actually means, how much of it survives scrutiny, and what a working researcher — or a working reader — should do instead.

"Whose side are you on?"

A researcher spends two years studying a homelessness service in a mid-sized city. She is there most weeks. She knows the staff, the residents, the night shift, the man who sits by the door and will not come in.

At the end she writes it up, and she finds three things.

One: the service's own outcome figures overstate its success, because people who leave without notice are recorded as "exited to unknown destination" rather than as failures, and they are the majority.

Two: the strict rules on drinking, which the staff believe are what keep the place safe, are the single most common reason residents give for not coming back — and the residents are, she thinks, right that the rules do not do what the staff think they do.

Three: the staff are working impossible hours for poor pay in a building that should have been condemned, and they are the only reason anything works at all.

She presents it, and gets attacked from two directions in the same afternoon.

The service manager says she has betrayed people who trusted her, that she has handed ammunition to a council looking for reasons to cut the budget, and that she was never neutral — she arrived sympathetic to the residents and found what she came for.

An activist in the audience says the opposite: that she has gone native, that her account of the staff is sentimental, that describing exhausted workers with warmth is how institutional failures get excused, and that she was never neutral either.

Both of them are partly right. She did arrive sympathetic to the residents. She did come to like the staff. Neither of those facts, by itself, tells you whether anything in her three findings is true.

And she has no idea how to say that without sounding defensive.

The instruction to be objective contains a confusion, and it needs pulling apart.

There are at least three different things people mean by "objective", and they have wildly different prospects.

One: having no values. Impossible, and not obviously desirable. A researcher with no interest in anything would study nothing.

Two: not letting your values determine your findings. Difficult, partly achievable, and worth trying — this is the one that matters.

Three: the findings being checkable by people who do not share your values. Achievable, and it turns out to be the one that does the real work.

Most arguments about objectivity are people defending the third while being accused of failing at the first.

And there is a second confusion underneath. The instruction is usually aimed at the researcher's feelings. But the biases that most reliably distort research are not feelings at all. They are structural — who paid, what gets published, what was countable, which populations are easy to reach. A researcher can be entirely even-tempered and still produce systematically skewed knowledge, because the skew was built into the conditions of the work before they arrived.

Weber's answer, which is more careful than its reputation

Two German words, and the whole position sits in the gap between them.

Wertbeziehung — value-relevance. Values decide what is worth studying. Out of the infinite mass of things that happen, we select some as significant, and there is no way to make that selection except by reference to what we care about. Weber thought this was inescapable and said so plainly. There is no value-free way to choose a research question.

Wertfreiheit — value-freedom, or better, value-neutrality. Once the question is chosen, values must not determine the answer. The evidence decides. Whether the answer is the one you hoped for is not a relevant consideration.

These are not in tension. They apply at different stages. Weber's position is: care intensely about what you study, and hold yourself rigidly to the evidence about it.

He added a third claim, which is the sharpest. Science cannot tell you how to live. It can tell you what a policy will probably do, what it will cost, what else follows from it, and whether your stated goals are consistent with each other. It cannot tell you what to want. In "Science as a Vocation" (delivered 1917) he described the modern condition as a polytheism of values — irreconcilable ultimate commitments, "warring gods", between which no science can adjudicate. You choose, and you are responsible for the choice.

From which came his fierce rule about teaching: the lecture hall is not a platform. A professor addressing students who cannot answer back, and who need his signature, has an unfair advantage, and using it is a misuse of the position. The teacher's duty, he said, is to teach inconvenient facts — inconvenient to the students' party, and to his own.

The attack: value-freedom as a myth

Gouldner, 1962: the myth is doing work, and the work is not scientific.

Alvin Gouldner's essay "Anti-Minotaur: The Myth of a Value-Free Sociology" made an argument that has never been fully answered.

His claim was not that objectivity is impossible. It was that the doctrine of value-freedom had come to function as a professional convenience. It let sociologists take money from anyone without asking what it was for. It let them supply findings to clients while disclaiming responsibility for the use. It marked moral engagement as unprofessional, which conveniently excluded the people most likely to make trouble. And it allowed a discipline to describe injustices in a tone of studied indifference and call the indifference rigour.

"A value-free sociology is a myth that serves the interests of the sociologist," was roughly the charge — a doctrine whose institutional effects were the opposite of its stated purpose.

Gunnar Myrdal made a related argument from a different direction. Having produced An American Dilemma (1944), the enormous study of race in the United States, he wrote in Objectivity in Social Research (1969) that valuations enter at every stage and cannot be filtered out by good intentions. His remedy was procedural and it is still the best short answer available: state your value premises explicitly, at the start, in the text. Not to excuse them — to make them available for the reader to discount.

Becker, 1967: "Whose Side Are We On?"

Howard Becker's presidential address to the Society for the Study of Social Problems is the other essential text, and it is frequently quoted for the wrong sentence.

His argument: the demand that we take no sides is not satisfiable, because the question is always already answered by the situation. Research is conducted in a world arranged hierarchically, and that arrangement produces what he called the hierarchy of credibility — the assumption that people at the top of an institution have the right to define how things are, and that their account is the accurate one.

So a study that simply reports what officials say is not neutral. It has adopted the official account by default and will not be accused of bias, because bias is only noticed when the unofficial account is taken seriously. The appearance of neutrality is itself a position — the position of whoever is currently believed without effort.

Becker's conclusion is therefore not "abandon standards". It is: we take sides whether we like it or not, so we should be aware which side we are on, say so, and then apply the same standards of evidence we would apply if we were on the other one.

And Gouldner attacked Becker too, in "The Sociologist as Partisan" (1968) — arguing that the underdog stance had become its own sentimentality, a romance of the deviant that flattered the researcher and did the subjects no good. Being on the side of the subordinated is not a method either.

What "the same standards either way" actually looks like.

Return to the homelessness researcher.

She is sympathetic to the residents. So the discipline requires her to test the residents' account hardest. When residents say the drinking rules are why they do not come back, she has to ask whether that is a reason or a reasonable-sounding account offered after the fact — and go looking for evidence that would distinguish them. Do people who were never subject to the rule return at different rates? Do the same people say the same thing to staff? Did the pattern change when the rule changed?

She is also fond of the staff. So she has to test her own account of them. She thinks they are heroic. Would she have written that if she had not been given tea every week for two years? What would an unfriendly reading of the same observations look like?

She does not achieve neutrality by these means, and she is not trying to. She achieves something more useful: the reasoning is on the page, so that a reader who does not share her sympathies can check where it goes.

Reflexivity, and what it is not

Bourdieu's version is the demanding one, and it is not about your feelings.

Reflexivity means turning the tools of the analysis onto the analyst. Everyone in sociology says they do it. What they usually mean is a paragraph about their own biography.

Bourdieu thought that version was worthless, and called for something he named participant objectivation: objectifying the objectifier — subjecting the researcher's position to exactly the analysis the researcher applies to everyone else (see 5.8.1).

He identified three sources of distortion, and only the first is the one people usually confess to.

One — social origins. Class, gender, ethnicity, nationality, education. This is the one that gets the paragraph.

Two — position in the academic field. Where the researcher sits in the structure of their own discipline: which school they trained in, whose approval they need, what a publishable finding looks like this year, which rivals a result will annoy. Bourdieu thought this was more distorting than the first, and much less often admitted.

Three — the intellectualist bias, the deepest and the least visible. The researcher's relation to the world is that of someone contemplating it rather than being caught in it — with time, distance and no urgent stake. So practices that are lived under pressure get rendered as though the people performing them were solving an intellectual puzzle. He called the resulting error the scholastic fallacy: attributing to actors the theory that the observer needed in order to describe what they did.

And he insisted reflexivity is collective, not confessional. It is a property of a discipline that examines its own conditions of production, not a paragraph of personal disclosure. Narcissistic reflexivity — the researcher's autobiography as the main event — he regarded as the disease posing as the cure.

The reflexivity failure modes, in ascending order of harm.

The disclaimer. A biographical paragraph early in the text, never referred to again, discharging the obligation. It changes nothing about the analysis.

The competitive confession. Positionality as credential — establishing standing to speak rather than examining what the position made hard to see.

The pre-emptive defence. Naming a bias so that having named it, you are protected from the charge. Stating that you are sympathetic to your subjects is not a substitute for testing your sympathetic conclusions.

The universal solvent. "Everyone is positioned, therefore all findings are just perspectives." This is the slide from 7.1.1, and it ends the conversation. If it were true, it would apply to itself, and there would be no reason to prefer the reflexive researcher's account to anyone else's.

Useful reflexivity produces a change in the research: a question asked that would not have been, a claim weakened, a case sought out because it would be inconvenient. If nothing in the study is different, no reflexivity happened.

The biases that are not about you

Four structural sources, none of which is a feeling.

One — who paid. The funding effect is one of the better-documented findings in the study of research itself: across medicine, nutrition and tobacco, studies funded by an industry are substantially more likely to report conclusions favourable to that industry than independently funded studies of the same question. The mechanism is rarely fraud. It is the choice of comparison, the choice of outcome measure, the dose, the duration, the decision about which results are worth writing up, and — most powerfully — which studies get commissioned at all.

Two — what gets published. Journals prefer positive, surprising, clean findings. A study that finds nothing is harder to publish, so the published literature over-represents effects that are real-but-smaller or not-real-at-all. This is publication bias, and it is a property of the literature rather than of any study in it (see 7.8.2).

Three — what was countable. Official statistics record what an institution decided to record, for its own purposes (see 7.4.4). Unpaid work, informal employment, undocumented people, harms that no agency is responsible for — these are not measured badly, they are structurally invisible. A discipline that studies what is measured will find that the world is shaped like the measurements.

Four — who is reachable. Some populations are easy to survey and some are not. Students, employees of participating firms, users of a service, people with stable addresses and phones. The result is the WEIRD skew of the published literature (see 6.8.1), and its smaller domestic versions.

None of these is corrected by a researcher examining their conscience. They are corrected — partially — by procedures, disclosure and institutional design. Which is the clue to what objectivity actually is.

Objectivity is a property of a procedure and a community, not of a person.

This is the move that makes the whole argument tractable, and versions of it are held by Popper, by the critical realists and by Harding alike.

Nobody achieves the view from nowhere. What can be achieved is a set of arrangements under which a claim can be checked by people who do not share the claimant's commitments, and would enjoy showing them wrong.

That is what these things are for, and it is why they matter more than sincerity:

Making the reasoning inspectable — the data, the code, the interview schedule, the coding frame, the decisions taken and the ones rejected. The test is whether a hostile competent reader could retrace your steps.

Pre-registration — stating hypotheses and analysis plan before seeing the data, which removes the freedom to decide afterwards which analysis counts.

Adversarial collaboration — researchers who disagree designing the study together and agreeing in advance what result would count as what. It is rare and it is the strongest available instrument.

Negative case analysis — deliberately hunting for the cases your account cannot handle, and reporting them. The qualitative equivalent of a robustness check.

Declared interests and independence of publication — who funded it, and whether the funder could have stopped it.

Replication and re-analysis by strangers — the community-level version of all of the above.

On this account the homelessness researcher's answer to both her critics is the same. You are right that I am not neutral. Here is the interview schedule, here is the coding frame, here are the cases that did not fit, here is what the outcome data look like recoded three other ways, and here is what would have changed my mind. Check it. That is what objectivity in social research realistically consists of — and it is a great deal more useful than a claim to have had no opinions.

Two errors that come from getting this wrong in opposite directions.

The first: dismissing research by locating a motive. "That study was funded by the industry, so it is worthless." Funding is a reason for heightened scrutiny — it tells you where to look, at the comparison group, the outcome measure, the duration, the unreported arms. It is not itself a refutation. A finding is defeated by evidence, not by biography. Otherwise every study can be dismissed by anyone, and the person with the most motivated audience wins.

The second: treating declared values as a licence. "I am writing from a committed standpoint" does not lower the evidentiary bar; it raises it, because the reader now knows which way the errors would run. Declaration is a service to the reader, not a permit.

Both errors have the same structure: treating the researcher's position as a substitute for examining the argument. It is a shortcut in both directions, and it is what makes public argument about research so bad.

Because the alternative to this position is one of two dead ends, and both are common.

Dead end one: the pretence of no position. Research written as though produced by nobody, from nowhere, with the funder in small print and the choices invisible. It is not more objective; it has simply made its commitments harder to inspect. Becker's point stands — the account that requires no defending is usually the official one.

Dead end two: the collapse into position. Everything is a perspective, all research is politics by other means, and the only question is whose. This is self-defeating — it undermines the evidence used to argue for it, and in practice it hands the field to whoever has the loudest platform, because if nothing can be checked then nothing constrains anyone.

The workable position is narrow and it is unglamorous. Choose your question because you care. State what you are committed to. Design the study so that it could have come out the other way. Test hardest the conclusion you want. Show your working to people who would enjoy refuting you. Then let the argument be about the evidence, which is the only place it can actually be settled.

Three different things are called objectivity: having no values (impossible), not letting values decide the findings (partly achievable, and the goal), and producing claims that others can check (achievable, and where the real work happens).

Weber separated value-relevance from value-neutrality. Values must choose the question — there is no other way to choose one. Values must not decide the answer. And science cannot tell you how to live: between ultimate commitments — the warring gods — no evidence adjudicates. His rule for teaching was to present the facts inconvenient to one's own side.

Gouldner showed the doctrine doing institutional work: value-freedom as professional convenience, insulating sociologists from responsibility for the use of their findings. Myrdal's remedy — declare your value premises in the text. Becker showed that neutrality has a default setting: the hierarchy of credibility means the official account is the one believed without effort, so reporting it looks unbiased. And Gouldner attacked Becker in turn — sympathy for the underdog is not a method either.

Bourdieu's reflexivity is the demanding version: three biases — social origins, position in the academic field, and the intellectualist bias that produces the scholastic fallacy — examined collectively, not confessed autobiographically. Reflexivity that changes nothing in the study did not happen.

And the biases that most reliably distort research are structural, not emotional: who funded it, what gets published, what was countable, who was reachable. None is fixed by examining your conscience.

Which is why objectivity is best understood as a property of procedures and communities — inspectable reasoning, pre-registration, adversarial collaboration, negative cases, declared interests, replication by strangers. The test is whether a hostile competent reader could retrace your steps.

Value-relevance (Wertbeziehung) — values necessarily select what is worth studying.

Value-neutrality (Wertfreiheit) — values must not determine the findings once the question is set.

Polytheism of values / warring gods — Weber's claim that ultimate commitments conflict irreconcilably and science cannot choose between them.

Hierarchy of credibility — Becker's term for the assumption that those at the top of an institution have the right to define how things are; the reason official accounts look neutral.

Value premises — Myrdal's remedy: commitments stated explicitly in the text so readers can discount them.

Reflexivity — turning the analysis onto the analyst and the conditions of the research.

Participant objectivation — Bourdieu's demanding version: objectifying the objectifier.

Scholastic fallacy — attributing to actors the theoretical reasoning that the observer needed in order to describe them.

Funding effect — the documented tendency of industry-funded studies to reach industry-favourable conclusions, mostly through design and commissioning rather than fraud.

Publication bias — the literature's over-representation of positive and surprising findings.

Pre-registration — recording hypotheses and analysis plan before seeing the data.

Adversarial collaboration — opponents designing a study jointly and agreeing in advance what each result would mean.

Negative case analysis — deliberately seeking and reporting cases the account cannot handle.

One — find the default. Take any news report about a public institution — a school, a police force, a hospital. Ask whose account is being treated as the description of what happened, and whose is being treated as a claim requiring support. That difference is the hierarchy of credibility, visible in the wild.

Two — write your value premises. Take a question you would want to research and write three sentences stating what you already believe about the answer. Then write, for each, what evidence would change your mind. If you cannot answer the second part for any of them, you have found something important.

Three — do Bourdieu's second bias. Pick any academic field you know a little about and ask what kind of finding is currently rewarded in it. Then ask what kind of finding would be true but unpublishable there.

Four — run the hostile reader test. Take something you have written or argued and ask: could someone who disagreed with me retrace my reasoning and see exactly where I decided what? If the answer is no, the fix is not more conviction.

Five — grade a funded study. Find a study funded by an interested party. Do not stop at the funding. Check the comparison group, the outcome measure, the time period, and whether anything was measured and not reported. Either you will find the effect, or you will have learned that funding alone was not enough to condemn it.