Acknowledge What You Know — Analects, “On Governing”#

An account of science’s social conditions must begin by explaining how knowledge is warranted. Otherwise criticism can miss its object: discovering that a company funded a study becomes sufficient to invalidate its conclusion; discovering that a scholar treats people badly becomes sufficient to find a flaw in their proof. The reverse mistake makes technical achievement a guarantee of conduct, as though solving an equation entitled someone to explain away another person’s suffering.

A knowledge claim must be examined according to what it claims. A mathematical proof requires definitions, premises, inferences, and conclusions that connect. An experimental finding requires scrutiny of measurement, confounding conditions, and error. A historical judgment requires attention to provenance, dates, context, and corroboration. An engineering performance claim also needs to specify operating conditions, interfaces, and failure conditions. Evidence from different fields cannot simply substitute for one another. Even within a field, proof, estimation, explanation, prediction, and demonstrations of feasibility warrant different things.

Showing that a device works on a test bench initially supports its performance under those conditions. Extending the conclusion to another temperature, scale, or population requires explaining which reasons remain valid after conditions change and which need new tests. A model that fits past data still needs examination on new objects. Scientific language makes these connections intelligible: which assumptions are used, which quantities were not measured, which steps are approximate, and how far the conclusion reaches. A necessary technical term is more than a badge of membership. It can make an otherwise blurred distinction open to argument.

No individual, however, can personally repeat every check. Researchers use other people’s theorems, reagents, instruments, and databases. They depend on calibration, records, and maintenance performed by people they have never met. Even a single-author paper employs languages, methods, and objects already worked upon by many others. The reliability of modern research therefore includes organized trust: reasons to provisionally use others’ results while retaining the possibility of checking crucial points.

Trusting others’ work requires neither abandoning checks nor making everyone an expert in everything. Confidence can be graded. Mature conclusions supported through several approaches deserve greater initial trust; novel claims with consequential implications call for closer scrutiny. Trust in an expert should likewise remain within tasks for which they can explain their grounds. Someone who has repaired an instrument for ten years may detect its drift better than a newly appointed project leader. An expert in a statistical method may know less about the constraints on collecting a sample. Trustworthiness must be judged through particular work, rather than ordered solely by administrative rank.

In 2011, the OPERA collaboration reported an anomaly in neutrino flight times. CERN made the result public and invited wider scrutiny. Its June 2012 update reported measurements from four experiments consistent with the speed of light, indicating that the earlier anomaly could be attributed to a faulty timing-system component.² The significance cannot be exhausted by saying that science eventually corrects itself. Checking occurred because results were available for discussion, colleagues could propose different tests, and equipment, time, and organizational support allowed those tests to proceed. The object did not accommodate the researchers’ hopes for a startling discovery. Yet its resistance became apparent through human technical activity.

This example brings out a distinction within “social construction.” People establish measurement procedures, conceptual classifications, apparatus, and forms of argument under particular historical conditions. Inquiry proceeds through these mediations, but researchers cannot therefore arbitrarily determine its results. Thermometers are made by people; their readings do not invariably obey their makers. Suspect readings can prompt a change of instrument, a comparison of methods, or an examination of environmental conditions. Explaining how people establish tests also explains how results can in turn revise their judgments.

Social conditions also affect which questions have an opportunity to be researched. Whether a disease mechanism exists differs from whether funding to investigate it is available. Funding arrangements can affect which evidence appears sooner and which populations remain poorly documented, thereby changing the knowledge we possess. This gives criticism a more definite object than “everything is interests”: through topic selection, sampling, publication arrangements, or secrecy, which interest affects what, and at which stage? A further judgment about biased conclusions still requires examination of the data and methods.

Warrant and community recognition must also remain distinct. A conclusion can have substantial support before receiving much attention, or enjoy considerable prestige while retaining an undiscovered error. But unrecognized correctness is not an automatically visible status. Those who describe themselves as suppressed dissenters must still provide grounds others can examine. Distinguishing recognition from warrant makes room for neglected work while protecting inquiry from the romantic belief that minorities must be right.

This distinction matters particularly to readers outside a field. They can provisionally accept a judgment on the basis of a professional community’s overall evidence while asking about its scope and uncertainty. They can question conflicts of interest or review procedures without first refuting an entire discipline. Mature public trust knows where its confidence rests and what would prompt re-examination. It requires scientists to explain limitations honestly and communicators to avoid turning developing judgments into declarations that can never change.

People produce knowledge, yet knowledge makes demands that cannot be adjusted at will. People rely on communities for judgment, yet must retain routes back to evidence and reasons. Research’s social conditions can help us discern objects, or keep certain objects, data, and objections out of discussion. Access to materials, the organization of tests, and opportunities for different judgments to meet therefore matter to how knowledge is warranted. The equipment, time, and shared labour supporting these activities are the conditions we need to examine next.

The Exemplary Person Is Not Different by Nature, but Makes Good Use of Things — Xunzi, “Encouraging Learning”#

For a claim to survive such scrutiny, work beyond the paper cannot be omitted. The paper is often only the final part of research that can circulate. Before it may come sample preservation, field contacts, software maintenance, animal care, literature organization, equipment cleaning, records of failed experiments, and teaching someone to notice an inconspicuous anomaly. There is also work absent from methods sections: purchasing, reimbursement, scheduling, covering for an ill colleague, and coordinating access to a machine.

These activities do not play the same role as proposing a theory, but they help determine whether a theory can be tested. If the provenance of data is lost, elegant analysis cannot recover it. If an experienced technician leaves without a successor, an expensive instrument may retain little beyond its price. Research’s material conditions include money and equipment, but also capabilities deposited in people’s hands, records, and ways of cooperating.

In Xunzi, making good use of things follows examples of height, wind, vehicles, and boats: conditions can extend human action. Research similarly draws upon enabling conditions, though using equipment and cooperating with people involve different relations. Equipment can be put to use; cooperation requires a division of work, commitments, and responsibilities among people. An organization can give someone capacities far beyond those they possess alone. It can constrain an individual while also making previously inaccessible questions researchable.

Resources both make work possible and select the work that can proceed. The first effect is readily visible: access to a microscope or archive, sufficient computing time, or training without the immediate need to earn a living. The second appears in funding conditions and completion requirements: only certain topics qualify, only certain results count as completion, and contributions visible within a term of office enter plans more readily. Nobody need directly order researchers to abandon other questions. Before writing an application, they may exclude problems with uncertain near-term returns, indivisible outputs, or no suitable reviewers. Resource constraints thereby enter the formulation of questions.

This does not make a funder’s purposes inherently illegitimate. Public agencies and businesses have limited budgets, and researchers cannot invoke freedom to demand unlimited use of others’ labour. What needs explaining is the relation between funding purposes, room for exploration, and responsibility for results. A specified engineering task and basic research whose destination is unknown need different timescales and forms of assessment. Asking people to explore the unknown while guaranteeing their discoveries in advance passes an institutional contradiction to applicants.

Consider two kinds of research. One has clear specifications and must demonstrate that a device meets them by a deadline. Another investigates an uncertain mechanism; presupposing the answer would undermine the inquiry. Both require accountability, but for different things. The first must explain performance and how failures are handled. The second must explain why a question matters, whether the design can distinguish competing explanations, and how changed evidence will alter the approach. The same short-term output sheet can misjudge both.

Tasks in shared research often have to wait upon one another. A programmer’s early completion need not bring a whole project forward; the bottleneck may be sample collection or an unresolved theoretical problem. Some people must wait, while others sustain foundations that produce little conspicuous output for extended periods. If assessment recognizes only achievements each person can display separately, the organization needs cooperation while members need contributions divisible into individually redeemable units. Interdependence in the project may fail to connect with individual accounting in assessment.

Waste must be distinguished from reserve capacity. Pointless duplicate reporting takes research time; preserving samples, records, and technical capabilities with no immediate use may enable later correction. Exploration also needs room to fail. Treating everything that does not become a paper as inefficiency encourages work confined to what is readily confirmed and published. Conversely, calling every delay basic research makes public accountability impossible. The question is whether the investment has an intelligible purpose, what has been learned, and which findings would change the direction.

The treatment of failure is especially revealing of how an organization distributes research risk. A hypothesis not supported by a well-designed experiment may yield an important constraint for the team. For a researcher on a short contract, that same result may mean having no acceptable output to submit. Epistemic gain and career loss diverge: the team can continue using what was learned, while the individual may receive no renewal because of it. The risk of an unsuccessful inquiry can become that person’s livelihood risk.

An organization that encourages risky inquiry needs ways to share its risks: room in budgets to change course, recognition of informative negative findings, training that survives the failure of one project, and staffing that does not push all uncertainty onto the shortest contracts. Such measures cost money and can be misused. But celebrating exploration while leaving the explorer’s life unsupported after failure means that weaker participants carry much of the risk.

Research agendas therefore require attention at three levels: the difficulties presented by the object, the organization’s division of those difficulties into tasks, and the people society enables to bear the uncertainty. An object’s difficulty may be unavoidable; labour allocation involves choices. An instrument requiring continuous operation does not determine that the same student must take every night shift. An immovable field season does not determine that researchers’ families alone must absorb its care costs.

Only with these distinctions does “research is hard” become a discussable judgment. Which difficulties belong to the object? Which can be reduced technically? Which arise from failed coordination, and which from costs transferred by organizations to individuals? Recognizing difficulty leaves these questions open and gives us reason to pursue them.

Those Who Labour with Their Minds Govern; Those Who Labour with Their Strength Are Governed — Mencius, “Teng Wen Gong I”#

This passage in Mencius belongs to a debate with Chen Xiang over Xu Xing’s teachings. It moves from the impossibility of one person performing every craft to a justification of intellectual and physical divisions of labour and political order. Division of labour may be necessary; its conversion into rule still requires further reasons. Modern research also needs to scrutinize that rapid transition.

Conception, execution, and maintenance frequently overlap in practice. A task initially assigned as repetitive operation can change an entire explanation when its operator notices an anomaly. A theoretical judgment can arise from discrimination developed through sustained contact with materials. A fixed hierarchy between those who think and those who work with their hands decides who can count as a contributor before contributions have occurred.

Once work has been divided, allocating shared results still requires answers to four questions: who did what, who receives public recognition, who controls subsequent use, and how benefits are distributed. Authorship addresses part of this; wages, patents, database access, and project authority address other parts. A name on a paper does not necessarily confer permission to take materials away or continue using equipment. Higher pay does not automatically replace an appropriate record of contribution. Different rights address different relations; “you have already been looked after” cannot settle them all.

Ross and colleagues’ 2022 study connected research-team personnel with credited outputs and combined that analysis with surveys and qualitative responses. Women in the study were less likely to receive authorship or patent credit for their teams’ outputs.³ Differences in credited names therefore cannot simply be assumed to represent differences in actual work. The finding does not, conversely, establish that every omission has the same discriminatory cause. Specific disputes still require attention to contributions, negotiations, and applicable rules.

Contribution itself is not a natural substance awaiting a scale. The significance of software maintenance, correction of a mistaken hypothesis, research design, and funding coordination requires shared judgment. Research achievements are often not simply additive: without essential materials, other work may produce nothing; without theoretical organization, materials may remain a heap. We need to examine how each activity contributed to the result and explain how different contributions are compared. Author order contains judgments; it is not a queue naturally formed by objective contribution.

CRediT describes research contributions through fourteen roles, including data curation, software, conceptualization, validation, and writing. Its official account also makes clear that the taxonomy does not itself determine authorship.⁴ That boundary is useful. Detailed records can name previously obscure labour; authority over the resulting list and the handling of disputes still require organizational arrangements. If a contributor must accept unilateral revisions by the project leader, a detailed form may merely transcribe existing power more neatly.

For contribution records to constrain allocation, prior agreements must remain open to later revision. Stating principles for authorship and the use of results at the start can reduce conflicting retrospective claims. Yet contributions change as research develops, and early agreements should not freeze out work that subsequently occurs. Records, confirmation by contributors, and independent handling of disputes allow agreements to follow contributions. If agreements cannot change with actual contributions, signing earlier may fix an original position prematurely.

Ownership is more complicated still. Whether publicly funded knowledge should be open, industry collaborations may remain confidential, or participant data require protection cannot be settled by authors’ personal freedom alone. Openness can extend testing and reuse, but can also expose participants or communities. Confidentiality can protect privacy or commercial advantage. These reasons need to be distinguished rather than packed into the phrase “intellectual property” and left there.

Custody of data, permission to use results, and recognition of contribution likewise concern different relations. A departing researcher may lack permission to copy sensitive data without thereby losing credit for organizing them. Authorship does not confer an unrestricted right to remove shared materials. Fair arrangements make departure compatible with handover and protection compatible with explanation, rather than allowing every restriction to lead back to one person’s final discretion.

Allocation has consequences beyond the present result. Those already credited are more likely to be invited to present work, collaborate, or seek resources. Unrecorded contributions struggle to become credentials in the next competition. An authorship decision can therefore influence who later has opportunities to perform work considered important. Past records enter the future through invitations, review, and allocation: these are the routes through which accumulation acquires force. Sustained, reliable contribution merits trust. What still needs asking is whether new evidence can change established judgments and whether opportunity travels only through existing lists.

Project leaders also perform real work. Securing resources, integrating questions, and accepting compliance and delivery responsibilities may be indispensable. Criticism of concentrated power cannot be achieved by describing all organizational labour as parasitic. The relevant questions concern what remuneration, recognition, and authority that contribution merits; how authority is limited; and who can review a leader’s decision when the leader is also a party to the dispute. Acknowledging contribution remains far from granting unlimited power.

These distinctions make deprivation more definite: labour goes unrecognized, control over results becomes detached from contribution, weaker members carry risks alone, or someone is indefinitely confined to tasks that do not develop portable capabilities. These need different remedies. Another author credit may not change working hours; a larger allowance may not restore opportunities to judge. Specifying what has been lost identifies which relation needs to change.

Learning Alone, without Companions, Leaves One Narrow and Poorly Informed — Book of Rites, “Record of Learning”#

At its best, an academic community appears as people who criticize one another while pursuing understanding together. This image preserves something worth maintaining. Companions can pose questions one person has overlooked; local work can acquire significance through others’ inquiries; younger researchers can enter accumulated knowledge instead of inventing a discipline anew.

To establish who makes these allocation decisions, we need to look inside the community. Laboratories organize daily work, universities arrange training and employment, degree-granting bodies decide qualifications, journals select work for publication, societies sustain professional exchange, and funders distribute some resources. Researchers also inhabit collaborations across institutions. These organizations and networks connect, but their membership, authority, and responsibility do not coincide. Rejection by a journal differs from termination of training by a university; praise from peers does not resolve problems in an employment relationship.

“The academic community believes” therefore needs unpacking. It may mean that several teams obtained similar findings, or simply that one review committee made a decision. It may identify a relatively stable disciplinary understanding, or a small circle’s preference. Specifying which community reached a judgment and through which procedures helps keep recognition proportionate to its grounds.

Communities take historical forms. In the American postwar policy debate, Vannevar Bush’s 1945 Science, the Endless Frontier advocated public support for basic research and scientific training while preserving institutions’ freedom of inquiry.⁵ This was a situated institutional proposal: research freedom needs resources, and resources connect it to national purposes and public accountability. The report is no common origin for every country or discipline. It does show that arrangements now treated as natural once required argument and political commitment.

Different requirements for equipment, staffing, and duration enable different organizational forms. Expensive facilities need coordination sustained beyond one person’s enthusiasm. Research based on public texts or proofs may support more dispersed collaboration. Fieldwork depends on communities, trust, and long-term presence. Such differences affect access, control over irreplaceable conditions, and what it actually means to leave an organization and continue researching.

With different research conditions, assessing communities requires more than asking whether they oppress or liberate. Do they preserve transmissible knowledge? Train newcomers? Enable errors to be found and those who find them to be heard? Sustain long-term questions? These are practical functions of organized cooperation. Where knowledge lives only in a few private memories and departing members lose every connection, gathering people together may still fail to establish shared work that can continue.

Professional standards likewise have more than one possibility. Shared terms, reporting formats, and testing methods help strangers cooperate and reduce repeated explanation. But standards fixed as the sole recognizable form of thought can exclude questions not yet expressible in an established language. The tension requires asking what a standard protects. Reporting sample provenance protects the grounds of judgment; forcing every question into a fashionable model may protect reviewers’ familiarity. Both are called norms, yet their effects differ substantially.

Politeness, ritual, and informal exchange also affect entry into cooperation. A conversation before a talk may help a newcomer understand it; introductions by acquaintances can lower the cost of establishing trust. Yet if important information is available only at private dinners, those unable to attend fall behind before formal competition begins. Public recruitment can coexist with restricted information. Fairness concerns both the final measuring standard and when people learn what that standard is.

Some people profoundly affected by research remain outside the community’s formal membership. Participants, patients, database contributors, and people supplying field knowledge may lack academic credentials and knowledge of every technical detail, yet know consequences researchers overlook. Counting only authors and reviewers as people entitled to ask questions creates blind spots beyond professional checks. Participation should follow the kind of question: discerning effects on everyday life and deciding methodological details require different experience and responsibilities.

A community capable of continuing to learn needs both concentrated technical expertise and the ability to explain decisions beyond itself. The first protects complex problems from being flattened by immediate opinion; the second prevents professional language from shielding decisions against accountability. Their friction cannot be resolved once and for all by slogans of complete autonomy or comprehensive oversight.

A firmer demand can therefore replace the comforting image alone: cooperation should give members capacities unavailable in isolation while making those capacities increasingly understandable, usable, and transmissible. A community’s maturity lies partly in its outstanding people, and partly in whether knowledge and judgment can survive any one person’s departure.

The Exemplary Person Seeks Harmony without Sameness — Analects, “Zi Lu”#

Although these organizations have different responsibilities, each needs usable judgments about the matters it decides. Journals cannot postpone decisions forever, projects cannot wait for every disagreement to end, and clinical or engineering practice often requires choices under uncertainty. Consensus has practical value because members can draw upon accumulated evidence, methodological experience, and professional division of labour without repeating every test from the beginning.

Shared judgment can nevertheless emerge through very different processes. Several mutually supporting lines of evidence, with disagreements open to examination, can reduce duplicated checking. Shared dependence on one authority may add little independent support, however numerous the adherents. Ten people repeating one unverified document do not provide the same epistemic weight as ten teams examining a claim through different approaches. How much trust consensus deserves depends on the process through which it formed.

Standing to participate in discussion must accordingly be distinguished from an argument’s credibility. Human dignity does not vary with expertise; arguments can nevertheless differ greatly in quality. Allowing newcomers to object does not require equating guesses made without basic methods with extensively tested conclusions. Equality should make reasons expressible, errors explainable, and judgments reviewable. Credibility remains answerable to evidence and inference.

Beyond methods and evidence, disagreements sometimes concern what counts as a problem or a satisfactory explanation. One team values predictive accuracy; another seeks an intelligible mechanism. One wants a rigorous general proof; another needs something workable under particular conditions. These orientations can each be useful and may compete over the same object. Simply calling one insufficiently scientific hides a disputed aim inside a definition.

Shared exemplars, methods, and ways of organizing problems concentrate inquiry while influencing how anomalies are noticed. An unexpected finding may prompt researchers to check measurements, revise auxiliary assumptions, or question the framework itself. The appropriate response depends on existing support, the anomaly’s persistence, and the strengths of alternatives. One anomaly does not automatically overturn a framework; past success does not license shelving anomalies indefinitely. Knowledge develops through reasoned retention and revision.

Criticism of a prevailing framework must therefore also make further inquiry intelligible. Where does the old explanation encounter difficulty? What distinctions does the new question make possible? How can their gains and losses be compared? Competing orientations may lack a predetermined, wholly neutral overall scale. They can still be compared in explanatory reach, consistency, responsiveness to new materials, and practical testing. Standards of comparison may themselves change through research. Giving reasons for those changes allows comparison to continue.

Differences in ability cannot provide a permanent reason to withhold explanation. A researcher unable to teach every discussion from first principles can identify prerequisites, existing answers, or suitable learning routes. An organization that indefinitely requires subordinates to execute tasks while denying them access to reasons has less ground for citing their lack of judgment as vindication of its arrangements. Training responsibilities and discussion efficiency require coordination that “you do not understand” cannot postpone forever.

Objections also do different work. One identifies a calculation error, another questions whether a sample can answer the question, and a third argues that an answerable question does not merit so much public investment. Recalculation may resolve the first; methods and objects must be discussed in the second; purposes and opportunity costs enter the third. Calling all three academic disputes assigns value decisions to professional committees without adequate representation. Putting all three to a public vote could instead eliminate technical scrutiny.

Once questions are distinguished, someone must handle them in an appropriate setting. Relevant expertise should carry technical judgment and record its reasons. Decisions about resource purposes, participants’ rights, and social consequences require corresponding public procedures. These procedures must also communicate. Public concern about a risk may require researchers to measure an overlooked consequence, and those measurements may then alter public debate. Participants have distinct domains of judgment while needing to let others’ questions enter their work.

This also identifies when expertise overreaches. Explaining a proposal’s possible consequences supplies reasons for a decision. Moving directly from “this can increase total output” to “therefore these people should bear its costs” adds a distributive judgment. That judgment may be defensible, but it needs reasons for the allocation, participation by affected people, and consideration of alternatives. Technical competence can clarify a choice without erasing its value conflicts.

Scientism tempts us to omit this step. A locally successful form of knowledge appears sufficient to prescribe the purposes of life as a whole. Sometimes only what can be quantified is treated as real; sometimes a model’s optimal solution relegates human objections to emotion. Yet optimality is always relative to goals, constraints, and weights. How goals are chosen, which losses are permitted, and whose interests enter the calculation do not become legitimate merely because the calculation is accurate.

Criticism of scientism should equally avoid treating accuracy as an enemy. Reliable measurements of pollution, disease, or production processes may finally make denied harms publicly discussable. Numbers can reveal as well as conceal; professional language can supply shared evidence to weaker parties as well as silence them. The questions concern how a number was obtained, what it represents and omits, and who can use it to make a claim.

Harmony without sameness here means shared scrutiny that still accommodates different judgments. It does not require endless disagreement or an unlimited range of research directions. An organization with finite resources can make provisional decisions. Those decisions should not become permanent verdicts on dissenters’ qualifications. Whether suspended proposals remain recorded, whether changed conditions allow reconsideration, and whether dissenters can continue working with others determines whether disagreement serves as a resource or a stain on identity.

The heroic story of dissent also needs scrutiny. Solitary persistence may arise from insight or refusal of counterevidence; majority agreement may arise from extensive knowledge or conformity. Numbers, loneliness, and suffering cannot replace reasons. Institutions protecting dissent need not declare dissenters right in advance. They should allow potentially sound objections to be tested regardless of the objector’s position while enabling unfounded claims to be rejected for stated reasons.

Trust formed this way does not require permanent cordiality. Sharp criticism can serve a shared object, while gentle language conceals an unchallengeable verdict. Healthy discussion depends on what criticism addresses, how others respond, whether errors can be corrected, and whether status terminates argument. A community able to say a serious “no” may sustain cooperation better than one that agrees with everything.

If Names Are Not Right, Speech Does Not Follow — Analects, “Zi Lu”#

Consensus helps people judge research, but distributing positions and funding also requires comparison. Time, positions, and money are finite: organizations must choose which work to support and who will undertake it. Assessment therefore has a particular use. Which decision does it serve? What needs to be known? How much work will assessors devote to understanding the work they judge?

Publication counts, citations, journal metrics, and grant totals attract attention because they make complex, difficult-to-compare activities aggregable. Organizations can rank quickly; applicants can more readily identify gaps. Public numerical standards may even improve on decisions made entirely through personal connections. Defects in metrics do not make an environment without them naturally fair.

The difficulty is the distance between a record and what it is intended to represent. Publication counts register public outputs without directly recording what each resolved. Citations indicate how often work is mentioned in subsequent literature, but may reflect criticism, reviews, or field size. A journal’s aggregate statistics cannot simply become the quality of every paper or author within it. Using numbers still requires saying what they actually record and why they warrant a further judgment.

The general recommendation of the San Francisco Declaration on Research Assessment, DORA, is to avoid using journal metrics as proxies for individual paper quality, personal contributions, or appointment and funding decisions. It also advocates attention to outputs such as datasets and software.⁶ A declaration states a normative direction; an institution’s signature alone cannot establish a change in practice. That requires examining what was subsequently read, how decisions were made, and who acquired different opportunities.

Once metrics become targets, they also organize behaviour. Imagine a publication count initially used as a rough indication of activity becoming a hard renewal threshold. Researchers then have reasons to favour publishable tasks, shorten exploration with uncertain returns, and split integrated work into separately creditable parts. Not everyone will respond identically, and division is not always inappropriate. The arrangement nevertheless increases the rewards for certain strategies, making it inadequate to attribute the resulting behaviour entirely to a lack of personal virtue.

In this imagined case, the organization first selects visible records and connects them to renewal and resources. Participants adjust their work; work meeting the threshold is more likely to remain, while other work becomes harder to sustain. The achievements and successful researchers later visible have already been filtered through the standard. It appears merely to discover excellence while also influencing which work is likely to occur. Judging this feedback requires comparing the behaviour encouraged with research purposes, rather than checking only whether the indicator rose.

Whether a record can support a judgment also depends on its distance from the assessed aim. In a calibration task required to meet a specified error tolerance, that tolerance is itself central. Posing questions, maintaining shared foundations, training newcomers, and correcting important mistakes contain values that cannot readily substitute for one another. Compressing them into one score makes the choice of weights determine how those values are balanced. Weights can be publicly discussed; they cannot present themselves as natural constants nobody chose.

Assessment also creates temporal bias. Easily counted work becomes visible sooner; infrastructure and training take longer to show effects. A software tool may quietly support research for years before becoming a prerequisite for many projects. A fashionable direction may quickly lose significance. Organizations settling accounts at one moment reward work matching their cycle. Longer cycles are no universal remedy, but may allow previously invisible contributions to enter judgment.

Representative-work assessment and peer judgment are common alternatives to numerical predominance. They can bring decisions closer to content while increasing reading burdens, discretion, and opportunities for conflicts of interest. Without sufficient time, relevant expertise, and requirements to explain reasons, assessing quality can revert to assessing authors, institutions, and familiarity. Changing the scale requires supplying the work and constraints the new scale needs.

Transparency has limits of its own. Publishing an algorithm can clarify assessment and make gaming more precise. Publishing review comments can support improvement while creating privacy or retaliation risks. The aim is not indiscriminate exposure. People affected by decisions should know the key reasons, authorized reviewers should have necessary materials, and decision-makers should carry corresponding responsibility.

Assessment can also move from a standard for allocating work into a standard for describing oneself. A person first learns to present achievements to secure a position. Repeated applications, evaluation, and comparison may gradually make the same records their only answer to “what have I done?” Reading, maintenance, teaching, and failed attempts without papers become apparently unproductive days as well as gaps in a CV. Organizational accounting begins to shape how someone remembers work and even how they look back upon that period of life.

Changing that self-description requires assessment to accommodate the distinct reasons different kinds of work matter. Papers remain important and should be discussed through their specific contributions. Training deserves recognition without being reduced to a headcount. Infrastructure maintenance should count, with an account of what was actually maintained. Several descriptions are not a device for finding everyone something flattering. They allow different work to receive serious judgment according to its objects.

Assessment systems themselves need assessment. Do they enable better-grounded decisions? Encourage the work they profess to value? Generate new unpaid labour or exclusions? If indicators rise while these questions cannot be asked, assessment has ceased to help an organization understand itself and has become a barrier against such understanding.

Lead Them through Government, Regulate Them through Punishment — Analects, “On Governing”#

Once assessment affects resources and qualifications, who controls it cannot be passed over. Concentrating several powers in one relationship creates particular difficulties. When one person controls the topic, funding, data access, authorship, references, and training assessment, a student objecting about one matter may fear losing every other source of support. Anticipated consequences can change what the student will say and how they say it, even without an explicit threat.

A student needs a supervisor’s knowledge. When the organization also places many career and livelihood gateways in the same hands, professional dependence has conditions for becoming organizational subordination. An asymmetry initially concerning knowledge and ability acquires, through resource allocation and judgments of qualification, sanctions extending far beyond technical matters. Criticizing this concentration need not first prove abuse on every occasion. Rights need stable safeguards, rather than recurring reliance on the goodwill of whoever holds power.

Constraining these powers requires distinguishing their sources. Knowing how an experiment works supplies professional reasons; allocating funds rests on organizational authorization; setting training requirements is constrained by educational rules and scholarly standards; judging character requires identifiable conduct and grounds. Professional competence helps with some duties without replacing authorization, procedures, or evidence of conduct. Calling all of them supervisory authority instead conceals their separate responsibilities within a single title.

China’s Academic Degrees Law, effective from 1 January 2025, assigns decisions to award, refuse, or revoke degrees, among other responsibilities, to academic degree evaluation committees. Article 28 provides for decisions on master’s and doctoral degrees after consideration of defence-committee resolutions and review of applications. The law also provides a route to request academic reconsideration of relevant academic evaluations.⁷ These provisions mean that a degree decision cannot simply be equated with a supervisor’s personal permission. How procedures work at a particular institution requires examination of its rules and implementation.

Formal rules and practical control may be separated by many stages. Who decides a thesis is ready for review? Who holds essential data? When are references submitted? Can someone changing supervisors retain research conditions? If entry points lack alternative routes, an ostensibly collective final decision may depend heavily on one person’s upstream judgment. Actual power must be traced through the process, not just located at the final signature.

Moving from a provision for reconsideration to making an actual application requires further conditions. Applicants need procedural knowledge, access to records, time to wait, and somewhere to continue learning or working. If livelihood and research conditions disappear before reconsideration is complete, even a favourable decision may not restore opportunities lost in the meantime. Judging usability must include the person’s circumstances while waiting.

Independent handling also requires more than a different name. Does the reviewer share interests with the original decision-maker, depend on their reference or cooperation, have access to complete records, or possess authority to enforce remedies? These conditions affect the power of review. Procedures should also allow respondents to answer allegations and distinguish academic disagreement, poor management, and serious mistreatment. Unexamined accusations should not settle every question of responsibility. Procedural protection constrains arbitrary decisions for all parties.

Informal rules can particularly confuse newcomers. Formal requirements may specify research completion while actual expectations include constant availability, extra organizational tasks, or emotional loyalty. Much mutual assistance is unobjectionable: covering for a colleague is part of shared life. The distinctions concern whether help can be refused, burdens rotate, and refusal is interpreted as unworthiness for professional opportunities. The ability to refuse while retaining basic rights makes voluntary assistance possible.

Guidance does require demands. Supervisors can require careful checking and explain consequences for stalled work. Whether a demand is justified depends on its relation to the task, the clarity of standards, adequate resources and time, and room for adjustment to new circumstances. “Not enough” or “the wrong attitude,” indefinitely detached from anything improvable, does little to help a student advance. It mainly keeps them waiting for the next verdict.

Organizations face genuine difficulties of responsibility. Project leaders may answer for students’ mistakes; supervisors may have many students and little time; committees may lack detailed knowledge. Distributing power by adding signatures can create delay without accountability. Improvement requires distinguishing initial judgment, provision of support, dispute resolution, and authority to act temporarily in emergencies subject to later review.

Legitimate authority therefore cannot be inferred solely from a task’s importance. Important tasks require coordination, but coordination must explain why powers are necessary, how long they last, and what responsibilities accompany them. Laboratory safety may require clear command; that does not extend the same command to people’s friendships, bodies, or entire futures. Powers concentrated for an emergency should return afterward to arrangements that permit discussion and review.

The Analects passage on government, punishment, virtue, and ritual concerns how order acts upon people. Modern research organizations face different conditions; moral cultivation cannot replace procedure. We may hope for virtuous supervisors while ensuring students still have somewhere to turn when that hope fails. Organizational responsibility gives that possibility conditions independent of private favour.

All under Heaven Is Held in Common — Book of Rites, “The Evolution of Rites”#

Researchers often have sound reasons to demand autonomy. A question may require prolonged investigation beyond a short-term manager’s understanding. A funder should not require a result to be rewritten because it disappoints expectations. Without time and room for judgment, research loses exploratory capacity and becomes a search for attractive proofs of predetermined aims.

Autonomy nevertheless exists in relations. Research uses social resources and can expose others to risk; knowledge enters policy, industry, and technical systems. Public responsibility already operates in topic selection, data collection, experimental design, confidentiality, and use of results. It is not merely an ethical afterword written when the work is finished.

Connecting autonomy and responsibility can begin by distinguishing three kinds of decision. Researchers must judge conclusions according to evidence, rather than announce facts to suit funders. Funding bodies may state their purposes and make demands concerning resource use. People affected by research and its applications have reasons to participate in decisions about risk and benefit distribution. These decisions involve different grounds and participants. Payment does not confer permission to alter data; expertise does not confer sole authority over social purposes; being affected does not make every technical claim someone offers correct.

States, businesses, and universities shape research differently. A state can support long-term infrastructure and direct work toward strategic purposes. Companies can bring practical problems and engineering capabilities while requiring secrecy or restricting publication. Universities can sustain training and exchange while reorganizing topics around rankings and budgets. Particular collaborations should be judged through agreements, decision processes, and consequences, rather than the funder’s category alone.

Funding also acts upon the next round of research. Sustained support for certain technologies accumulates instruments, curricula, skills, and careers. Neglect of other questions can make the eventual complaint that nobody studies them appear strangely sudden. Funders do not merely purchase existing knowledge; they help determine which capabilities can develop over time. Public responsibility requires looking beyond individual outputs toward the research conditions gradually formed by many decisions.

Global cooperation involves unequal resources. Imagine a cross-regional project in which one party supplies long-standing field relationships and samples, while another controls analytical equipment, the language of publication, and funding. If the latter alone chooses questions, interprets results, and allocates authorship, international cooperation can still concentrate judgment in one direction. Fairness concerns more than the countries named. Can local researchers and relevant communities help frame questions, receive results, and develop capabilities usable in future work?

Cross-regional cooperation also needs shared conditions of communication. A common language reduces costs while adding writing, editing, and presentation burdens for people whose first language differs. Open data can extend scrutiny but may let those already possessing computing resources and skills benefit more rapidly while providers of materials lack those conditions. Obtaining a file and using it to continue research are connected but distinct opportunities. Download access alone does not resolve the latter problem.

Research participants likewise need to remain able to exercise their rights. Consent, privacy, and reasonable handling of risk cannot stop at clearing a review. A signature documents a procedure at a particular time. Whether participants understood, could withdraw without losing existing services, or faced changed subsequent uses requires attention to the relationships involved. Procedures can continue to protect participants if these questions remain open to scrutiny. Reduced to certification of complete paperwork, they can become detached from their original purpose of protection.

Applications also change purposes. A method may improve diagnosis or enable inappropriate screening and surveillance. A security technique may protect systems or increase the capacity for harmful action. Researchers cannot foresee everything, but can judge visible uses, partners’ demands, and major risks. Responsibility should neither expand into personal omniscience nor shrink to the empty position that one handles only the technology.

Accountability also needs matters researchers can explain and affected people can examine. Specifying a project’s expected beneficiaries, potentially harmed groups, unresolved uncertainty, and routes of correction is more useful than proving it will ultimately benefit all humanity. The grander promise can become untestable rhetoric. The more definite account can change as research proceeds and allow affected people to identify omissions.

Public purpose can also support exploration whose immediate use cannot be demonstrated. Society may reasonably fund basic knowledge, preserve linguistic texts, or study phenomena not currently urgent, because future understanding cannot be fully commissioned by present needs. This does not remove competition for resources. It asks public deliberation to accommodate long-term, indirect, and uncertain value while researchers explain questions and progress seriously.

The most defensible research autonomy enables judgment answerable to objects and reasons; it does not exempt organizations from questions. The most defensible public accountability makes resources, risks, and consequences answerable to relevant people; it does not allow immediate preferences to dictate findings. They constrain and support one another. Without autonomy, accountability may receive only compliant reports. Without accountability, autonomy may become a privilege through which a few control shared resources.

Ancient Learners Studied for Themselves; Today’s Learners Study for Others — Analects, “Xian Wen”#

These relations also change the people participating in research. Researchers do not first become complete selves outside society and then enter a laboratory. Learning changes what they can see. An image once composed only of shades begins to disclose background, noise, and structure. A fluent passage reveals weak evidence and shifting concepts. Repeated checking, once difficult to endure, becomes a demand they themselves accept as appropriate to the work.

This exceeds the acquisition of skills. People learn to consider certain pursuits worthwhile and certain mistakes necessary to correct. They learn to explain judgments to others. A research subject forms: someone able to pose questions, accept constraints from the object, use a community’s language, and answer for claims. This person has agency while already inhabiting relations of training, expectation, and assessment.

In “The Subject and Power,” Foucault draws attention to the word subject’s connections both to control and dependence upon others and to being tied to an identity through self-knowledge.⁸ This suggests a line of inquiry: how researchers understand themselves and how others organize their actions are connected processes. The question here is how capacities, identities, and dependencies develop together in research, rather than how an external force simply suppresses an interior self.

A student initially uncomfortable explaining a methodological choice may develop judgment through repeated practice. A teacher’s demand becomes a standard the student can actively use, a valuable form of internalization. Through demands, responses, and repeated practice, people may also learn to equate immediate replies with seriousness, never expressing fatigue with strength, and always accepting tasks with academic promise. Each standard can be actively followed, but its grounds require separate scrutiny.

The distinction between internal and external is therefore insufficient to judge a demand. Sincere commitments can constrain someone, while criticism from others can support growth. We need to ask which activity a demand serves, whether reasons can be given for it, whether changing capacity or circumstances permit revision, and whether it expands a limited role into a person’s entire value.

We also need to distinguish how activity takes a form others can use from how people become constrained by those forms. Writing thought into a paper or organizing observations into data makes activity available to others; such objectification enables accumulated knowledge. When an organization treats a record of ability as someone’s fixed essence and sees only deployable productive capacity, living relations tend to be treated as fixed things. When collectively produced results and rules become unchallengeable forces determining their producers’ lives, obstructing their understanding of and participation in the use of results and the making and revision of rules, we can speak of alienation in this specified sense.

Imagine a team establishing an output sheet to improve communication. It initially helps members see one another’s work. Later its ranking determines all resources, while maintenance and training remain absent. Members know important work is missing but must reorganize inquiry around the ranking. A form they helped create becomes an external command they struggle to revise together. The problem is neither the sheet’s material existence nor calculation itself, but the separation of activity, expression, control, and revision.

Alienation, exploitation, and domination point to different relations. Judging wrongful appropriation of results requires examining contribution and distribution; judging largely unconstrained authority requires examining powers and their limits. Someone may also feel estranged from work without being directly dispossessed by a particular person. Concepts are useful because they specify what needs investigating. Alienation understood merely as dislike explains little. Understood as loss of understanding and participation in shared activity, it clarifies directions for change.

People usually occupy several research positions at once: learner and worker, project member and potentially dissenting author, recipient of public support and bearer of short-contract risk. These positions carry different rights and responsibilities. An organization that treats students as skilled employees when output is needed but only as educational beneficiaries when compensation arises can evade responsibility by switching roles. Participants likewise cannot claim only favourable aspects of roles while neglecting clearly agreed work.

Clarifying positions changes the language of dispute. Is an extra task necessary training, shared maintenance, or organizational labour unrelated to education? As training, what feedback and capacity development should it provide? As work, how should it count toward time, pay, and contribution? As emergency assistance, why is this person doing it, and when does it end? Ambiguity gives those with higher status greater freedom to interpret relations according to the occasion.

Repeated assessments can also consolidate identity. Someone whose caution is continually called procrastination may gradually hesitate to ask questions needing further checks. Someone whose compliance repeatedly receives praise may learn to agree before understanding. Eventually these assessments enter judgments about personal strengths and suitability for research. Established self-demands can carry existing behaviour forward without someone issuing commands every day.

Experience does not always fit an established identity. Researchers may accept assessment standards yet discover important omissions in practice, respect supervisors yet recognize particular mistakes, or desire positions while remaining attached to work with no immediate return. Encounters with objects, other people, and life can disclose things their established self-descriptions cannot explain. Reflection often begins in such mismatches.

Studying for oneself and for others in the Analects should not be directly equated with present-day selfishness and altruism. Here the passage helps distinguish learning that develops usable understanding from learning oriented principally toward a recognizable appearance. These are not exclusive alternatives. Public research necessarily faces others’ judgments, and people legitimately care about income, identity, and recognition. The question is whether explaining oneself to others tests the work or becomes merely arranging oneself to be accepted.

Autonomy should accordingly be understood through relations of dependence. Independent researchers still use others’ methods, seek help, and accept criticism. Independence includes explaining what one relies on, identifying where checks are needed, choosing suitable collaborations, and revising arrangements when reasons warrant it. Autonomy includes the capacity to identify, choose, and arrange relations of dependence, supported by conditions that make this capacity usable.

Because autonomy needs these shared conditions, demands upon individuals alone cannot establish it. Students need not become fearless and omniscient before earning the right to object; telling someone to believe in themselves does not yet make objection possible. Judgment, access to materials, livelihood support, and other accessible sources of informed assessment jointly determine whether a different view can be expressed, examined, and, when necessary, change an arrangement.

Teaching and Learning Develop One Another — Book of Rites, “Record of Learning”#

Developing this capacity for judgment often begins in an asymmetric educational relationship. Teachers know things students do not and can better anticipate some mistakes’ costs. Requiring identical decision-making powers at every stage may prevent training from proceeding. Novices need clear restrictions before handling dangerous equipment and may need to practise established methods before grasping basic concepts. The question is how temporary dependence helps people form judgments of their own.

Explaining procedures, identifying the origins of mistakes, and gradually transferring more complex judgments constitute training and give this educational asymmetry its grounds. Teachers’ decisions on students’ behalf today should connect to students making more decisions later. Greater ability also needs opportunities for actual use. If students can only ever execute instructions, training may become a permanent name for maintained dependence.

The process needs observable development. Initially students follow procedures. Later they should explain crucial conditions, compare methods, and recognize cases where procedures fail. They can then frame local questions, design tests, and accept corresponding responsibility. Progress will not be uniform, and fields cannot share one growth chart. Teachers should nevertheless explain which capacities a current restriction addresses and which developments would lead them to revise it.

Whether a student has developed the relevant ability usually still requires a teacher’s judgment. If only teachers can declare maturity, while every disagreement counts as proof of immaturity, the original asymmetry may never change. Shared supervision, periodic discussion, and external assessment allow others to examine demonstrated abilities and the teacher’s reasons. They do more than prevent malice: they make “not yet mature” a judgment open to review.

The wish that today’s differences in knowledge and ability should enable tomorrow’s shared judgment is a powerful educational aspiration.⁹ A teacher’s declaration cannot establish its realization. Teachers may sincerely want students to develop while needing them to complete projects. The aims do not always coincide. When independent topic selection reduces project efficiency, institutional support for the former commitment is tested.

Good supervision therefore involves more than a gentle disposition. Kindness can reduce fear without supplying effective guidance; demanding standards can promote growth without making humiliation necessary. The National Academies’ 2019 report on mentoring discusses both career development and psychosocial support.¹⁰ What matters here is support as a usable relation: someone helps the learner understand tasks, identify difficulties, receive feedback, and connect to other resources when needed. One person need not supply every function.

Sustained guidance allows teachers to know students over time, locate a difficulty, and place one performance within a process of development. This has more educational value than an unfamiliar evaluator’s single score. Risks arise from the same continuity: early impressions can harden, personal relationships can influence assessment, and students can struggle to escape their initial roles. Other relationships are therefore needed to supplement or review established judgments, with changes of supervisor possible where necessary. Long acquaintance should not become a permanent verdict.

Training must distinguish educationally useful difficulty from depletion without learning. Repeating an operation may develop practical sensitivity to error. But repetition without explanation of results or access to design may merely supply skilled labour for the team. Difficulty should be judged by what it enables someone to learn and whether new understanding and decision-making follow. Difficulty itself proves no educational benefit.

Independent exploration can develop capacity, yet independence can also name absent guidance. The distinction concerns whether a task matches preparation, resources and feedback are available, help exists after a dead end, and teachers retain educational responsibility. Autonomous inquiry needs somewhere to return for discussion. Being left alone with failure does not become freedom simply because nobody intervenes.

Students also teach one another beyond formal guidance. A senior student explaining an easily missed step or a colleague sharing the cause of failure can address immediate difficulties more closely than formal courses. Such work transmits everyday experience. If it remains unpaid while those doing it are judged insufficiently productive, an organization benefits from mutual assistance while penalizing its providers. Teaching and maintenance need corresponding time and recognition.

Development need not culminate in students surpassing teachers in every respect. No one can eliminate all knowledge differences across such an extensive professional world. What matters is learners’ ability to form questions, tests, and responsibilities within a domain, collaborate with teachers, and eventually work beyond that relationship when appropriate. Educational success can include someone no longer needing the same kind of dependence.

The original passage on teaching and learning begins with learning revealing one’s insufficiency and teaching revealing one’s difficulties. A useful implication is that teachers discover the limits of their methods through teaching. With this feedback, a student’s difficulty can become information about courses, guidance, and organization rather than always being a student deficiency. Whether a community receives that information determines whether education can change educators themselves.

Knowing Is Less than Loving; Loving Is Less than Delighting — Analects, “Yong Ye”#

Learning to research can also develop a love of research. A long-blocked proof yields a crucial opening, untidy data disclose a stable relation, or a collaborator supplies a missing step. These experiences have pleasures of their own. People rejoice in understanding and feel close to others through accomplishing something difficult. Rewards and pressure cannot explain this commitment alone. Criticism that treats pleasure simply as deception also struggles to understand why researchers cherish what they do.

In his 2009 retrospective on the open mathematical collaboration Polymath1, Gowers described several weeks as exceptionally exciting and emphasized how participants’ ideas stimulated one another. He also noted fewer active contributors than expected and the substantial time needed to follow rapidly accumulating discussion.¹¹ Pleasure and conditions of participation coexist in the same account. Their relation deserves more attention than the assumption that openness gives everyone equal participation.

The sources of these pleasures can be distinguished. An obscure object becomes intelligible; a person develops the ability to do something previously beyond them; others take up an isolated inquiry and connect it to wider understanding. All can receive recognition and rewards without depending entirely on them. A finding no one immediately cites may still give its author a real understanding of something.

Recognition has legitimate value too. Inquiry cannot remain perpetual speech to oneself. People need to know whether others understand their work and whether contributions enter shared memory. A serious response from a colleague familiar with the problem can matter more to a young researcher than general praise. Recognition supplies information about work and connects a person to relations in which development can continue.

Recognition becomes difficult when only one form remains available. If only one kind of journal, title, or position can establish a person’s worth, other real contributions may provide little security. Someone can achieve a desired qualification and immediately fix upon the next threshold. The first success gave them entry into a position without changing the relationship in which repeated qualification is necessary to confirm their worth.

This is more than an accusation of vanity. Positions affect income, housing, family arrangements, and the conditions for further research. Concern about them often has ample grounds. Concentrating those conditions in a few qualifications intensifies the desire for qualifications. Reducing recognition’s hold requires broader viable routes for life and work rather than performances of detachment in the face of risk.

Love of research can gradually become an organizational demand through everyday arrangements. Liking a problem prompts extra effort. Repeated extra effort may become the organization’s expected baseline, which then enters the next assessment. A voluntarily extended working day can become everyone’s required pace. The attachment remains real, but the steps from personal commitment to a collective requirement alter its consequences.

Whether such commitment remains voluntary depends on whether people can refuse without losing basic standing, understand the costs, reconsider later, and prevent their choice from becoming a requirement for everyone. Freedom to commit needs conditions for stopping and adjusting. Otherwise, someone’s willingness to do extra work can let the organization leave them carrying time and costs it should have arranged.

Professional ideals also shape interpretations of suffering. Difficulty can indicate proximity to an important problem, or poor design, insufficient resources, or an unbalanced relationship. A culture permitting only the first interpretation turns every obstacle into a personal ordeal to endure. This can harm people and research alike, since methodological errors lose opportunities for timely recognition.

Pleasure, conversely, cannot be the sole test of worthwhile work. Organizing materials, checking proofs, and cleaning data can be tedious but necessary. Important research may proceed for long periods without the excitement of discovery. Reasonable demands connect these tasks to intelligible purposes and distribute their burdens rather than requiring daily passion. Calm, sustained work can be a mature relation to inquiry.

Cynicism offers another response. Someone recognizes defective metrics, complies, and invokes the fact that everyone does it to protect against disappointment. This may help them survive an unchangeable situation in the short term. If everything becomes a transaction in advance, however, serious judgment becomes harder to sustain. Organizations may most readily tolerate criticism that has already promised to change no behaviour.

A way forward need not begin with perfectly pure motives. Someone can want both to solve a problem and secure a position, both to cooperate and receive credit. What matters is making these motives intelligible and negotiable without granting one unquestioned supremacy. Strong professional commitment can coexist with other sources of life: relationships, care, interests, and forms of existence that do not enter a CV.

When research no longer bears the entire burden of proving a life’s value, researchers gain another way to face failure. A hypothesis lacking support can prompt examination of evidence and methods before a decision about revising it, without immediately negating the whole person. Recognizing several sources of value need not diminish research’s depth. It may make unwelcome findings easier to face: a researcher need not preserve every hypothesis in order to preserve themselves.

Tension and Release: The Way of Kings Wen and Wu — Book of Rites, “Miscellaneous Records II”#

Sustained commitment nevertheless also needs bodily and practical support. Hands tire, attention falls, and sleep loss makes fine judgment harder. Reading, experiments, and writing require food, housing, medical care, and life with others. Calling these private matters outside research makes another portion of the work supporting research disappear from view.

The Book of Rites passage on tension and release occurs in a discussion of popular festivity after Zi Gong observes the year-end zha sacrifice. It cannot directly establish an ancient precedent for modern working-hours rules. It can remind us that sustained activity includes recovery and that recovery requires collective arrangements. Counting only time visibly spent working can deplete the capacities that make work reliable.

Tasks have different temporal requirements. Experimental windows can be brief, field access seasonal, and deadlines demanding. These differences do not justify indefinite emergencies. An arrangement should explain why an intense phase is necessary, who bears it, what recovery follows, and whether shifts, handovers, and preparation can reduce the burden. If every week is critical, urgency has become a routine management method.

The same demand on time can impose different burdens upon different lives. People available for travel, relocation, and evening meetings differ from those with fixed caring responsibilities under the same standard of commitment. Formal eligibility for opportunities can coexist with unequal time and mobility. Treating constant availability for work as evidence of passion mistakes a particular household arrangement for an individual virtue.

Using longitudinal data on American STEM professionals, Cech and Blair-Loy studied employment changes after becoming a parent. They found substantial departure from full-time STEM employment among new parents, with a more pronounced loss among women.¹² American STEM employment is not equivalent to university research everywhere, and leaving full-time positions is not necessarily leaving all scientific activity. The substantive question is how continuity of careers and caring can become jointly possible.

When households alone provide care, growth, recovery, and life support keep research going without the organization recording these costs. Someone still bears them: partners, parents, paid carers, or researchers through reduced rest. Those apparently most able to endure may also receive more hidden support. Particular conditions need examination; outward performance alone cannot determine who is stronger.

Research failure also requires subsequent work to deal with it. A design that has not produced the expected evidence still leaves materials to clear, delays to explain, and resources to await. If project completion, graduation, or renewal dates cannot respond to research uncertainty, honestly explaining unresolved questions can threaten livelihoods. The prospect that apparently complete findings are safer than admitting something remains unknown is an incentive requiring scrutiny.

Pressure does not inevitably cause misconduct, nor does difficulty erase responsibility. The point is to avoid sustained opposition between honesty and survival. Recognizing informative failures, distinguishing deficient methods from reasonable exploration without success, and assigning responsibility and support for delays help make honesty sustainable rather than heroic.

When failure becomes “I am unworthy,” judgment of a research result has become judgment of the self. Work may be closely tied to identity, and identity to life chances. If surrounding responses offer few other languages of appraisal, a local mistake can appear to threaten a whole future. Addressing this requires support for particular people and renewed explanations of how the result arose, who bears which responsibilities, and what it actually shows about ability. One result cannot settle a person’s worth.

Institutional criticism must not cancel individual support. Someone in difficulty may need rest, trusted relationships, and professional help without first producing a structural analysis. Nor can personal support replace institutional change. Teaching tolerance while excessive workload, humiliating interaction, or uncertain training arrangements remain in place returns organizational problems to individual management.

Accounts of distress should preserve differences of experience. Similar arrangements can affect people differently; one person’s endurance does not establish fairness, and another’s earlier harm does not establish inferior ability. Suffering is information deserving serious attention without automatically identifying its entire cause. Tasks, relationships, duration, and other life conditions need to be considered together.

Research ethics thus includes responsibility for sustaining capabilities. Sleep, care, predictable living conditions, and recovery deserve protection beyond whatever effect they have on publication output. Participants have lives of their own. Even a sacrifice that demonstrably raises short-term efficiency still raises questions of authority, refusal, and shared cost.

A community sustaining output only by continually replacing exhausted people may keep work going in its accounts while interrupting the lives and capabilities of previous participants. Its achievements should be considered alongside these departures, silences, and unfinished work. Reading only the CVs of those who remain can make the whole process appear to be a path everyone could traverse smoothly.

Act When Employed; Withdraw When Set Aside — Analects, “Transmitting”#

The easiest advice for someone in an unreasonable research relationship is to leave: change supervisors or laboratories, enter industry, or stop researching. All can be important actions. All require conditions. Presenting a theoretically available route as one everyone can take immediately turns the difficulty of acting back into insufficient courage.

The ability to leave is often connected to how someone entered. Differences begin accumulating before entry into research. Early exposure, knowledge of how to find supervisors, time to prepare applications, and support after unsuccessful attempts influence what later appears as talent and commitment. Opportunities are formed before selection as well as allocated at it.

Morgan and colleagues surveyed 7,204 tenured and tenure-track faculty members across eight American disciplines and compared their family backgrounds with relevant national data. Their findings identify a strong association between entry into the professoriate and advantaged socioeconomic origins.¹³ They do not show that every privileged person lacks ability, nor directly establish the situation of Chinese graduate students. They clarify that those finally selected have already survived many earlier stages of selection.

After entry, identical records may also be acquired at different costs. One person can extend training on low income; another must support a family sooner. One can relocate for opportunities; another is constrained by health or care. Comparing only final CVs can mistake the conditions for affording a wait for perseverance and the ability to move freely for enterprise. Actual ability still needs serious judgment, but that judgment should not also become a verdict on character as a whole.

Exit concentrates these differences again. Leaving a supervisor can mean losing a topic, equipment, references, and existing collaborations. Leaving a city can transfer employment and care costs to family members. Leaving an academic job does not instantly convert knowledge into another income. Meaningful exit needs portable records of contribution, reasonable handovers, alternative references, and continuity of living conditions. “No one is forcing you to stay” otherwise ignores the accumulated weight of the relationship.

Departure involves both past investment and future possibilities. Time and money already spent and no longer recoverable do not alone justify continuing. The capacities and relationships developed through training, however, remain conditions potentially usable in the future. Treating the former investment as a sunk cost does not mean discarding the latter conditions. Judgment needs to compare the consequences of different paths from the present onward while allowing time for transition and for established expectations and feelings to be reoriented. Economic calculation can inform a decision without completing the work of departure.

These costs also prevent staying and leaving from directly establishing a person’s stance. Someone who stays may value a problem, colleagues, and students, believe local change remains possible, or temporarily lack alternatives. Leaving need not expose research ideals as false. People may use their training after leaving a position or find a better life in different work. Career paths and relations to knowledge do not coincide completely.

Leaving academia also needs a specified level. Leaving a project differs from leaving graduate training; leaving a university differs from ceasing to create knowledge; abandoning promotion differs from abandoning serious work. Calling them all surrender removes intermediate choices. Transfers, pauses, different roles, and re-entry routes reduce the need to choose between total compliance and complete rupture.

Multiple paths cannot exist only in guidance documents. Teaching or technical positions accompanied by treatment as a failure do not become genuinely respected options. Roles may differ in duties, pay, and assessment. Those differences should follow the work rather than ranking human worth by distance from one ideal scholar. Research depends on multiple abilities and should allow several to support stable, dignified careers.

Discussion of personal action should first establish the decision facing someone, which materials they can lawfully retain, who can provide trustworthy judgment, and which changes their life can bear. Someone with an established position may have conditions for changing procedures or speaking publicly. Someone without security may first need to preserve learning and livelihood. Demanding identical risks from everyone reproduces existing inequality.

Organizational responsibility for choice goes beyond wishing departing members well. Clear handovers, accurate contribution records, non-punitive references, and reasonable transitions reduce unnecessary loss. Departure need not negate previous work. Available exit routes also change the situation of those who stay: cooperation depends less on having nowhere else to go, and commitment can arise more readily from relations that remain open to renewed choice.

The Analects passage about acting when employed and withdrawing when set aside belongs to a conversation between Confucius and Yan Hui. Its composure cannot simply be demanded of people with unequal living conditions. Entry and departure require roads and provisions. Respecting a person’s choice can coexist with asking why every turn costs some people so much more.

To Err and Not Change Is Truly to Err — Analects, “Duke Ling of Wei”#

Whether people can stay, change direction, or leave is also affected by how organizations handle errors. Scientific correction is a commitment worth preserving, but different mistakes need different procedures. Reanalysis can correct a data-processing error. Recalculating data cannot compensate someone for years of unfair labour. Retracting a paper does not automatically restore a degree, income, or damaged life.

Epistemic correction must be distinguished from organizational correction. The first examines whether claims have adequate grounds; the second examines appropriate use of authority and reasonable arrangements of responsibility and cost. They can support one another. Access to records helps identify methodological problems, and protection for people who raise concerns supports honest reporting. Yet they are not automatically equivalent.

A rigidly hierarchical team can produce correct findings, and an egalitarian group can collectively believe something false. Fairness cannot replace evidence, and achievement cannot absolve mistreatment. Recognizing only reforms that improve output makes rights perpetually vulnerable to efficiency comparisons. Considering only participants’ satisfaction removes research’s obligation to its objects. Both tests must remain.

Distinguishing the two kinds of correction helps identify what a reform changes. Registered Reports move part of peer review before research outcomes are observed and offer in-principle acceptance after the relevant review, conditional on quality assurance, following the protocol, and adequate reporting, among other requirements.¹⁴ By changing review timing and researchers’ publication expectations, this arrangement attempts to weaken incentives to publish only sufficiently striking findings. It does not guarantee correct designs or suit mechanical application to every exploratory activity.

Other reforms must likewise specify which link they change. Contribution statements increase some work’s visibility. More independent dispute handling allows different people to review an original decision. Stable transitional support lets someone object without immediately losing a livelihood. These arrangements address different problems and need separate examination. One improvement cannot establish justice throughout a system; remaining problems do not erase the relationships it has already changed.

New arrangements generate new labour. Open data need organization, explanation, and maintenance. Reproducible code needs environment records and technical support. Careful review takes time. Independent appeals require staff, relevant expertise, and enforcement powers. Assigning all of this to those already short of time can turn good purposes into new burdens. Budgets and workloads belong inside reform, rather than being left for enthusiastic members to absorb.

New work created by oversight can also impede existing work. Records help review; reporting every action can consume research and living time. Matching checks to the seriousness of risks, recording reasons for consequential decisions, and allowing relevant people to obtain necessary materials often serve the purpose better than indiscriminate documentation. Those supervising must also be constrained, or concentrated power merely moves to another office.

Reform can also move a difficulty to another stage. Requiring several approvals to protect students may stall training if nobody is responsible for timely handling. Excluding everyone familiar with a project to avoid conflicts can deprive review of expertise. These trade-offs need to be explicit, with time, responsibility, or professional capacity supplied where the new arrangement lacks them, and adjustment through practical feedback.

One especially fragile reform changes terminology while preserving resource allocation. Publication counts become comprehensive contribution, but promotions reward the same outputs. Failure is encouraged, but renewal still requires every exploration to succeed on time. Communication channels appear without investigative or enforcement authority. Whether reform occurred must be traced through consequences. Whose time changed? Who gained a new opportunity to judge? Whose risks actually fell?

Assessment of reform cannot merely count measures adopted. It can examine whether disputes are handled within reasonable periods, research conditions survive an objection, contribution records affect decisions, and training expands independent judgment. Institutions need different indicators, but indicators should answer these substantive questions while leaving room for experiences beyond their records.

These consequences also need comparison over time. More complaints after a change can mean more problems or a newly usable channel. Increased movement can indicate deteriorating conditions or fewer barriers to transfer. One number seldom interprets its own change. Time, populations, process records, and participants’ experiences must be connected, with examination of whether other simultaneous changes affected the outcome.

Improvement is therefore continuing organizational work. Someone must formulate a problem, obtain authority to address it, bear the costs, and later acknowledge remaining deficiencies. Research asks findings to face counterevidence; research organizations need a related capacity to revise their arrangements. The similarity lies in facing unwelcome facts. The difference is that institutions must also answer judgments about rights, responsibility, and life.

Learning from One Another’s Strengths Is Mutual Cultivation — Book of Rites, “Record of Learning”#

Reform also needs someone to organize shared problems into changes that can be carried out. Individual researchers can alter some relationships but rarely rewrite shared assessment, employment, or resource systems alone. Everyone may recognize a time-consuming requirement while fearing the consequences of being first to refuse. Everyone may want a shared tool maintained without being able to maintain it alone. Problems borne together require conditions for shared judgment and action.

Bringing these experiences together can make previously isolated difficulties discussable and give private requests an opportunity to become common rules. New organization also creates divisions of labour, representation, and power. Who represents members, controls information, does coordination, and preserves disagreement? Shared experiences of harm can bring people together, but decisions made afterward still need reasons and scrutiny.

Forms of action should match problems. Humiliating discussion in a group meeting may be addressed through clear discussion rules and accessible support. Barriers to transfer across a training institution require broader coordination and resources. Cross-institutional assessment may require societies, funders, journals, and employers to change together. Local mutual assistance can support people without being expected indefinitely to replace institutions holding formal powers.

The division of action tests whether reform follows its own reasons. Securely employed reform advocates who leave students or temporary staff to bear the main consequences of speaking once again shift risk onto people with less protection. Greater ability and security provide reasons to assume more coordination, public explanation, and protection. Vulnerable members may participate while choosing lower-risk routes without accusations of cowardice.

Mutual assistance can make immediate differences: sharing procedural information, interpreting reviews, preserving publicly shareable research experience, and creating discussion across supervisory relationships. These practices expand the resources available to individuals confronting organizations. Continuity requires maintenance, however. If the same few people always listen, organize, and coordinate, they may exhaust themselves while helping others avoid exhaustion. This labour needs visibility, rotation, support, or the ability to stop.

Some therefore seek new communities beyond formal academic organizations. Independent reading groups, open research teams, public writing, and collaborative platforms can free questions from established assessment demands. They may lower barriers, develop new language, and include people without institutional positions. Their value should be judged through these concrete capacities.

New communities must also arrange life and work. Livelihoods, platform and material costs, time allocation, agenda-setting, contributions, and disputes do not disappear in spontaneous groups. Members’ income from other jobs is a source of material support; founders’ reputations are forms of authority. Leaving a university changes the form of dependence. New dependencies still need to be understood and arranged together.

Openness does not give everyone the same capacity to participate. The Polymath1 retrospective illustrates how rapidly accumulating discussion creates a barrier to keeping up. Broader practical openness may require summaries, introductory materials, clearer problem allocation, and records others can resume.¹¹ These demand labour. Communities rewarding only the most conspicuous ideas while neglecting entry points may repeatedly attract a few already-prepared participants.

Which research conditions can be rebuilt also depends on the work. Public-text discussion can proceed at comparatively low cost. Research requiring expensive equipment, long-term samples, or strict safety conditions cannot easily subsist on platforms and enthusiasm. An alternative needs to explain how it can obtain and sustain these capacities. Criticizing a bridge’s location still leaves the question of how people will cross the river.

Alternative organizations need to handle their own errors and misconduct. Strong common purposes can make criticism appear disloyal, and a founder’s large contribution can make limits on authority difficult to accept. New relationships remain more open when decision-making, records, resource use, and recognition after departure are explained early. Rules can be simple; responsibility cannot depend only on familiarity.

Relations with formal institutions need not reduce to joining or breaking away. Independent communities can supply new reading, discussion, and criticism while professional institutions supply equipment, sustained support, and quality checks. They can cooperate on particular questions. Such cooperation should be judged by whether it expands capabilities, concentrates authority at a single gateway again, and supports durable commitments. Several dependencies can sometimes protect autonomy better than searching for a place without any.

Collective action needs both to change unreasonable decisions and to form relationships capable of continued shared judgment. After a particular dispute is resolved, members still need training, trust, and maintenance to keep cooperating. Pursuing victory alone can leave people without capacities for continuity. Preserving relations without touching decisions can turn solidarity into polite conflict avoidance. Changing decisions and sustaining cooperation must test one another in practice.

A community’s worth ultimately concerns what it enables members to do: understand objects more clearly, ask previously difficult questions, make cooperation’s costs visible, and revise relationships after mistakes. Its name can be new and its scale small. What matters is whether these capacities actually develop.

People Can Enlarge the Way; the Way Does Not Enlarge People — Analects, “Duke Ling of Wei”#

Looking back, knowledge, organization, and the making of researchers have met at many points. Methods and shared labour produce results; organizations use assessment to distribute resources and opportunities; participants arrange the next round of work around those conditions while learning to demand things of themselves through the same standards. Reliable knowledge may accumulate, while difficult dependencies may deepen. Each step involves particular judgments and arrangements. Changing the process requires identifying where reasons are insufficient and where usable alternatives are missing.

Scientific language can make claims’ evidence and inferences available for scrutiny. Using it to assess contributions, allocate tasks, or judge qualifications requires separate reasons for allocation and qualification. An objective tone cannot supply missing grounds. Correct papers do not establish fair labour arrangements; fair arrangements do not guarantee correct papers. Taking both seriously makes it possible to uphold knowledge and its participants together.

People’s abilities and dependencies can grow within the same process. Training develops judgment, organization extends action, and recognition brings responses to work. The same process can tie people to a single identity and make disproportionate demands hard to refuse. Actual activities must show how capacities develop, who allocates opportunities, and how people understand their effort. An overall celebration or denunciation of research has not yet answered these questions.

At least five considerations bear on whether an arrangement deserves continuation: stronger grounds for knowledge, participants’ continuing development, reasonable rights and burdens, durable resources, and consequences for relevant publics. These standards arise from different reasons and cannot become one automatic ranking.

These standards have distinct grounds. Reliable knowledge requires claims to answer to evidence and inference. Training that develops people needs opportunities for capacities to grow and be used, rather than fixing participants as instruments. People’s lives cannot be governed solely by organizational convenience, so rights and burdens need reasons of their own. Continuing today’s achievements depends on tomorrow’s resources and personnel. Research using shared resources and affecting others must also be answerable to relevant people. Each consideration needs concrete examination; achievements elsewhere cannot cancel its demands.

They can conflict. Rigorous research may need more time than a student’s livelihood allows. Wider participation can add important experience and coordination costs. Privacy restricts data openness. Exploratory freedom may reduce visible short-term output. There is no context-free total formula. We should first identify basic requirements that are not readily tradable, then seek resources and arrangements that reduce conflict, and finally make reasoned choices where aims cannot be jointly met.

Participants’ basic dignity cannot be overridden on grounds of greater efficiency; research cannot announce a known falsehood to preserve relationships. Between these requirements, many arrangements remain changeable: narrowing claims, adjusting time, finding alternative tests, adding support, and separating educational assessment from one project’s success. Before presenting knowledge and people as directly opposed, we should examine the organizational choices between them.

Return to the data needing repetition. Better handling need neither declare the supervisor wrong in advance nor cancel further experiments. It should first explain the evidential gap and which additions could alter judgment; then discuss feasible approaches and resources; then specify labour, contribution credit, and the consequences of delay. Serious remaining disagreement should be examinable by suitable others. Every stage returns a decision to the reasons it needs.

The result may remain difficult. Research may miss its original schedule, a conclusion may require withdrawal, and a student may not yet meet the required level. Respect does not guarantee everyone’s preferred outcome. It requires naming difficulty accurately, assigning responsibility properly, keeping assessment within its grounds, and preserving practical routes to improvement, change of direction, or exit.

Even after change occurs, the original judgment needs checking. Distributed supervision that confuses feedback calls for clearer responsibility. New assessment that neglects important labour needs revision. Open collaboration that merely overburdens a few calls for examination of entry and maintenance. A particular protective arrangement that obstructs necessary research requires an account of the conflict and better arrangements. Precisely because improvement has costs, it needs continued adjustment through this feedback.

Our concepts can also change through these adjustments. Someone finally obtains a contribution record, raises a question without losing a livelihood, or carries learned abilities into new relations. Autonomy, recognition, fairness, and community then acquire more definite content. Practice participates in forming theory rather than merely executing recommendations when theory is finished. Theory’s depth also depends on whether it can distinguish these changes and identify where established concepts need revision.

The Analects passage concerns people and the Way. Applied here, it cannot mean that determination alone overcomes institutions. Making reasons real often needs many people, resources, and durable arrangements. Science’s noble name does not itself protect researchers, nor does an elegant declaration deliver fairness to a laboratory.

We can build research relations in which serious workers need not surrender entire lives to a single verdict: conclusions remain examinable, authority bounded, labour recognizable, training enlarges participation in decisions, and departure does not erase contribution. A person can love an activity for a long time while retaining room to judge it, change it, and leave when necessary.

When a researcher can explain that the evidence is insufficient and also give reasons why an arrangement must change, scientific discourse preserves both rigour toward its objects and clarity toward shared life. The community may not gain more compliant members, but it may gain people more capable of continuing to research together.

The cover is an original AI conceptual illustration of shared research labour and individual judgment. It depicts no real institution, person, or historical scene.