Full chapter text
Arbitrary
Power–Knowledge–Truth test cycle: question → access → classification → evidence → interpretation → counterargument →
independent verification → correction → accountability
Knowledge is not produced in an empty sky. Who has the right to ask questions, who has access to libraries and archives, which
language is treated as ‘standard’, which data are collected, which forms of expertise receive prestige, which research receives
funding, and whose testimony is regarded as ‘credible’—all these are influenced by social and institutional relations of power. To
deny this ordinary fact would be naïve realism.
But the opposite extreme is equally mistaken: if power influences the production of knowledge, then truth is merely a story
made by power. This conclusion conflates the ‘social conditions of knowledge’ with the ‘truth-value of a claim’. A government
cannot stop rainfall by issuing an order against rain; a university cannot change a historical event by changing a vote count; a
company cannot eliminate the biological effects of a toxic substance by renaming its risk.
Parallel Philosophy proposes that the relation between power and truth is neither complete independence nor complete identity.
Power can alter the route, speed, cost, visibility and interpretive options through which truth is approached; yet the world, the
body, documents, memory, measurement, counterargument and independent sources repeatedly offer resistance that causes
arbitrary claims to fail.
The aim of this chapter is not to defend a naïve ideal of the ‘view from nowhere’, but to build institutional conditions for better
objectivity: source provenance, disclosure of conflicts of interest, methodological transparency, space for dissent,
reproducibility, data access, version history, appeal, error correction and multi-evidential testing.
The Problem
The first problem concerns the selection of questions. Funding, prestige, administrative priorities and public pressure influence
which subjects of research are regarded as ‘important’. A question that was never asked will have no answer in the record; but
the absence of an answer does not make the object of the question unreal.
The second problem concerns archival access. Documents may be destroyed, classified, unclassified but inaccessible, locked in
private collections, poorly catalogued, or available only in one language. A historian’s claims are affected by this structure of
access; even so, critical examination of the available sources can still yield conclusions of greater or lesser reliability.
The third problem is classification. States, schools, hospitals, courts, companies and platforms continuously create categories—
resident/non-resident, risky/safe, eligible/ineligible, crime/unrest, diseased/normal, spam/legitimate. Categories have
consequences; but the social effect of a category does not answer the separate question whether its underlying factual basis is
sound.
The fourth problem is expertise. Complex knowledge is impossible without communities of experts, but expertise is not
automatically infallible. Credentials, peer networks, citation prestige, journal access, linguistic dominance and disciplinary
boundaries affect whose voice is heard. The solution is not anti-expertise, but transparent expertise.
The fifth problem is the unequal weight of surviving evidence. The records of powerful institutions are often preserved, while
the oral memories of weaker communities may disappear. An archive is therefore not a perfect mirror of ‘what happened’. Yet
recognising archival bias and treating every oral claim as automatically true are two different things.
The sixth problem is propaganda. Power does not merely hide facts; through information overload, repetition, framing, selective
disclosure, emotional salience and algorithmic amplification, it can raise the cost of finding truth. Misinformation is not always a
complete falsehood—an incomplete, contextless, or technically true statement can also mislead.
The seventh problem concerns resistant standpoints. Marginalised experience can reveal dimensions ignored by dominant
models; but ‘a harmed person said it’ is not, by itself, final proof of every factual proposition. Experience is epistemically
important, not infallible.
The eighth problem is relativism. If factual contradiction is dissolved by saying that every group has ‘its own truth’, justice,
science, history and public policy become impossible. Two contradictory claims cannot both be true in the same sense at the
same time.
The ninth problem is objectivity. Merely saying ‘I am impartial’ does not eliminate bias. Better objectivity is created when
procedures contain mechanisms for discovering bias: adversarial review, preregistration, independent replication, open data,
audit trails, dissenting notes, conflict-of-interest disclosure and rights of reconsideration.
The tenth problem concerns the political use of truth. Claims to truth can themselves become instruments of power: phrases
such as ‘science says’, ‘history proves’, and ‘the data show’ demand authority. It is therefore necessary to make public the level of
evidence, the uncertainty, the assumptions of the model and alternative interpretations.
Core Proposition
Parallel Philosophy distinguishes three levels in the power–knowledge relation. First, the production level: questions, data,
institutions, language, funding, access and method. Second, the dissemination level: publication, platforms, censorship, ranking,
prestige and pedagogy. Third, the truth-testing level: evidence, contradiction, prediction, replication, independent corroboration
and defeating evidence.
Power can exert enormous influence on the first two levels and can obstruct the third as well; but it cannot make the third
wholly arbitrary, because claims collide with the world, with sources and with other claims. A badly designed bridge may
collapse, a wrong medical dose may cause harm, and a fabricated date may prove inconsistent with independent archives.
‘Truth’ in this chapter does not mean magical absolute certainty. Truth-claims may be revisable, confidence may be graded, and
evidence may be incomplete. Fallibility and arbitrariness are not the same thing. From the fact that we can be wrong, it does not
follow that anything goes.
Twenty-Five Maxims
1. Power can change the agenda of questions; it cannot abolish the factual constraints on answers.
2. Archives are selective; selective does not mean imaginary.
3. Silence can be evidence, but investigate the cause of the silence independently.
4. Experience is important evidence; experience is not omniscience.
5. Expertise is necessary; a credential is not infallibility.
6. Numbers are powerful tools; measurement design is not value-free.
7. Categories create social consequences; test their empirical adequacy separately.
8. A standpoint can reveal what was overlooked; a standpoint is not itself the final judge.
9. Consensus can be evidence; examine the independence of the agreement and the incentives behind it.
10. Dissent matters; dissent is not proof of truth.
11. Censorship can diminish knowledge; absolute freedom of speech does not automatically increase truth.
12. Transparency is useful; a raw data dump is not the same as transparency.
13. Fluency without provenance is not reliability.
14. A conflict of interest does not automatically make a claim false; disclosure increases scrutiny.
15. Institutional trust is not blind faith; it is audited reliability.
गजेन्द्र ठाकु र
16. A repeated claim is not independent corroboration.
17. Source diversity is useful only when the sources are genuinely independent.
18. Language changes framing; framing does not itself change the event.
19. Power inequality changes the uptake of testimony; its factual content remains testable.
20. Facts and values belong to different levels, but they interact in policy.
21. Acknowledging uncertainty is not weakness; fabricated certainty is anti-knowledge.
22. Correction mechanisms are central to a truth-seeking institution.
23. Algorithmic authority without a right of reconsideration is democratically weak.
24. Multi-evidentiality does not mean ‘all sides are equal’; evidence may carry different weights for stated reasons.
25. The critique of power can itself become power; self-critique is indispensable.
How Power Shapes Knowledge
Agenda and Funding
The availability of funding changes the direction of research. Limited funding for rare diseases, local languages, public health,
environmental damage or low-profit technologies can produce less research in those fields. Do not treat this inequality as
evidence that the subject is unimportant. A map of funding is itself epistemically relevant evidence.
But the existence of a funder’s interest does not by itself invalidate a result. Examine the method, data, preregistration,
replication, independent teams and conflict-of-interest disclosure. A conflict of interest increases scrutiny; the evidence
determines truth-value.
Archives and Memory
State archives preserve the administrative viewpoint more readily; domestic letters, women’s voices, workers’ memories, oral
traditions, local accounts or minority materials may survive less often. Here the historian must identify ‘archival silence’.
Archival silence is not a licence for arbitrary filling-in. A methodological distance must be maintained between ‘there is no
source’ and ‘I shall fill the gap by imagination’. Possibility, inference, plausibility and established fact should be clearly separated
by labels.
Education and the Curriculum
Curricula create canons. Authors, events and languages excluded from the curriculum may disappear from the intellectual
horizon of a new generation. Canon revision may be necessary, but representational balance alone does not automatically
guarantee factual accuracy.
A better curriculum will include multiple sources, historiographical disagreement, the reading of primary evidence and training
in method. Students should be shown not merely conclusions, but the process by which conclusions are formed.
Publication and Prestige
Peer review can be a quality filter, but it can also become gatekeeping. Acceptance by a prestigious journal is not a guarantee of
truth; rejection of a paper is not proof of falsehood. Publication bias, language bias, incentives for novelty and citation networks
are all open to examination.
Open peer review, registered reports, negative results, access to data and code, and post-publication correction can become
institutional remedies.
Law and Administration
Law makes categories binding: ‘minor’, ‘resident’, ‘scheduled’, ‘protected’, ‘hazardous’, ‘eligible’, and so on. Legal truth and
empirical truth can differ—legal definitions serve administrative decisions, whereas scientific classifications serve causal
explanation.
Even an official legal record is not the whole event. FIRs, affidavits, judgments, censuses and notifications should be read
according to source type, with attention to provenance, purpose, incentives and evidentiary limits.
Digital Platforms and Algorithms
Search ranking, recommendation, moderation, virality and ad targeting affect which information comes before us. Platform
visibility is not a truth score. ‘Everyone is seeing it’ indicates popularity, not evidence.
In algorithmic classification, examine training labels, proxy variables, thresholds, feedback loops and the right of
reconsideration. Who bears the burden of classification error is an ethical and power-related question; a model’s predictive
accuracy is a separate empirical question.
Why Truth Is Not Arbitrary
The first reason is resistance. The world does not always behave according to our wishes. A wrong map can spoil a journey; a
wrong dose affects the body; a wrong date collides with a documentary sequence.
The second reason is independent convergence. When different instruments, different teams, different archives, different
languages or different methods point towards the same conclusion, reliability rises—provided the sources are genuinely
independent.
The third reason is predictive or inferential success. A theory that correctly anticipates new observations or outcomes carries
epistemic weight beyond authority alone. But overfitting, selective reporting and base rates must still be examined.
The fourth reason is replicability or reproducibility. A repeated experiment is not possible in every field, especially history;
nevertheless, procedural transparency, re-analysis, source verification, alternative coding and independent reconstruction can
perform comparable functions.
The fifth reason is contradiction. In the same sense, ‘the event occurred in 1900’ and ‘the event occurred in 1910’ cannot both be
true. Once ambiguity is removed, contradiction requires factual adjudication.
The sixth reason is defeating evidence. New, credible evidence can reduce confidence in an older belief. A rule of revision of this
kind marks the difference between power-protected dogma and truth-seeking.
The seventh reason is public scrutiny. Compared with private conviction, a claim is more reliable when it shares evidence,
withstands criticism, accepts error correction and explains its method.
Standpoint, Experience and Epistemic Injustice
Some social positions make available information that a privileged observer may see less clearly—for example humiliation
within service systems, workplace harassment, disability access, linguistic discrimination, or the insecurity of informal labour.
In this sense, standpoint is an epistemic resource.
Miranda Fricker’s concept of testimonial injustice shows how a speaker’s credibility may be unjustly reduced because of social
identity. In hermeneutical injustice, shared conceptual resources may be insufficient for expressing an experience.
Parallel Philosophy accepts this insight but rejects identity-based infallibility. Experience is important evidence for the claim
‘this happened to me in this way’; the further claim ‘therefore this alone is the complete causal explanation’ is a different
proposition and requires additional evidence.
A better institution will create structures in which marginalised testimony can be heard, retaliation is reduced, language access
is provided, and procedures for corroboration and reconsideration are clear. Respect and truth-testing are not opposed.
Foucault, Genealogy and the Parallel Response
Michel Foucault’s analysis of power/knowledge raises the important question of how categories such as ‘normal’, ‘ill’, ‘criminal’,
‘mad’ and ‘disciplined’ are connected with institutional practices, examinations, records, surveillance and professional
discourse.
गजेन्द्र ठाकु रक समानान्तर दर्शन — खण्ड २
Genealogy reveals the contingent histories of categories: what now appears natural was formed when, through which
institutions, amid which disputes and for which administrative needs? This question does not simply strip a category of
authority; it clarifies its origins.
But the historical contingency of a category does not automatically imply that the category is wholly false. The history of a
disease concept may be social, while biological regularities remain real. A statistical category may be politically constructed,
while the measured disparity remains real.
The Parallel response is: critique of origin + present evidence. Always ask two separate questions together: ‘Where did this
concept come from?’ and ‘How strongly does current evidence support it?’
Science: Institution, Power and Self-Correction
Science is a human institution and therefore is not free from funding bias, prestige hierarchies, fraud, publication bias, exclusion
by gender, race or caste, national rivalry, corporate interests and the like. Recognising these social facts is not anti-science.
The special strength of science lies not in infallible scientists but in mechanisms of correction: explicit methods, measurement,
peer criticism, replication, instrumentation, cumulative records, correction and retraction. When these mechanisms are weak,
reliability declines.
Instead of saying ‘science says’, ask: which study? What sample? What method? What uncertainty? Is there a systematic review?
Replication? Conflicts of interest? Where does the evidence sit in the hierarchy? What degree of agreement exists among domain
experts? What disputes remain unresolved?
In policy, empirical evidence is joined by value judgements. Science can estimate risk; deciding ‘how much risk is acceptable’ is
also a social and ethical judgement. Collapsing these two levels creates technocracy.
History, Genealogy and Archival Truth
Laboratory replication is limited in history, but source criticism, chronology, provenance, palaeography, archaeology, material
evidence, independent accounts, linguistic dating, administrative context and historiographical comparison are instruments of
truth-testing.
A royal inscription may be panegyric; this does not make every piece of information it gives about date, ruler or place false.
Disaggregate claims according to genre: self-praise, administrative fact, genealogy, ritual claim, territorial aspiration—each
carries a different evidentiary weight.
A genealogical book may preserve family memory, social prestige, marriage networks or chronological information; but copying,
interpolation, retrospective ordering, honorific claims and missing branches must be investigated.
Oral tradition can preserve living memory and local geography; exact chronology requires cross-checking. ‘Oral’ does not mean
unreliable; ‘an old story’ does not automatically mean historical fact.
Indian Philosophical Dialogue
In the Nyāya tradition, the distinction between pramāṇa and semblances of proof offers a useful discipline for the power–
knowledge question. The issue is not merely whether the speaker is prestigious; the conditions of validity and defects of
perception, inference, comparison and testimony must each be examined.
In verbal testimony, the question of āptatā—the trustworthiness of the speaker—can enter into dialogue with modern ethics of
testimony. Trustworthiness is not identical with social prestige; knowledge, honesty, context, conflicts of interest and
independent corroboration matter.
Buddhist epistemology raises subtle questions about cognitive error, inference and conceptual formation. Its distinction between
experience and conceptual construction can enter into dialogue with modern critiques of classification.
Mīmāṃsā develops sophisticated rules for textual authority, but in modern public reason the claims of traditional authority
must be re-examined in terms of reasons accessible to a plural citizenry with equal civic standing.
Jain anekāntavāda disciplines us to see multiple aspects; it does not mean that all contradictory claims are equally true. The
conditionality of standpoints does not magically dissolve logical contradiction.
Navya-Nyāya analyses of delimiters, relations, qualificands and qualifiers can inspire greater precision in defining categories:
‘about whom?’, ‘in what relation?’, ‘at what time?’, ‘under what condition?’ Such clarity reduces political ambiguity.
Mithila-Centred Parallel Application
When writing the history of Mithila, Vajji, Anga or the Nepal–India border region, do not project modern administrative
boundaries directly onto the past. Place inscriptions, manuscripts, gazetteers, oral memory, linguistic geography, river routes,
pilgrimage networks, land records and archaeological evidence on maps appropriate to their different periods.
The archives of prestigious local families may be valuable, but preservation itself can also be an indicator of power. Families
able to preserve records become more visible in history; the absence of a community whose records were destroyed is not
evidence that it had ‘no history’.
Examine linguistic power. Sources in Maithili, Sanskrit, Persian, Urdu, Hindi, Nepali, English or local speech forms may reveal
different administrative and cultural contexts. Do not treat an archive in one language as the voice of an entire region.
Place-names change, river courses change, district boundaries change, scripts change; therefore a failed search does not
necessarily mean the source is absent. Authority files, alias mapping, historical gazetteers and GIS crosswalks are useful.
The Parallel view of history will preserve both local pride and self-critique: where the evidence is strong, take pride; where it is
weak, reduce confidence. Regional identity is not a substitute for factual standards.
Pūrvapakṣa
Pūrvapakṣa 1: ‘All knowledge is constructed by power; therefore objective truth is a myth.’ Reply: construction explains the
context of production; it does not erase truth conditions. A claim can still be inconsistent with the world or the evidence.
Pūrvapakṣa 2: ‘Victors write history; therefore the defeated side’s account is automatically true.’ Reply: victors’ bias is real;
sources from the defeated can provide counter-evidence, but they are not automatically infallible. Both require source criticism.
Pūrvapakṣa 3: ‘Experts are elites; popular experience is superior.’ Reply: local experience can reveal overlooked dimensions;
specialised methods may nevertheless be necessary for complex causal inference. Collaborative expertise is better.
Pūrvapakṣa 4: ‘Data are neutral; narratives are biased.’ Reply: data collection, coding, missingness, denominators and model
selection are affected by values and assumptions; even so, transparent measurement imposes greater constraints than narrative
alone.
Pūrvapakṣa 5: ‘Every truth-claim is a seizure of power; therefore neutral adjudication is impossible.’ Reply: complete neutrality
is difficult, but comparative procedural fairness is possible—open evidence, recusal, rights of reconsideration, reason-giving and
independent review.
Pūrvapakṣa 6: ‘Censorship is justified in the name of truth.’ Reply: narrow rules against demonstrable harm may be justified, but
suppressing disagreement weakens truth-seeking. Rules should be evidence-based rather than viewpoint-based.
Pūrvapakṣa 7: ‘If there is uncertainty, suspend decision.’ Reply: policy often has to operate under uncertainty. Calibrated
confidence, reversible action, monitoring, precaution and update rules are useful.
Pūrvapakṣa 8: ‘Multi-evidentiality will create false balance.’ Reply: no; evidence should be weighted explicitly and unequally. A
peer-reviewed replicated study and an anonymous viral post are not equal.
Uttarapakṣa: Seven Pillars of a Truth-Seeking Institution
1. Access: as equal an access as possible to relevant data and sources.
2. Transparency: make methods, funding, assumptions, edits, versions and conflicts of interest public.
3. Counterargument: protect dissent and shield critics from retaliation.
गजेन्द्र ठाकु र
4. Independence: make the dependency graph of corroborating sources explicit.
5. Correction: correction, retraction, corrigenda and rights of reconsideration must remain living processes.
6. Representation: include diverse experience and expertise to reduce blind spots; diversity is not a substitute for evidentiary
standards.
7. Accountability: make clear who owns the decision, the reasons for it, the date of review and the remedy available to harmed
parties.
A Fifteen-Step Method for Testing Power–Knowledge Claims
1. State the claim precisely; identify whether it is factual, causal, normative, legal or predictive.
2. Define the key terms and categories.
3. Map the stakeholders who may benefit or be harmed by the claim.
4. Identify the primary source and its provenance.
5. Record the source creator’s purpose, incentives, access and constraints.
6. Investigate why sources or voices are missing.
7. Build a dependency graph of source independence.
8. Note uncertainty arising from measurement, coding, translation and versions.
9. Seek the strongest counter-evidence.
10. State alternative hypotheses.
11. Run prediction or consistency tests where possible.
12. Measure the real extent of expert disagreement; do not confuse it with headline disagreement.
13. Disclose conflicts of interest, but do not turn disclosure into an ad hominem conclusion.
14. State the level of confidence—high, medium or low—and give reasons.
15. State the update rule: what new evidence would change the conclusion?
Twenty Application Scenarios
Scenario 1: A government says the unemployment rate has fallen. Examine definitions, the labour-force denominator, survey
redesign, seasonal adjustment and the raw series; apply the same standard to opposition claims.
Scenario 2: A company says its drug is safe. Examine trial registration, adverse events, comparators, dropouts, funding,
regulatory review and independent meta-analysis.
Scenario 3: A community says a historical site ‘belongs to our ancestors’. Oral tradition is important; examine it alongside
archaeology, inscriptions, land records, chronology and competing memories.
Scenario 4: A viral video appears to show police brutality. Examine the full sequence, timestamp, location, metadata, witnesses
and official records; even an authentic video may have been stripped of context.
Scenario 5: A court judgment declares what happened. A legal finding may be authoritative under a stated evidentiary standard,
yet appeal or new forensic evidence may remain possible; understand the limits of the legal standard.
Scenario 6: A university ranking claims to measure ‘excellence’. Examine the metrics, weights, reputation surveys, self-reported
data and field mix. Ranking is not total quality.
Scenario 7: A history book relies only on colonial archives. Administrative detail may be useful; add local, oral and vernacular
sources, but do not automatically discard colonial sources.
Scenario 8: An activist report describes pollution exposure. Advocacy affiliation increases scrutiny; examine sensor calibration,
sampling, laboratory chain of custody and independent readings.
Scenario 9: A corporation-funded climate study. Disclose the funding; independently review the method and results. The identity
of the funder alone does not determine truth-value.
Scenario 10: An AI chatbot gives a confident answer about history. Fluency is not evidence without citation provenance, edition
control, hallucination checks, source access and independent verification.
Scenario 11: A social platform shows a ‘community consensus’ trend. Examine bot activity, recommendation bias, self-selection
in the sample and the silent majority.
Scenario 12: A census category appears to reduce the size of a group. Examine questionnaire wording, coding, migration, self-
identification, non-response and changes in classification.
Scenario 13: A school textbook compresses a disputed event into one sentence. Provide primary sources, historians’
disagreements and the evidentiary basis in a footnote or appendix.
Scenario 14: A hospital labels a patient ‘non-compliant’. Examine structural causes such as transport, cost, language, side effects,
work schedules and informed choice. The label may conceal blame.
Scenario 15: A ‘crime hotspot’ algorithm increases police deployment; more police then produce more recorded crime. Feedback
loops, unreported crime, denominators and bias tests are necessary.
Scenario 16: From an archival absence, someone concludes, ‘such a tradition never existed’. Examine survival bias, cataloguing,
private collections, and variation in script and language; the evidentiary force of absence is context-dependent.
Scenario 17: Expert consensus is reported as 95 per cent. Examine how consensus was measured, the independence of
respondents, the boundaries of expertise and the evidentiary strength of the minority argument. Do not rely on headcount alone.
Scenario 18: A marginalised witness’s testimony is denied by a dominant official. Examine whether credibility discounting is
biased; seek corroborative evidence; do not treat the witness as automatically infallible.
Scenario 19: A political party calls an inconvenient report ‘fake’. Examine the full report, method, source data, correction history
and external review; a partisan label is not a substitute for evidence.
Scenario 20: Parallel Philosophy itself encounters weak evidence for a favoured regional or philosophical thesis. Apply the self-
critique of Chapter 96: narrow the claim, lower confidence, issue an erratum and remain open to future evidence.
Power–Knowledge in the Digital and AI Age
In the digital age, visibility is a major form of power. The first page of search results, recommendation feeds, trending labels,
verified badges and moderation decisions distribute epistemic attention. Attention is not truth, but it affects the practical
probability of finding truth.
In generative AI, the structure of the training corpus determines which languages, authors and perspectives are more strongly
represented. Corpus imbalance shapes output priors; nevertheless, the factual truth of each output must be independently
verified.
A model provider’s system prompts, safety policies, retrieval index, ranking, update date and hidden fine-tuning shape outputs.
In high-stakes use, records of model, version, date and provenance are therefore necessary.
Deepfakes weaken older heuristics of authenticity. ‘I saw it with my own eyes’ or ‘I heard it with my own ears’ is no longer
enough; add cryptographic provenance, chain of custody, independent camera or audio evidence, contextual corroboration and
forensic uncertainty.
In platform governance, the right of reconsideration is part of epistemic fairness. Erroneous moderation does not merely harm
speech; it also affects the visibility of evidence, reputation, research datasets and public memory.
Limits and Self-Critique
The first limit is power reductionism. Power is not the cause of every disagreement; factual error, methodological difference,
random noise and genuine value conflict can also explain disagreement.
गजेन्द्र ठाकु रक समानान्तर दर्शन — खण्ड २
The second limit is cynical symmetry. Saying ‘all sides engage in propaganda’ does not make stronger and weaker evidence
equal. Making asymmetry explicit is part of fairness.
The third limit is identity essentialism. Disagreement within groups is real. It is wrong to treat ‘the women’s viewpoint’, ‘the Dalit
viewpoint’, ‘the scientific viewpoint’ or ‘the local viewpoint’ as a single homogeneous voice.
The fourth limit is anti-institutional romanticism. Institutions possess power, but they can also provide quality control, archival
preservation, standardisation and accountability. Reform is needed, not meaningless destruction.
The fifth limit is procedural fetishism. A transparency checkbox is not enough; publishing a thousand-page data dump is not
public reason. Usable explanation, metadata, tools and access to time and resources are necessary.
The sixth limit is false neutrality. Where harms are unequal, saying ‘both sides are equal’ is not neutrality. Evidence-weighted
adjudication is required.
The seventh limit is the moralisation of fact. A conclusion does not become factually false merely because it is morally
unwelcome. Value-based disagreement should be named clearly.
The eighth limit is self-exemption. Scholars, activists and media critics of power themselves possess powers of prestige, platform,
funding and group loyalty. Apply the same standard to one’s own claims.
Chapter Conclusion
‘Power influences knowledge, but does not make truth arbitrary’ is the balanced maxim of this chapter. Power can alter whose
questions are heard, whose sources survive, whose language carries prestige, and whose evidence circulates. The critique of
power is therefore an indispensable part of epistemology.
But abandoning the concept of truth does not protect the weak; it may instead allow the powerful to say that ‘our truth’ is
sufficient. Justice requires shared factual adjudication, documented evidence, contradiction tests, independent corroboration
and mechanisms of correction.
Parallel Philosophy chooses neither naïve objectivism nor total relativism here. Its ideal is ‘situated, revisable, multi-evidential
objectivity’—an objectivity that acknowledges its own location, interests and language without thereby abandoning evidentiary
standards.
Chapter 96 teaches self-critique, Chapter 97 says that structure is not destiny, Chapter 98 treats language as a limit but not a
prison, and Chapter 99 accepts the real relation between power and knowledge while refusing to let truth become arbitrary.