Full chapter text
The Structure–Agency Cycle: constraint → opportunity → practice → resistance → collective action → institutional change → new
constraint
structure → position/resources → possible options → agency → response → outcome → change in rule/institution/habit → modified
structure
Human beings do not decide in a vacuum. Language, family, caste, class, gender, school, law, market, technology, geography,
memory, institutions and historical inheritance shape the field of their options. Yet this does not eliminate choice. The maxim of
Parallel Philosophy is: structure is real, but structure is not destiny; agency is real, but agency is not groundless omnipotence.
Problem
The first extreme is structural determinism. On this view, the individual becomes almost entirely the product of social location—
caste writes the future, class consciousness, gender the role, language the capacity and technology the behaviour. This view
correctly sees the reality of unequal opportunity, but it can become weak in explaining resistance, innovation, deviation, moral
responsibility and historical change.
The second extreme is voluntarist individualism. On this view, the principal causes of success are ‘hard work’, ‘mindset’ or ‘the
right choice’, while poverty, discrimination, school quality, inheritance, debt, health, networks, law, violence, language barriers
and geography recede into the background. This view recognises agency, but it can turn structural inequality into a personal
story of moral failure.
The third difficulty concerns levels of causation. A student fails an examination—is the cause inadequate preparation, a weak
school, household labour burdens, linguistic mismatch, nutrition, the examination system, digital access, social expectations, or
the relation among all of these? Selecting one cause produces a simple story; reality often requires a multi-level causal account.
The fourth difficulty is that the word ‘structure’ is itself extremely broad. If every stable pattern is called a structure, the
argument can become unfalsifiable. To identify structure, one must specify rules, resource distribution, roles, incentives,
sanctions, access, expectations, networks and mechanisms of reproduction.
The fifth difficulty concerns agency. The mere availability of options is not agency. Knowledge of the options, capacity to bear
risk, social support, legal rights, time, money, language, confidence and expectations about consequences alter the practical
range of agency.
The sixth difficulty concerns exceptions. If some individuals succeed within an oppressive structure, does the structure therefore
not exist? No. Exceptions do not refute a structure; they may reveal exceptional resources, networks, talent, chance, support or
openings within the structure itself.
The seventh difficulty concerns the responsibility of the disadvantaged. Structure influences behaviour, but does that free a
person from moral responsibility? The answer is not binary. Degrees of coercion, the reality of alternatives, information, fear,
dependency and the scale of harm can alter degrees of responsibility.
The eighth difficulty is institutional agency. Universities, corporations, panchayats, religious bodies, states and platforms are not
‘persons’, yet they make decisions, create rules, collect data and distribute benefits and harms. Institutional agency must
therefore be analysed separately from individual intention.
The ninth difficulty is historical path dependence. A past decision—a railway route, language policy, land settlement, caste
classification, water-management regime or software standard—can make future options more expensive or cheaper. An
inherited path is not automatically inevitable, but lock-in can increase the cost of change.
The tenth difficulty concerns cultural meaning. The same material resource can acquire different meanings in different cultures
or communities. Rules are not only external constraints; people internalise them through expectations, prestige, shame,
sacredness, familial obligation and identity.
The eleventh difficulty is language. Categories such as ‘backward’, ‘merit’, ‘normal’, ‘traditional’, ‘modern’, ‘illegal’ and
‘deserving’ can affect the possibilities of individuals and groups. Linguistic structure does not merely describe; it can guide
institutions and behaviour.
The twelfth difficulty is digital structure. Recommendation, ranking, moderation, credit scoring, biometric databases, search
visibility and automated eligibility present ‘choices’, but the order, visibility and cost of those choices are algorithmically
structured. A user’s click is an act of agency, but the architecture of choice is also causally significant.
Core Proposition
Parallel Philosophy treats structure not as an invisible force that mechanically produces fixed outcomes, but as a relational
configuration of recurring rules, distributions of resources, roles, expectations, networks, technological affordances, rewards
and sanctions, and institutional memory. Such a configuration changes the probability distribution of outcomes; it does not
determine every individual result.
Agency is not a pure will standing outside structure. Agency means choosing among available or imaginable options,
reinterpreting rules, cooperating, dissenting, resisting, innovating, exiting, bargaining, building institutions and forming new
habits. Agents are born within structures and use structural languages, yet they can also transform structures.
The Parallel view therefore uses the language of ‘structured possibility’ rather than determinism. Poverty can make higher
education more difficult—this may be a causal claim; ‘poor students cannot obtain higher education’ is a deterministic claim. The
first claim seeks intervention; the second closes the future.
Structure does not merely constrain; it can also enable. Language sets limits but makes communication possible; law imposes
sanctions but secures rights; schools discipline but provide knowledge resources; platforms capture data but can open new
public spheres. The same institution can be both enabling and constraining.
The moral value of agency is not measured only by outcomes. Resistance may fail and still embody agency; obedience may
succeed while diminishing autonomy. The aim of a just society is therefore not merely ‘successful outcomes’, but the
enlargement of real options, dignity, voice, rights of reconsideration and the capacity for collective change.
Twenty-Five Principles of Structure and Agency
1. Structure does not determine outcomes; it distributes the probability, cost and risk of outcomes.
2. Agency does not arise from nothing; it depends upon available resources, language, knowledge, relationships and imaginable
options.
3. Exceptions do not invalidate a structure; the pathway of an exception may reveal openings within the structure or special
resources.
4. An average pattern is not a moral judgement on an individual. Inferring an individual destiny from group statistics can
become an ecological fallacy.
5. Rules need not be written. Prestige, shame, expectation, imitation, silence and informal sanctions can also form structures.
6. Resources are not only money; they include time, safe space, language, credentials, networks, information, mobility, health
and digital access.
7. Power is not only command; it also includes setting the agenda, making options invisible, creating categories and setting
defaults.
8. Enabling structures remain open to criticism; institutions that provide benefits can also reproduce inequality.
गजेन्द्र ठाकु र
9. Agency should not be measured merely by the fact that ‘a choice was made’. Ask whether the options were real, informed and
accompanied by a meaningful possibility of exit.
10. Path dependence is not destiny; inherited pathways raise the cost of change, but they do not make change impossible.
11. Habit and habitus can be causes rather than character defects. Social learning and embodied routine may require policy
interventions of their own.
12. Resistance can itself become structure—movements, unions, women’s groups, cooperatives and open-source communities
create new rules.
13. Local autonomy does not automatically guarantee justice; rights of reconsideration are necessary where local domination
occurs.
14. Central law does not automatically guarantee equality; implementation networks, bureaucracy and local practice alter
outcomes.
15. Language creates categories, but a category is not the whole reality of its object. Keep labels revisable.
16. Technology is not merely a neutral tool; interfaces, defaults, interoperability and data requirements structure behaviour.
17. An algorithmic output does not remove responsibility under the phrase ‘the system decided’. Identify the human institutions
responsible for design, deployment, oversight and appeal.
18. Moral responsibility can be graduated according to coercion, ignorance, dependency, capacity and foreseeable harm.
19. Structural explanation is not the same as excuse; explanation refines the distribution of responsibility.
20. Institutional change without individual change in practice is limited; individual reform without institutional change is also
limited.
21. Seek leverage points for change—rules, resources, information, networks, defaults, sanctions, narratives, representation or
exit.
22. Make claims about structure empirical: specify the mechanism, indicator, comparison, timescale and counterfactual.
23. ‘Tradition’ is not a uniform structure; seek internal disagreement, reinterpretation, regional variation and change over time.
24. ‘Market’, ‘state’, ‘society’, ‘caste’ and ‘family’ are not singular actors; examine internal levels and conflicts.
25. The highest maxim is this: arrangements made by human beings can be changed, but the cost, time and risk of change
require realistic assessment.
Seven Levels of Causation
Level 1 — Biological/physical: body, age, illness, disability, nutrition, sleep and stress. Social structure can influence this level,
but biological facts are not merely social imagination.
Level 2 — Individual skill/belief: knowledge, habit, confidence, fear, memory and goals. Even when socially learned, these may
require intervention at the individual level.
Level 3 — Relationship/family/network: support, care, violence, recommendation, marriage, debt, acquaintance and informal
information.
Level 4 — Institution: school, hospital, bank, panchayat, court, company, religious body, university, media and platform.
Level 5 — Social classification: caste, class, gender, language, religion, citizenship, region, disability and age—categories upon
which stereotypes, rules or resource allocation may rest.
Level 6 — Material/technological structure: roads, rivers, electricity, housing, internet, devices, databases, software standards,
transport and flood control.
Level 7 — Historical/political economy: land regimes, colonial classifications, borders, migration regimes, capital accumulation,
state capacity and climate change.
These seven levels are not competing explanations; they often form a joint causal chain. Instead of asking for the one ‘correct
level’, it is more fruitful to ask which mechanism operates at which level.
Pūrvapakṣa and Uttarapakṣa
Pūrvapakṣa 1: if structure is not destiny, the expression ‘structural injustice’ becomes weaker.
Reply: no. Structural injustice does not claim that every individual outcome is predetermined; it claims that opportunities, risks,
respect or resources are systematically distributed unequally. Probabilistic effects can also be morally serious.
Pūrvapakṣa 2: emphasising agency can become victim-blaming.
Reply: recognising agency is not victim-blaming when constraints, coercion and resource gaps are kept explicit. Treating the
disadvantaged only as passive objects can itself diminish dignity.
Pūrvapakṣa 3: structural explanation will excuse individual wrongdoers.
Reply: explanation and justification are different. Social pressure can explain violence without making violence right. Degrees of
culpability can vary with context.
Pūrvapakṣa 4: if everything is called multi-level, policy will become vague.
Reply: multi-level explanation does not prevent prioritising interventions. Leverage points can be chosen by effect size,
feasibility, rights, cost, reversibility and distributional consequences.
Pūrvapakṣa 5: merit is ultimately an individual capacity.
Reply: capacity is real, but opportunity, training, language, nutrition, time, test design and inherited advantage shape
performance. Treating merit as a context-free substance is misleading.
Pūrvapakṣa 6: if people themselves accept a tradition, how much structural pressure can there be?
Reply: internalised norms can be experienced as voluntary. The quality of consent should be tested against exit costs,
alternatives, childhood socialisation, sanctions and information.
Pūrvapakṣa 7: market choice is the best evidence of agency.
Reply: market choices are structured by resources, information, monopoly, defaults, advertising, credit and bargaining power. A
purchase can be an exercise of agency without amounting to complete autonomy.
Pūrvapakṣa 8: an algorithm merely learns user preferences.
Reply: data selection, objective functions, ranking, interface design, moderation, monetisation and feedback loops also shape
preferences. Preference is both an input and an output.
Pūrvapakṣa 9: once a law is enacted, the structure has changed.
Reply: formal rules are only one level. Without enforcement, awareness, bureaucracy, access to courts and changes in informal
practice, outcomes may remain limited.
Pūrvapakṣa 10: if structures are made by human beings, why can they not be changed immediately?
Reply: coordination problems, sunk costs, vested interests, network effects, habits, fear, uncertainty, material infrastructure and
unequal power raise the cost of change.
Pūrvapakṣa 11: successful exceptions are inspiring, so highlighting them is appropriate.
Reply: inspiration can be useful, but exceptions should not be treated as general policy evidence. Survivor bias can conceal
structural barriers.
Pūrvapakṣa 12: identity categories are necessary for measuring structural injustice; why criticise categories?
Reply: categories are useful operational tools, but they simplify historically heterogeneous realities. Use them for measurement
without turning them into essential destinies.
Pūrvapakṣa 13: local communities should determine their own rules.
गजेन्द्र ठाकु रक समानान्तर दर्शन — खण्ड २
Reply: subsidiarity is important, but basic dignity, protection from violence, equal civic rights and rights of reconsideration may
require safeguards above a local majority.
Pūrvapakṣa 14: if structure is everywhere, freedom is an illusion.
Reply: freedom need not mean absolutely unconditioned choice. The capacity to reflect, revise, resist, cooperate and alter
institutions within a causal world can provide a sufficient substance of freedom.
Pūrvapakṣa 15: structural change is not always progressive.
Reply: correct. Change can create new forms of domination. That is why impact testing, dissent, reversibility, pilots, monitoring
and correction are necessary.
Indian Philosophical Dialogue
In the Nyāya tradition, agency, desire, effort, knowledge, defects and results are considered within causal sequences. Moral and
practical action is not treated as isolated will; relations among cognition, desire, effort and circumstance are examined. The
Parallel view benefits from this analytical discipline while adding the levels of modern institutional power and social
classification.
Buddhist pratītyasamutpāda, or dependent origination, teaches relational arising against the idea of a single self-sufficient cause.
Its conditional form—‘when this is, that comes to be’—can enter into dialogue with structural thinking, although the Buddhist
metaphysical project is not identical with modern theories of social structure.
Jain anekāntavāda cautions against treating an object or claim as exhausted by a single standpoint. This is useful in the
structure–agency debate: a person is neither merely structure nor merely an independent agent; different nayas disclose
different aspects.
Mīmāṃsā offers subtle analyses of action, injunction, entitlement, context and linguistic expectancy. In understanding the causal
power of social rules, it is worth remembering that the force of a rule lies not only in a written sentence; eligibility, context,
tradition and structures of compliance also matter.
The Bhagavad Gītā raises the questions of prakṛti, guṇa, action and agency in complex form, and different philosophical
traditions interpret the text differently. Parallel Philosophy does not turn the Gītā into a direct theory of modern sociology, but
treats it as a historical site of dialogue about conditioned action and self-understanding.
The Sāṃkhya distinction between prakṛti and puruṣa is not equivalent to the modern structure–agency distinction; nevertheless,
the contrast between conditioned processes and witnessing discernment can open comparative questions about the
philosophical imagination of agency. The purpose of comparison is conceptual contrast, not a declaration of identity.
Indian Dharmaśāstra traditions can be read as articulating normative structures concerning varṇa, āśrama, family, inheritance,
marriage and obligation; yet historical practice varied by region, period, caste, political authority, economy and local custom.
Text does not equal the whole of social reality.
The rise of bhakti, śramaṇa traditions, folk traditions and regional languages shows that new religious and literary agency can
emerge within established linguistic and institutional structures. Resistance is not always anti-institutional; it can also create
new institutions.
Mithila’s Parallel Perspective
Mithila is a multi-level field for studying structure and agency: the physical geography of rivers and floods, the India–Nepal
border, rural settlement, agriculture, migration, education, linguistic plurality, caste institutions, Panji memory, religious
centres, towns and a digital diaspora all shape the range of agency at the same time.
Treating floods merely as ‘natural disasters’ is incomplete. River systems, embankments, land use, roads, relief arrangements,
administration, poverty, migration networks and household strategies together shape vulnerability. A family’s migration can be
an act of agency while still containing elements of forced choice.
Panji and genealogical records can be analysed both as historical sources and as social institutions. Written classifications can
affect marriage networks, memory and prestige; yet no individual life is wholly determined by a Panji category. Dates, copying
history, usage and historical change require examination.
Maithili, Bajjika, Angika, Nepali, Hindi, Sanskrit, Urdu and English open different kinds of opportunity. The language of
schooling, justice and administration can become a structural gatekeeper; multilingual capacity can expand agency; low-
resource digital status can create new barriers.
Migration can alter income, aspiration, gender roles, elder care, education and housing in Mithila families. Migrants do not move
‘outside structure’; they develop new networks of agency between the structures of origin and destination.
The expansion of women’s education and employment can result from the combined effects of law, schools, household
bargaining, safe transport, digital access, role models and labour markets. Neither ‘tradition’ nor ‘personal ambition’ alone is an
adequate explanation.
Discussion of caste inequality should identify structural patterns without imprisoning individuals within categorical stereotypes.
Changes in historical categories, locality, class, education, gender, migration and inter-group relations produce intersectional
differences.
A digital platform such as Videha can alter one part of publication gatekeeping: geographical distance is reduced, archives
become searchable and Maithili material gains global reach. Yet devices, bandwidth, search ranking, font support, accessibility
and platform dependence create new structures.
Philosophical and historical writing about Mithila bears an ethical responsibility: make regional pride a source of agency, but
not an evidence-resistant narrative of destiny. Parallel Philosophy does not accept the claim, ‘our tradition has always been this
way, therefore it must remain this way.’
A Practical Protocol for Testing Structure
Step 1 — Clarify the outcome: what result, behaviour or inequality is to be understood? Avoid the vague statement ‘society is like
this’.
Step 2 — Identify the actors: individuals, families, groups, institutions, platforms, the state, markets and networks—which
decisions are relevant?
Step 3 — State the rules: distinguish formal law or policy from informal expectations and sanctions.
Step 4 — Map the resources: money, time, education, language, mobility, networks, data, devices, care and legal access.
Step 5 — Map the option set: what real alternatives were available to the actor? Distinguish theoretical options from affordable
and accessible ones.
Step 6 — Identify the mechanism: how do rules or resources affect outcomes—through incentives, exclusion, information,
coordination, stigma, defaults, violence, dependency or network effects?
Step 7 — Compare: does the pattern change across regions, periods, groups or institutions? If so, which part of the structure is
causally relevant?
Step 8 — Test exceptions: trace the pathway of successful or unsuccessful atypical cases and avoid survivor bias.
Step 9 — Trace feedback loops: does an outcome strengthen or weaken the structure? For example, credit score → loan denial →
low investment → future score.
Step 10 — Locate points of agency: how much leverage exists at individual, collective and institutional levels—education, unions,
appeal rights, technological redesign, law, budgets, narratives or exit?
गजेन्द्र ठाकु र
Step 11 — Map responsibility: designers, implementers, supervisors, beneficiaries, actors and regulators—how much
accountability belongs to each?
Step 12 — Define revision signals: which indicators will be monitored after intervention? Establish in advance how policy will
change if unintended harms appear.
Structure in the Age of Artificial Intelligence and the Digital World
Digital structures can be invisible. At the interface, users may see only ‘recommended’, ‘eligible’, ‘trending’, ‘safe’ or ‘verified’;
behind these labels lie ranking objectives, moderation rules, training data, thresholds, business models and legal requirements.
Structural analysis must therefore move from front-end choice to back-end governance.
Recommendation feedback loops do not merely measure user preferences; by changing exposure, they can shape future
preferences. ‘People want to see this’ is therefore an incomplete explanation. Exposure history is an independent causal variable.
In low-resource languages, scarce data, weak moderation, higher speech-recognition error and lower search visibility directly
affect agency. A Maithili user may formally be able to ‘choose the language’, while software support limits the actual option set.
Automated hiring, credit or benefit systems can learn patterns from historical data and reproduce structural inequality. Fairness
is not achieved merely by removing a protected attribute; proxy variables, targets, label quality, error distribution and appeal
rights must also be examined.
Generative AI can expand a writer’s agency through translation, drafting, search strategy and accessibility, but model defaults,
training representation, refusal policies, hallucination and platform lock-in can also become cognitive structures. Human
verification remains an indispensable part of agency.
Interoperability and open formats can increase structural freedom. If books or data are locked in proprietary formats, future
access depends upon a platform; open standards reduce exit costs.
Digital identity and biometric systems can make inclusion easier, but authentication failure can become denial of service for
vulnerable people. Structural design therefore requires fallback channels, human review and grievance routes.
The correct formula for AI governance is this: writing ‘human in the loop’ is not enough. Is human authority real? Is there
sufficient time and information? Is override safe? Is the appeal mechanism independent? Is responsibility traceable? Without
these, human presence becomes merely symbolic agency.
Structure–Agency Scenarios
Scenario 1: a poor student passes a top examination. The conclusion ‘poverty is not a barrier’ is wrong. Trace scholarships,
mentors, libraries, family support, unusual effort and selection pathways, and then return to population-level data.
Scenario 2: a woman ‘chooses’ not to work outside the home. Analysis of the choice should examine income alternatives,
childcare, transport safety, family sanctions, property rights, prior socialisation and personal values.
Scenario 3: a male relative makes the decisions for a woman occupying a reserved panchayat seat. The formal structure has
changed while informal authority remains old. Policy evaluation should measure actual participation in decision-making.
Scenario 4: caste discrimination is prohibited by law, but surnames or networks are used as proxies in the rental market. Formal
rules and informal screening must be examined together.
Scenario 5: a platform says toxic material goes viral because users demand it. Examine the structural contribution of ranking
objectives, engagement incentives, recommendation cascades and moderation delays.
Scenario 6: a migrant worker moves to a city and increases income but has weak access to housing, identity documents or health
care. Agency has increased in one dimension while vulnerability has increased in another.
Scenario 7: a family in a flood-prone region repeatedly rebuilds its home. Before asking ‘why do they not move?’, examine land
prices, livelihood, kin networks, schooling, compensation, trust in relocation and cultural attachment.
Scenario 8: a school offers a mother-tongue option on paper but has neither teachers nor books. Formal choice has not become
real agency.
Scenario 9: a university claims admission is based on ‘merit only’, while access to coaching is unequal. Examine test validity,
variation in schooling, language, disability accommodation and the cost of preparation.
Scenario 10: a reformer within a religious community offers a new interpretation of scripture. Agency does not stand outside
tradition; reinterpretation within the language of tradition can become a mechanism of structural change.
Scenario 11: an algorithm rejects a loan and staff say ‘the system decided’. Map the training target, bureau data, threshold,
override rules, regulator and vendor responsibility.
Scenario 12: a rural entrepreneur gains market access through the internet, but income falls when platform commission or
visibility rules change. Digital agency depends upon platform structure.
Scenario 13: early marriage declines in a society. The cause is not simply a ‘modern mindset’; schooling, law, transport,
employment, media, peer norms, fertility preferences and household bargaining act together.
Scenario 14: an old sub-caste label becomes stabilised in a community during the nineteenth or twentieth century. Its present
social effects can be real, but it is wrong to turn the historical category into an eternal destiny.
Scenario 15: an employee obeys an unethical order for fear of losing a job. Coercion may reduce culpability, but in cases of
severe harm the expectation of refusal or whistleblowing does not fall to zero.
Scenario 16: an open-source tool expands support for a local language. Volunteer agency alters a structural gap in the software
ecosystem and creates new infrastructure.
Scenario 17: a reservation policy increases representation, but workplace culture remains exclusionary. The entry structure has
changed while retention and promotion structures have not; multi-stage reform is necessary.
Scenario 18: a climate-adaptation grant is available, but applications are digital-only. A nominally universal policy produces
unequal uptake because digital literacy and device access are structurally unequal.
Scenario 19: a family educates a daughter but denies her inheritance. One structure has liberalised while property relations
remain old; ‘the family has changed’ is not a complete claim.
Scenario 20: Parallel Philosophy itself becomes popular and followers turn its maxims into rigid doctrine. Apply Chapter 96’s
self-critique and Chapter 97’s principle of structure: community norms must not turn the author’s original openness into destiny;
dissent, new editions and revision should be institutionalised.
Strategies for Structural Change
1. Change the rule, but build implementation capacity at the same time.
2. Provide resources, but measure the friction of access.
3. Change defaults, but protect informed opt-out.
4. Increase representation, but test whether decision-making power has actually been transferred.
5. Provide transparency, but add usable explanation and rights of reconsideration.
6. Provide local autonomy, but preserve a floor of rights and external review.
7. Provide education, but do not forget structures of labour markets, transport and care.
8. Provide technology, but examine maintenance, language, accessibility, privacy and lock-in.
9. Change narratives against stigma, but also change rules and resources that produce material discrimination.
10. Use pilots or experiments, and measure distributional effects and unintended harms.
11. Support collective agency—associations, unions, cooperatives, community archives and peer networks.
12. Measure the success of reform not by ‘policy adopted’, but by real options, outcomes, dignity, voice and reversibility.
गजेन्द्र ठाकु रक समानान्तर दर्शन — खण्ड २
Limits and Self-Critique
The first limit is over-structuralisation. Not every individual failure is society’s fault; skill, decision, chance, temperament, health
and effort can be real causes. A structural lens must not replace evidence.
The second limit is romantic agency. Saying ‘people can resist’ does not make the cost of oppression small. The risks of resistance
—violence, job loss, exclusion, litigation and family rupture—must be made explicit.
The third limit is category reification. Categories created for analysis—Dalit, woman, migrant, rural, minority, youth—can
conceal internal diversity. Examine subgroups, interactions and self-identification.
The fourth limit is causal storytelling. Words such as ‘power’, ‘structure’ and ‘culture’ can become theatrical explanations
without mechanisms. Measurable or traceable pathways are required.
The fifth limit is policy hubris. Interventions that change structure generate adaptation, resistance and unintended incentives.
Feedback monitoring is indispensable.
The sixth limit is moral simplification. Powerful actors are not always completely free; marginal actors are not always innocent.
Power asymmetry changes responsibility, but does not end moral examination.
The seventh limit is temporal scale. Structural change can be slow, while emergency harms require immediate relief. Long-term
reform and short-term protection must proceed together.
Chapter Conclusion
‘Structure exists, but structure is not destiny’ is not merely an optimistic slogan of Parallel Philosophy; it is a discipline of causal
analysis. It holds that social patterns are real, measurable and morally significant, but they do not turn persons into perfectly
predictable objects.
Respect for agency does not mean denying structural injustice. On the contrary, the purpose of identifying injustice is to enlarge
the real range of agency—through more information, resources, security, voice, exit, rights of reconsideration, association and
the capacity to change institutions.
The test of structural reform is not only the average outcome: whose options increased? Whose risk fell? Whose voice was
heard? Whose dignity was protected? Who acquired a new burden? Is the reform reversible and contestable?
Chapter 96 teaches that philosophy itself is not outside criticism. Chapter 97 extends that lesson socially: no institution, tradition,
category, technology or historical arrangement can be declared destiny merely because ‘it has always been this way’.
The combined maxim of this chapter is: ‘Structure makes the field of our options; agency makes possible play within the field,
new meanings for its rules, collective resistance, and sometimes the reconstruction of the field itself.’