Map of the Two Paths of Science Natural science — measurement • experiment • law • causal structure Social science — meaning • institution • history • agency Shared discipline — evidence • causation • retesting • transparency Key difference — the social world can change itself through its own understanding of meaning and through its reactions Problem The relationship between natural science and social science is a long-standing question in modern philosophy. Disciplines such as physics, chemistry, biology and geology measure natural phenomena, conduct experiments, construct models and offer causal explanations. Sociology, economics, political science, anthropology, history and social psychology likewise rely on evidence, data, comparison and explanation; but their subject matter includes human beings, institutions, rules, language, meanings, beliefs, values and historical memory. Here the object of research itself holds views about its world, understands rules, resists, changes strategy and may react to the researcher’s classifications. The question, therefore, is not merely, “Are both sciences or not?” The question is how common standards of scientific rigour should be applied to different kinds of subject matter. One old position holds that the social sciences should imitate the methods of the natural sciences. In this naturalist view, causation, laws, measurement, testing and prediction are treated as ideals. A second position argues that human action is not merely an external event; its meaning, purpose, understanding of rules and historical context must also be grasped. In this interpretivist view, understanding the meaning of action becomes central to social explanation. The problem arises when the first side seeks to remove meaning as unscientific, or when the second side rejects causal testing as hostile to human understanding. Max Weber’s interpretive understanding, Verstehen, shows the importance of grasping the meaning of social action. But it would be wrong to reduce it to empathy or to the researcher’s private experience. Understanding meaning must be disciplined by historical documents, linguistic behaviour, consequences of action, institutional rules and comparative evidence. The question “Why did this person act in this way?” can be examined through the agent’s stated reasons, actual incentives, available alternatives and social pressures together. A second axis of the dispute is the relation between individual and structure. Methodological individualism insists that social phenomena be connected to individual-level action and intention. Yet institutions such as family, market, caste, law, state, school, currency, language or corporation are not simply private wishes of individuals; they create structures of rules, rights, resources and expectations that alter the possibilities of individual action. Understanding the relation between micro and macro levels is therefore a central challenge of social science. A third axis concerns objectivity and value. Choosing a research topic, defining poverty, crime, development, literacy, employment, caste or family, and fixing a threshold of risk all involve human purposes and institutional priorities. But it does not follow that findings are arbitrary. A value-influenced question and an evidence-free answer are different things. The task of scientific discipline is to make the relation among concepts, data, method and conclusion publicly testable. A fourth axis concerns prediction and explanation. The timing of an eclipse can be predicted with great precision; the exact prediction of elections, riots, migration, inflation or changing forms of marriage may be much more difficult. This difficulty is not a reason to deny the status of science to the social sciences. Complexity, open systems, agents’ reactions, institutional change and historical contingency limit prediction. Success in social science may sometimes be measured by precise prediction, and at other times by causal process, comparative boundary conditions or meaningful explanation. Core Proposition The central proposition of this chapter is that natural science and social science share an epistemic discipline, but not one universally valid method. Both require evidence, causation, error correction, clear definitions, independent checking and public criticism. Social science, however, includes additional dimensions—meaning, rules, institutions, agency, history, values and reflexive reaction—that alter the choice of method. For that reason a balance is needed between methodological unity and methodological plurality. Unity means that data cannot be fabricated, contrary evidence cannot be concealed, causal claims cannot be made without grounds, and theories are not exempt from criticism. Plurality means that experiments, surveys, archives, interviews, ethnography, historical comparison, statistical analysis, textual reading, fieldwork and causal modelling play different roles for different questions. The subject matter of social science is a meaningful causal world. Human beings are subject to causes, yet they also interpret causes; they live within rules, yet can change rules; they are shaped by institutions, yet can reconstruct institutions. Mere imitation of natural laws is therefore insufficient, and free-floating interpretation is insufficient as well. Parallel Philosophy will read this relationship at five levels: objective conditions, the agent’s meaning, institutional rules, historical pathways and public testing of evidence. Combining these five levels prevents social science from becoming either a weak imitation of natural science or an untestable exercise conducted in the name of literary interpretation. Principal Arguments First argument—scientificity is not the name of a single tool; it is the discipline of correcting error. Experiment may be central in natural science; in social science a natural experiment, survey, archival comparison or field study may sometimes be more appropriate. The question is not whether an experiment occurred, but whether the claim is supported by the right kind of evidence. Second argument—cause and meaning are not competitors. An unemployed person may leave a town to migrate because of income opportunities, while family obligations, prestige, hope for the future and social imaginaries of migration form the meaning of the action. Meaning without economic structure is incomplete; structure without human action is equally incomplete. Third argument—Verstehen is not private guesswork. It is not enough for the researcher to say, “I understand.” Interpretation of meaning must be tested against agents’ statements, behaviour, historical context, institutional rules and contrary evidence. Interpretation too must be disciplined by evidence. Fourth argument—methodological individualism is not political individualism. To say that social outcomes are generated through individual action is not to say that institutions or structures are unreal. Individuals act within language, rules, markets, families and histories. The search for micro-foundations is useful, but micro-foundation does not mean the abolition of structure. Fifth argument—large-scale social structures can play causal roles. Laws, reservation policies, tax systems, border controls, school-language policies or banking rules alter the costs and possibilities of individual choices. Structure should be analysed not as a mysterious entity but as an organised arrangement of rules, resources, sanctions, rights and expectations. Sixth argument—an intervening mechanism is needed between micro and macro levels. Statements such as “poverty increases crime” or “individuals decide” are incomplete by themselves. Intermediate processes—employment opportunities, social networks, state presence, schools, credit, stigma and distribution of risk—must be specified. Such mechanism-based explanation makes social causation more testable. Seventh argument—correlation is not causation. The fact that two variables change together does not establish causal direction. Reverse causation, hidden third factors, selection bias and measurement error must be examined. Counterfactuals, natural गजेन्द्र ठाकु र experiments, randomisation, longitudinal data or other identification strategies may be useful for causal inference in social science, but each has its own limits. Eighth argument—a randomised controlled trial is a powerful tool, not a universal answer. An intervention that succeeds in a small controlled group may not produce the same result in another region, language, caste structure, institutional capacity or period. Internal validity and external validity require separate examination. Ninth argument—observational data are not merely weak evidence. Historical records, censuses, archives, longitudinal surveys, court records, land registers or institutional data may be the only possible sources for many questions. The difficulty lies in source selection, incompleteness, changes in definition and causal interpretation; source criticism is therefore essential. Tenth argument—measurement can influence social reality. When a poverty line, school-performance score, productivity target, crime rate or ethnic category becomes an administrative goal, institutions and individuals may change their behaviour in response to the measure. A social indicator is therefore not merely a mirror; at times it also becomes an incentive. Eleventh argument—categories themselves can become reflexive. When people learn the social or medical classification applied to them, they may accept it, reject it or use it strategically. This creates an important difference from the ordinary classification of natural objects. Twelfth argument—objectivity is not the name of a value-free human being. A researcher’s social position, choice of questions and institutional interests can exert influence. Objectivity becomes stronger when definitions are clear, data are available, competing analyses are possible, methods are transparent and an active critical community exists. Thirteenth argument—the distinction between fact and value is useful, but complete separation is difficult. “How many children die?” is a factual question; “How much risk is acceptable?” is a normative one. Policy research should not collapse the two, but neither should it hide values under the name of fact. A sound method makes value judgements explicit. Fourteenth argument—quantitative and qualitative methods are not opposing camps. A survey may show how much migration occurs; interviews may reveal what migration means to people; archives may show how policy changed; fieldwork may reveal how an institution actually works. Combined use can make the question clearer. Fifteenth argument—large numbers do not automatically produce objectivity. Poor definitions, biased samples, missing data or faulty causal reasoning can yield false conclusions even after millions of rows of data. Numbers are evidence; they are not substitutes for definition and method. Sixteenth argument—a small field study does not automatically yield universal truth either. An intensive study of one village, institution or community may uncover causal processes and meanings, but the limits of generalisation must be stated clearly. Depth and breadth are distinct epistemic virtues. Seventeenth argument—history is an integral part of social causation. The same present structure can arise through different historical paths. Land relations, language policy, caste categories, borders or urbanisation remain incompletely understood without historical sequence. Path dependence does not mean that the past determines everything; it means that previous arrangements alter the costs of present options. Eighteenth argument—prediction and explanation are different achievements. A data-rich model may predict voting or credit risk accurately without clarifying the cause. Conversely, a deep historical study may explain a causal process without predicting individual outcomes precisely. These achievements should be recorded separately. Nineteenth argument—social laws are often domain-bounded. Hard laws of the form “if X, then always Y” may be rare, but this does not make causal science impossible. Conditional regularities, tendencies, mechanisms, probabilities and institutional boundary conditions can provide sufficient grounds for scientific explanation. Twentieth argument—the Parallel method favours mixed evidence. Social claims become stronger when statistical regularities, causal identification, agents’ meanings, institutional records, historical sequences and counter-examples are considered together. The maxim is: social facts are meaningful; meaning is not evidence-free. Pūrvapakṣa First Pūrvapakṣa: Social science has fewer experiments, so it is not a real science. Reply—experiment is an important instrument of scientific inquiry, not its sole criterion. Geology, astronomy and evolutionary biology also answer many questions through observation, comparison and historical evidence. Social-scientific claims should be judged by the appropriateness of their evidence and their capacity for error correction. Second Pūrvapakṣa: Human beings are free, therefore a causal science of human action is impossible. Reply—agency does not mean absence of causation. Desires, rules, resources, information, institutions and history influence action. Freedom itself may be a mode of choosing among options within a causal structure. Third Pūrvapakṣa: Meaning is the researcher’s interpretation, so social science is wholly subjective. Reply—interpretation of meaning can be tested against sources, language, behaviour, institutional rules, multiple interviews and contrary evidence. Interpretation can be made corrigible. Fourth Pūrvapakṣa: Social structure is only the sum of individuals. Reply—institutions are made by individuals, but organised rules, offices, rights and resources generate properties that belong to no isolated individual. The relation between individual level and structural level must be made explicit. Fifth Pūrvapakṣa: Once values enter research, objectivity ends. Reply—values may influence the choice of questions and policy goals; nevertheless data, definitions and causal conclusions can still be publicly tested. Hiding values is not objectivity; making their place explicit is a more honest method. Sixth Pūrvapakṣa: Quantitative method is rigorous, qualitative method is merely storytelling. Reply—bad statistics are not rigorous, and disciplined ethnography is not mere storytelling. The appropriate form of evidence depends on the question. Causal measurement combined with interpretive depth can produce fuller knowledge. Seventh Pūrvapakṣa: Every society is unique, therefore general social science is impossible. Reply—comparable processes can exist within unique histories. Generalisations should state their limits of region, time, institution and group; if knowledge is not universal, it can still be conditionally valid. Eighth Pūrvapakṣa: Social science is a product of power structures, so its truth claims are untrustworthy. Reply—power can influence funding, classification and the selection of questions, but this criticism itself requires evidence. The answer to institutional bias is greater transparency, plurality of sources, counter-position and public criticism—not abandonment of truth. Uttarapakṣa First reply—let method follow the object. It is improper to impose a single instrument on natural phenomena, human action, institutions and historical change alike. Experiment, comparison, archives, statistics or interpretation should be chosen according to the nature of the question. Second reply—write causal and interpretive explanations separately, then connect them. “What caused this?” and “What meaning did the agent give it?” are two questions that do not replace one another. Their conjunction is a distinctive strength of social science. Third reply—make the bridge between micro and macro levels explicit. Show the process through which individual action produces institutions and institutions shape individual action. Loose statements such as “society does this” or “the individual chooses” are not enough. Fourth reply—make the provenance chain of measurement public. How was an indicator defined, how were data collected, how were missing values handled, and why was a category created? These should be treated as part of the conclusion itself. गजेन्द्र ठाकु रक समानान्तर दर्शन — खण्ड २ Fifth reply—seek alternative explanations for causal claims. Without testing reverse causation, hidden factors, selection bias, historical shocks and institutional change, causal reasoning remains incomplete. Sixth reply—seek a counter-position for interpretive claims. Do not treat a single statement by an agent as final truth; examine behaviour, other evidence, documents, silences and competing structures of meaning. Seventh reply—state the limits of generalisation. Make clear the population, period, region, institutional arrangement or policy conditions within which a conclusion is supported. Limited truth is not failure; concealing its limits is. Eighth reply—understand objectivity as multi-sided criticism. When researchers with different methods, social locations, sources and theories test one another’s claims, the effects of individual bias can be reduced. Objectivity is not a lifeless viewpoint; it is a disciplined public process. Indian Dialogue Classical Indian philosophical traditions do not contain a ready-made precursor of the modern division between natural science and social science. That division is tied to the history of modern institutional science. Yet classical reflections on evidence, inference, testimony, action, rules, purpose and meaning can open useful dialogue at the level of problems. The Nyāya tradition developed a discipline for distinguishing perception, inference, comparison and testimony and for examining the sources and errors of knowledge claims. Its relevance to modern social science does not lie in saying “Nyāya = modern methodology”, but in asking what kind of evidence supports a claim, what defeaters are available and how far an inference may legitimately extend. Mīmāṃsā developed complex reflections on language, injunction and the meaning of action, but it would be anachronistic to call it a precursor of modern interpretive sociology. At the level of problems, its useful insight is that linguistic rules and meanings structure human action; social science cannot read rule-following and intention merely as outward motion. Arthaśāstra, Dharmaśāstra, inscriptions, royal bureaucratic traditions and historical narratives can provide sources on social life, but they are not neutral collections of modern social-scientific data. Their genre, purpose, power context, silences and normative language must be examined before use. The fact that an old text records a social fact does not make it equivalent to a modern survey. Buddhist and Jain reflections on causal relations, standpoints, conduct and community are likewise not directly identical to modern theories. The appropriate axes of comparison are the problems themselves—individual and relation, agent and condition, standpoint and claim, cause and consequence. Dialogue becomes meaningful only when historical difference is preserved. The chief conclusion of the Indian Dialogue is that a discipline of plural evidence is useful for social science. Direct data, inference, credible testimony, archives, behaviour, linguistic meaning and contrary evidence may carry different weights. Yet no classical list of pramāṇas can replace modern method; it can only provide methodological inspiration. Mithila’s Parallel Perspective Mithila can serve as a distinctive testing ground for social science because natural processes and social structures are interwoven in many areas. Flooding is a hydrological phenomenon, but its damage is altered by embankments, land rights, roads, class, caste, gender, migration, information and administrative decisions. River science alone cannot explain the social meaning of floods; social interpretation alone cannot alter the physical causes of water flow. A combined natural- and social- scientific method is therefore required. In studies of migration, census records, rail and bus mobility, employment data, remittances, family structure, marriage, language and interviews about aspiration are useful together. Income differentials alone do not explain the whole cause of migration, and the pain of leaving home alone does not explain the economic structure. Studies of land and agriculture require a combined analysis of revenue records, maps, satellite imagery, rainfall, soil, crop prices, credit, tenancy, inheritance and local terminology. Documentary categories and actual use may diverge; archival source criticism is therefore indispensable. Research on the Panji tradition, genealogy, marriage rules and social prestige should keep manuscripts, family memory, legal documents, oral history and current practice distinct. Panji is a source, not an automatically representative sample of the whole society. Limits of class, caste, gender and archival silence must be recorded. In the study of the Maithili language, census categories, school language, self-identification, script, home language, media use and the behaviour of migrant communities may vary independently. “Mother tongue” is not a simple natural property; the relation among administrative category, family identity and actual practice must be examined. Applying the discipline of Pūrvapakṣa–Uttarapakṣa to social research, Mithila’s Parallel method would say: first construct the strongest rival explanation to your own hypothesis; then test both against sources, data, agents’ meanings and institutional history. Regional pride, political commitment or cultural sympathy cannot take the place of evidence. Six Levels for Testing a Social Claim Clarify the definition → examine the provenance of the source → seek alternative causes → understand the agent’s meaning → state institutional and historical limits → conduct independent retesting One number does not explain society; one story does not explain society either — combine multiple levels of evidence Contemporary Applications Impact evaluation is highly useful in public policy, but before declaring a programme successful ask: successful for whom, in which region, for how long, compared with which group, and with what unintended effects? Average results may conceal inequalities among groups. In education research, examination scores are important data, but they are not the whole meaning of learning. Language comprehension, attendance, teacher availability, household resources, nutrition, fear, school culture and curriculum must be considered together. A rise in scores does not by itself prove a rise in genuine learning. In public health, biological and social causes are intertwined. Pathogens or biological risks belong to natural science; housing, sanitation, income, working conditions, access to healthcare, belief and stigma belong to social science. Policy succeeds when the two levels complement rather than replace one another. In the social study of climate change, conclusions from physical climate models do not automatically predict social behaviour. Perception of risk, insurance, agricultural choices, migration, land rights and state capacity shape social response. Climate causes are real; social outcomes are institutionally mediated. Artificial-intelligence-based social prediction requires particular caution. Models of credit, crime risk, employment or examination success may learn regularities from historical data; yet historical bias, feedback from classification and policy response can alter the future. High predictive accuracy is not the same as just causal reasoning. In studies of digital platforms, the volume of data may be enormous, but population representativeness is not guaranteed. Who is present on the platform, which behaviour is recorded, what the algorithm makes visible and which data researchers can access are all selections. Big data do not equal the whole society. In voting and opinion surveys, question wording, sampling, non-response, social desirability, timing and events matter. A survey is a useful measure; it is not democratic will itself. Uncertainty in the result should be stated openly. In criminology, recorded crime and actual crime may differ. Police presence, capacity to report, legal definitions, social stigma and administrative incentives can alter the data. Before claiming that crime has risen, the measurement process must therefore be examined. गजेन्द्र ठाकु र In economics, models are useful simplifications. Assumptions such as the representative consumer, perfect information or equilibrium can facilitate analysis, but they may conceal actual social institutions and inequalities. The purpose and limits of a model should remain explicit. Joint work between history and social science is especially useful for understanding policy change. The effect of a law may be altered not merely by immediate data but by long-term institutional adaptation, social resistance, judicial interpretation and administrative capacity. Time itself is a causal dimension. Chapter Conclusion The right account of the relationship between natural science and social science is neither one identical method nor two wholly separate worlds of knowledge. Their shared scientific discipline lies in evidence, causation, clarity, retesting, counter-examples and correction of error. The difference lies in the nature of the subject matter: the social world is meaningful, rule-structured, historical and reflexive. Social science must therefore connect causal explanation with understanding of meaning, institutions and history. Parallel Philosophy proposes not an artificial compromise between the rigorous evidential discipline of naturalism and the meaning-sensitive method of interpretive traditions, but a division of labour. When the question is causal, examine causes; when it concerns meaning, examine meaning; when it is institutional, examine rules and resources; when it is historical, examine pathways; and keep every conclusion open to contrary evidence. Chapter 83 maxim: “The natural world resists through causation; the social world gives meaning alongside causation. Science requires rigour in both, but method must change with the subject.”