Philosophy of the social sciences

Some epistemologies & methodologies

 * Group the following into objectivist (there is an absolute truth out there), constructivist (no absolute truth, interplay between object and subject), and subjectivist (meaning is imposed on object according to phenomena external to object)
 * Ontology → epistemology → theory → methodology → methods (see also "Scaffold of learning" by Sumner & Tribe, 2008)
 * Rationalism
 * Positivism
 * There is a reality, which is observable; the aim of knowledge is the acquisition of the single truth in the form of a universal, general law; the observer is objective, independent
 * Empiricism (Falsificationism; logical/neo-positivism)
 * Critical realism
 * Relativism
 * Logic of inquiry: abduction
 * Scientific knowledge should be derived by investigating the meaning of actors
 * Engage in the interpretation of those in a particular setting
 * Discover and describe the insider view and not imposing an outsider view of reality (subjectivities)
 * Multiple realities; the aim of research is a more informed construction or understanding; researcher is subjective and not independent
 * Realism
 * There is an independent reality and it can be described; the aim of knwoledge acquisition is to describe it, however, it is not possible to establish a truth about reality; researcher and their thoughts are part of reality and hence she is a dependent observer
 * Historical materialism
 * Philosophical realism
 * Scepticism
 * Solipsism
 * Instrumentalism
 * Functionalism
 * Structuralism
 * Interpretivism
 * Phenomenology
 * Post-structuralism
 * See here for an Outline of epistemologies.

General thoughts

 * Every social science has at the least one value underlying it as a discipline: epistemic values underpin every discipline. We always make decisions on what is important to research, to know something about and this is grounded in a prioritizing of knowledge based on values
 * Learn the trade of some empirical discipline (interpreted broadly, to include history for example). But also make sure that you do not become a full-time scientist. The difference between philosophy and science lies as much in the questions asked as in the methods used. Scientists usually do not ask the deep questions raised by philosophers, and also lack some of the distinctive tools that philosophical training provides.
 * In general, philosophers of social science are concerned with the differences and similarities between the social and the natural sciences, causal relationships between social phenomena, the possible existence of social laws, and the ontological significance of structure and agency
 * We want to know why things happen and not just what will happen (i.e. an instrumentalist view: don't mind explanation as long as it predicts); we want evidence that the model in question cites real underlying causes of the phenomena we observe. Claims as being qualified ceteris paribus renders them either non-falsifiable or else superfluous. Either we can specify what the “other things being equal” are or we cannot do so. If we cannot, then social claims qualified ceteris paribus seem unfalsifiable, for every failed prediction has an “out” – other things weren’t equal.
 * Ways to tell whether a model is explanatory:
 * It provides insight – an informal rationale common among social scientists as a defense of particular models.
 * It unifies, i.e. shows how different phenomena might be captured by the same model
 * It serves as an instrument – we can do things with it
 * It is isomorphic to the phenomena of interest
 * Issues that are at the core of the philosophy of social science: the nature of institutions, the emergence of norms, and the relation between individual behavior and aggregate social phenomena
 * Mechanisms that work at the social level include for example:
 * reflexivity: social entities are constituted by beliefs about beliefs (such as "I think that you think that I think...")
 * performativity/looping effects: if social entities are somehow made of beliefs, they (unlike natural entities) must be constantly re-created (or ‘performed’) by the individuals who belong to a given social group. Language is not just a tool to describe the world, but also to make it up. (e.g. by uttering "I hereby declare you husband and wife", I actually create a fact). Indeed a major task for a new social concept is to classify pre-existing phenomena in a new way that is functional to novel social, political, and scientific needs. Unlike natural-kind terms, however, social kinds are interactive: the very people who are classified react to the classification in ways that are only partly predictable. People react by changing the way they see themselves, their identities, goals, and role in society.
 * collective intentionality: it's impossible to reduce, for example, we-beliefs ("we as the XYZ believe") to an aggregation of I-beliefs without losing some of the intrinsically ‘social’ character of the institutional reality they contribute to bring about

An approach to the PoSS

 * The idealized positivist research is to have a theory, generate hypotheses, define concepts, operationalize them and then collect data
 * The other way ("grounded theory") is to start bottom-up data collection
 * Distinguish between the methodology of the philosophy of the social science and an overview of the methodologies of the specific disciplines
 * Distinguish between pre-scientifc ontology and social ontology
 * Approaching the social sciences means asking "What is the object of study" & "How is the object studied"?

Ontological issues

 * Closed system ontology: governed by law-like regularities which occurs across time and space; predictability and universality; intrinsic condition of closure (one cause leads always to one effect) and extrinsic (vice versa)
 * Open system ontology: no or only partial event regularity which excludes perfect predictability (i.e. demi-regularities)
 * Internal/necessary relations: nature of objects and occurences in the world depends on their relation to another object/occurence; the existence of one necessarily presupposes the existence of another object or occurence
 * Ontological atomism: denies existence of internal relations; all relations between objects/occurences are external or contingent or accidental
 * Ontology is the theory of what there is, or of the fundamental constituents of being. In our culture the ultimate authority concerning the nature and structure of reality is science. So, is there room for a non-scientific, pre-scientific, or extra-scientific theory of what there is?
 * ontology is always internal. It always involves a commitment to some (scientific, commonsensical, etc.) theory, and makes use of whatever mix of empirical, pragmatic, and logical considerations is considered appropriate to solve a given problem.
 * Philosophers of social science often address philosophical questions raised by pre-paradigmatic social science. According to Thomas Kuhn, famously, there is an inverse relation between the level of philosophical controversy and the potential for scientific progress within a given discipline
 * If you cannot take the results of a scientific discipline for granted (because there are no widely accepted results) then you end up discussing predominantly pre-scientific questions.
 * Despite its name, the debate about methodological individualism as always been more about metaphysics than methodology. A solution to the central question (Are explanations that refer to ‘macro’ entities legitimate?) has been usually sought in some ontological thesis about the independent existence of macro entities, the reduction of macro to micro properties, and so forth.
 * Is our philosophy of social sciences biased towards behavioral fields (economics etc) as opposed to for example anthropology

Conventions & norms

 * For evolutionary game theory, social phenomena (e.g. social conventions) can be depicted as coordination games in which there can be multiple equilibria. An equilibrium is achieved if both actors have the optimal outcome achieved through cooperation (e.g. on which side of the road do we drive)? Left, Left or Right, Right? Any of the two will do, but in the absence of other information or the possibility of communication, it is difficult for each player to decide on which side of the road one should drive. → thus, common knowledge is a crucial condition for such equilibria to exist.
 * Difference between social norm (which has an intrinsic ‘ought’, and is usually backed up by a system of informal/formal sanctions) and convention: Conventions are the outcome of repeated coordination games. Without precedence there would be no focal points (i.e. an idea/intuition of what I think others are going to believe/do, like driving on the left side), and no convergence on a common solution. Thus, several problems that are intractable in a one-shot game can be solved in repeated play when I know that since last time you drove on the left side I expect you (and to expect me )to drive on the left side, and so on. In short, the role of history and repetition as the central mechanisms for the creation and stabilization of social institutions
 * Norms are triggered or activated by contextual cues, little details that signal that such-and-such is a situation in which a certain norm applies. These cues tend to work at an unconscious level, which explains why we are often not very good at explaining why we act as we do.

Investigating social phenomena

 * Doing social research includes:
 * Claim (C): whose assertibility is being defended
 * Qualifier (Q): which indicates the strength of the empirical support; context-dependence of the phenomenon (e.g. not at all times, not for everyone)
 * Data (D): that present the grounds supporting the conclusions
 * Warrant (W): which demonstrates the guarantees that authorize our movement from the data to the conclusions; methodological guarantees regarding the legitimacy of drawing conclusions about a social phenomenon starting from the methods that were used
 * Backing (B): that supports, empirically for the most part, the warrant; empirical documentation which makes it possible to assess the match (e.g. official record of the phenomenon)
 * Rebuttal (R): which indicates the counter–arguments that challenge the conclusions


 * Social facts/objects are structured object, in the sense that they do not simply consist of whatever pattern of correlation is included in an empirical law. They refers to the structures, processes and tendencies which, by existing to some extent independently of the law-like pattern, make the occurrence of the latter causally intelligible.
 * At an epistemological level, we can emphasize that it is a complex object, which can be made intelligible by means of different empirical reconstructions, without having to assume that a phenomenon such as ‘slow productivity growth’ can be found in the outside world independently of its theoretical identification. The phenomenon as described becomes empirically relevant as an object of inquiry, only after some research-dependent formulations of it allow for some questions to be addressed and for some consequent answers to be explored.
 * Ontological questioning requires a descriptive inquiry. This does not mean that what there is simply amounts to what we can know about it (positivistic move), or say about it (hermeneutic move).
 * Can we characterize the objects independently of the methodology?
 * Taking the perspective of another discipline can teach you what you have been studying in your discipline so far
 * Asking "What is the social in the phenomenon?" What is the autonomously social in the social? What makes the social not just an aggregation of individual behaviors?
 * Distinction between disciplines doesn't mean boundaries
 * You have types and instantiations of the types (i.e. tokens)
 * It is not the high number of occurences (e.g. suicides) that makes it social, but the fact that it is social that generates the high numbers
 * Rather than assuming a certain picture of science and taking that as the standard by means of which the object of investigation is to be defined and judged (an epistemological direction), one might set out instead to investigate what such an object must be like in order for a science of economics to be possible, and to spell out the details of construction of such a science (an ontological direction). All theories of knowledge, science-oriented philosophies, social practices, etc. presuppose an ontology, that is some conception of the nature of reality. So, we must proceed by investigating what the world must be like for these theories, practices, etc. to be possible.

Durkheim & LeBon
Compare the 'social' of Durkheim vs. the 'social' of LeBon

LeBon on crowds:
 * A psychological crowd: A numerically strong agglomeration of individuals does not suffice to form a crowd. There are special characteristics of psychological crowds such as:
 * the turning in a fixed direction of the ideas and sentiments
 * their conscious personality vanishes
 * a collective mind/mental unity is formed and the crowd is always dominated by considerations of which it is unconscious
 * Psychological crowds vary in their organisation which is influenced by: race, composition, nature & intensity of the cause
 * It is only the uniformity of the environment that creates the apparent uniformity of characters → back to normal life all the crowd members might be entirely different characters
 * The psychological crowd is a provisional being formed of heterogeneous elements, which for a moment are combined, exactly as the cells which constitute a living body form by their reunion a new being which displays characteristics very different from those possessed by each of the cells singly.
 * The greater part of our daily actions are the result of hidden motives which escape our observation. It is more especially with respect to those unconscious elements which constitute the genius of a race that all the individuals belonging to it resemble each other, while it is principally in respect to the conscious elements of their character—the fruit of education, and yet more of exceptional hereditary conditions— that they differ from each other. Men the most unlike in the matter of their intelligence possess instincts, passions, and feelings that are very similar → It is precisely these general qualities of character, governed by forces of which we are unconscious, and possessed by the majority of the normal individuals of a race in much the same degree—it is precisely these qualities, I say, that in crowds become common property
 * The heterogeneous is swamped by the homogeneous, and the unconscious qualities obtain the upper hand. This very fact that crowds possess, in common, ordinary qualities explains why they can never accomplish acts demanding a high degree of intelligence.
 * Different causes determine the appearance of these characteristics peculiar to crowds, and not possessed by isolated individuals:
 * 1) Invincible power: the individual forming part of a crowd acquires, solely from numerical considerations (and through anonymity and irresponsibility), a sentiment of invincible power which allows him to yield to instincts
 * 2) Contagion: In a crowd every sentiment and act is contagious, and contagious to such a degree that an individual readily sacrifices his personal interest to the collective interest.
 * 3) Suggestibility: having entirely lost his conscious personality, he obeys all the suggestions of the operator who has deprived him of it, and commits acts in utter contradiction with his character and habits. He is no longer conscious of his acts. All this is even stronger in crowds than in hypnotized individuals since the suggestibility affects everyone in the crowd
 * The crowd is always intellectually inferior to the isolated individual, but, from the point of view of feelings and of the acts these feelings provoke, the crowd may, according to circumstances, be better or worse than the individual. All depends on the nature of the suggestion to which the crowd is exposed.

Methods & models in social research

 * The distinctive feature of modelling is that the system under investigation is studied indirectly: we build a model, we analyse the model, and then we relate the model to reality.
 * Distinguish the model form its description: The descriptions can consist of equations, computer code, pseudo-code, graphs, pictures and most importantly, words - the same model can be described in very different ways
 * The contemporary scientific method is biased toward investigating causal relationships. Predictions based on models and theories means to arrive at a formal representation of what we think (some of) the underlying causal processes are within a given social context, and then use deductive and mathematical tools to predict future states of the system
 * We can ask two sorts of questions about a model:
 * 1) Truth value of the model: We can ask whether the model is a good approximation of the underlying social reality—that is, the approximate truth of the theory or model. Likewise, we can ask whether the theory or model gives rise to true predictions about the future behavior of the underlying economic reality (subject to the time frame of the analysis)
 * 2) The warrant of the model: the strength of the evidence and theoretical grounds available to us on the basis of which we assign a degree of credibility to the model: does available evidence give us reason to believe that the model is approximately true, and does available evidence give us reason to expect that the model’s predictions are likely to be true? These questions are centrally epistemic
 * The question of the approximate truth of a model is separate from that of the approximate truth of its predictions. This might be the case because the ceteris paribus conditions are not satisfied, or because low precision of estimates for exogenous variables and parameters leads to indeterminate predictive consequences.
 * Theories give rise to models and models produce predictions. But there is a specification problem at two levels: the model can specify the theory in various ways, leading to different predictions. And the parameters of the model themselves can be estimated in various ways—again producing different predictions.
 * Oftentimes, the distinction between observational and experimental is meant to express a difference in quality; "if you can't do experiments, you do the observational thing"
 * Data consist of observations or measurements of characteristics of populations or individuals under study. Data are oftentimes indirectly generated by measuring other characteristics. For instance, school motivation can be measured by recording class attendance. Or 'age' can be an observed (and directly measured) variable, or a proxy, i.e. a variable that 'stands for something else' and that is not directly measured (e.g. in which life period the person is, school-age/work-age etc.). Once generated, data are then organised and grouped according to variables. Using a type of variable rather than another depends on reasons that can be methodological, empirical, or other. Once data are organised, we have to organise the variables. This is the task of quantitative models
 * What is a model? Differentiate between the mental conceptions natural scientists and social scientists have in mind when they hear 'model'
 * Scientific rigor is not what model you use (not about the object of study), but how you use it (about methodology)
 * Models may be used to explain, to establish causal relations, or to simulate a phenomenon.
 * Models lie at the `interface' between the epistemic agent (in this context, the scientist) and the system under investigation (be it physical, biological, or social). Thus models allow us to study, understand, and interpret the surrounding reality
 * Epistemic agents act in communities (e.g. the scientific community), and the epistemic activities involved in science, in which modelling is certainly central, are then distributed - distributed across all the individuals that belong to the scientific community
 * Instead of an incessant quest for objectivity in the social sciences, the role of the epistemic agent (the scientist) in model-based reasoning should not be reduced as much as possible, but should instead be studied and understood as much as possible.
 * Different types of models: qualitative/quantitative, experimental/quasi-experimental, theoretical/simulation
 * Two conceptions of models: models as representation (with the two variants: set-theoretic structures and family of probability distributions) and models as objects (with the two variants: models as fictional entities and as epistemic objects)
 * Concerning the relationship between models and reality: models as mediators, as isolations, and as maps
 * Models are like maps. Maps, to be sure, are not true or false, but useful or useless for a specific purpose


 * Models as representations:
 * A model as a representation of a phenomenon, or of a certain portion of reality in the sense that it captures the main features of the phenomenon, and expresses them in a formal manner. Define a model as a set-theoretic structure. In set theoretic models, the model is the `formal part' of a theory. More generally, a model is an abstract structure, such as a mathematical structure, or a set of statements formalised in first-order logic or other logic. Representations are structures (set-theoretic or probabilistic)
 * Since a model is not coincident with what it is modeling, it has to abstract away from certain features of the described phenomenon 'P'. So one will make simplifying assumptions, based on what one takes the most important features of P to be, and include certain features while omitting others: this will produce a schematized picture of the actual workings and actual nature of P. → A theory may abstract from all manner of phenomena, have nothing to say about them, and hence be "descriptively inaccurate" (or incomplete), but that does not make it false. The question of falsity only arises when a theory yields definite predictions about some phenomenon which turn out to be untrue (although this statement might be questioned, too). Descriptive inaccuracy is an inherent quality of any abstract model; but falsity is not.
 * That statements describing a set of already known facts may be deduced from a model is evidence, not of its truth, but of the logical skills of the person who constructed it.
 * Distinguish between between the ideal-as-descriptive-model (an ideal—in the sense of accurate—model of how P actually works) and ideal-as-idealized-model (an ideal—in the sense of an exemplar—model of how P should work). Ideal theory (i.e. models) either tacitly represents the actual as a simple deviation from the ideal, not worth theorizing in its own right, or claims that starting from the ideal is at least the best way of realizing it
 * The model has to have a axiomatic-deductive structure; there is a phenomenon that you can present mathematically
 * In the social sciences, a model (is often a statistical model that) represents a phenomenon by (family of) a probability distribution. A probability distribution is a function that assigns a probability value to each of the possible values of a variable. To say that a model is a family of such distributions means `putting together' the probability distributions for each of the variables in the database and then studying their behaviour
 * The use of probability theory and statistics to study phenomena (social or natural) presupposes a stochastic representation of reality, rather than a deterministic one. This is mirrored in the inclusion of `error terms' that can stand for measurement errors, latent variables, or even for the fact that the phenomena are genuinely indeterministic. A phenomenon might very well be deterministic while our representation of it is stochastic. The representation of a phenomenon is partial|the presence of terms of errors or latent variables means that we cannot take into account all possible aspects of the phenomenon.
 * Model as representation doesn't capture the interactive relationship the models have with the epistemic agent and the world


 * Models as objects:
 * Some models as physical objects (a globe) vs. as abstract object/fictional entities (e.g. Bohr's atom) vs. as epistemic objects (tools for understanding; as mediators because they mediate the relation between the epistemic agent and the world in constant interaction/provide understanding). Models mediate instrumentally between theories and reality, i.e. models mediate the access to reality.
 * Models are models of data, not of theories.
 * A model also allows us to learn about the two sides that it connects (in virtue of their representative function). An interesting aspect of this view is that we do not learn from the model just by `looking' at it, but by building it and manipulating it, and that is why they are tools
 * Models as fictional entities that we use instrumentally. Models are objects because they are concrete, tangible `products' that we can manipulate in different ways.
 * The model is somehow autonomous, somehow dependent. Models have partial autonomy compared to the theories on the one hand, and the reality on the other hand.
 * The function of a model determines which aspects are highlighted (like a map to navigate)


 * Validity has conventionally only looked at whether my theory corresponds to reality (is my theory right). Do look at the process of getting to know (the model I use, the scientific community review etc.)
 * Simulations are based on the assumption that a reduction of scale leaves unchanged the essential characteristics of the system.
 * The aim is to be able to reproduce the state of a system as it evolves from certain initial conditions that are set in the computer programme. Simulations have an interesting, hybrid status between experiment and theory.
 * The difficulty in finding an easy and clear correspondence between our scientific statements and some facts, state of affairs, or truthmakers stems from the fact that social reality is notoriously elusive and, to some extent constructed

Economic models

 * Usually optimization (choice) or fixed point problems (equilibrium), but there is a lot of stochastic processes too
 * Construction: tinkering, trial-error:
 * Use: which narratives are considered meaningful; models are not tested but used
 * Interpretation of econ models
 * Probabilistic modeling allowing to act as if the future were known
 * Models don't try to show what is most likely to happen but paint different possible futures and their effects
 * Models not as ideas or theories but as infrastructure
 * "In order to study the consequences of contemplated actions, a model (a causal description of the phenomena which are affected by the contemplated actions or affect the actions) is needed. Now this model should […] contain and describe all those factors or phenomena which are important. But what is more important than the causal model itself which determines what can be traced as a consequence? Therefore the model should be part of the model, because it influences what can be figured out as a consequence. In other words, a model should contain itself, and that is impossible.
 * Models intervene in ongoing events
 * Objectives of model are built-in - precisely because the model is built for a purpose
 * Model as a disciplining device for utopian policy makers, forcing the latter to formulate their objectives in terms of the model

Causality

 * A world of cause and effect is a world subject to change in historical time
 * To infer causal relations we need two fundamental elements: temporal asymmetry and comparisons between exposed and unexposed groups or cases and controls.
 * A causal inference may be based on a relation when three criteria are satisfied:
 * the "cause" precedes the "effect" in time (temporal precedence),
 * the "cause" and the "effect" are related (covariation), and
 * there are no plausible alternative explanations for the observed covariation (nonspuriousness)
 * Five different problems about causality:
 * Inference: the first problem of causality is the problem of inference: Is there a causal relation between X and Y? Does X cause Y? What are the causes of Y? What are the effects of X? How much of X causes how much of Y?
 * Prediction:
 * Explanation: We often want to know not just what happened or will happen, but how it happened, and why; how do we (causally) explain phenomena?
 * Control: how do we control variables that may confound a relation between two other variables? How do we control the world or an experimental setting more generally?
 * Reasoning: What reasoning underlies the construction and evaluation of scientific models? What conceptualization of causation underpins causal methods? How do we reason about all aspects of causality? Thinking about causal reasoning more broadly is when we worry about relations between the other four problems
 * We learn about causality by comparing different groups, different instances, and different situations (e.g. through observational studies)
 * The ‘common cause principle’, which says that if two events are correlated, either one causes the other, or there is a third event causing both of them. Similarly, formulation of the screening-off relation: in the presence of a common cause S (such as Smoking), of two other variables Y (such as Yellow fingers) and L (such as Lung cancer), S screens off Y from L. That is, if you hold S fixed, the association between Y and L disappears. This would appear correct: for smokers, Y and L are associated (although it is a spurious correlation), while for non-smokers they are not. So once S is held fixed, the association disappears, showing that it is due to S, rather than being a symptom of a genuine causal relation between Y and L, in either direction
 * Defining causality/causation or establishing truth conditions for causal claims are metaphysical/conceptual questions. Establishing what evidence supports causal claims lies between epistemology and methodology. Manipulations and tests such as invariance are tools to establish causal relations and fall in the remit of methodology
 * Negative causes: i.e. causes that prevent effects from happening. In probabilistic terms, either causes raise the probability of the effect, or they lower it (in which case they are customarily called preventatives)
 * The central idea of causal ascription is the idea of causal powers and causal mechanisms
 * Marx generally spoke of the economy as a single organism, in which each new condition is the result of all the conflicting forces at work within it in the preceding period; so we should speak not of causes, but of functional relationships.

Variance

 * To establish causal relations we track what varies with what, either by observing changes, or by provoking them in experiments
 * Without variation invariance is devoid of meaning.
 * Method of Agreement: comparing different instances in which the phenomenon occurs
 * Method of Difference: comparing instances in which the phenomenon does occur with similar instances in which it does not
 * Method of Residues: subducting from any given phenomenon all the portions which can be assigned to known causes, the remainder will be the effect of the antecedents which had been overlooked or of which the effect was as yet an unknown quantity
 * Method of Concomitant Variation: In the presence of permanent causes or indestructible natural agents that are impossible either to exclude or to isolate, we can neither hinder them from being present nor contrive that they shall be present alone
 * In comparing instances in the Method of Agreement and of Difference, we look precisely at things that vary, and at things that stay the same
 * Causal reasoning in Mill’s methods recommends that we do more than simply observe the effect (failed rising) following the cause (to be determined). The four methods tell us to look for how things vary or stay the same, when other factors change or stay the same
 * Between observation and experiment there is no logical distinction but only a practical one
 * There is only one way of proving that a phenomenon is a cause of another, that is comparing cases where they are simultaneously present or absent, and looking whether the variations they exhibit in these different combinations of circumstances prove that one depends on the other
 * The rationale that underlies our epistemology of causality is: variance. Quantitative causal analysis establishes causal relations by measuring variations, not by establishing regular sequences of events
 * We design, and should design, causal methods in the social sciences to uncover variations. Simply put, we do this by examining, in a data set, correlations between our target variable (the effect, such as child mortality) and possible causes (such as maternal education). We will typically consider those (joint) variations that are regular and invariant. These are the variations that are strong enough and that stick around when some things change. We find these by searching for correlations which remain across chosen partitions of the population being analysed
 * Variation does not tell you that all there is to the nature of causation is difference-making. It says that the way you find out about causal relations, the rationale by which you design methods, is tracking difference-making, and then substantiating the correlations you find with other sorts of arguments, such as mechanisms, exogeneity and invariance, in ways that depend on the scientific context
 * Exogeneity tests are used to check if cause and effect are properly `separated', i.e. if the (probabilistic) structure of the model is correct from a causal point of view. Exogeneity is typically explained by saying that exogenous variables are caused outside the model, while the endogenous ones are caused within the model (by the exogenous variables)
 * Invariance tests, instead, are used to check that the causal structure is sufficiently stable across different partitions of the population of reference, or under interventions or manipulations
 * In probabilistic approaches, we compare the unconditional (or marginal) probabilities of the effect and its probability conditionally on the cause. This means that the cause makes the probability of the effect vary
 * Regularity of the variation means that the joint variations between the variables of interest show up often enough (i.e. strength of association). Regularity can be in time as well as in other factors
 * In performing manipulations, we make the putative cause-variable vary and observe what variations occur in the putative effect-variables. Manipulations as a valid tool for testing causal relations.
 * Invariance helps us establish causal relations that are valid for the population as a whole; that is we are interested in establishing generic causal claims
 * We need to compare different instances of the putative cause with different instances of the putative effect (if everyone/nobody smoked it would be difficult to establish causal relation)
 * Establishing a causal relation means establishing, in the first place, a difference-making relation between two variables (e.g. smoking makes a difference to lung cancer). Often, however, to determine whether these difference-making relations are causal, we also need evidence of production, namely of how smoking causes cancer (i.e. mechanisms)
 * While quantitative models generate evidence of difference-making, it is less clear how they also generate evidence of production.
 * Remember: Internal validity refers to the possibility that the relationship between two variables, within a given model, is causal or, conversely, to the possibility that, given the lack of correlation between two variables, we conclude that there is no causal relation between them. External validity concerns the possibility of generalising a causal relationship, established within a specific model, to different populations or settings.
 * An epistemological notion of causality represents, in other words, the scheme of causal reasoning, i.e. it is a rationale. A rationale is the principle or notion underlying some opinion, action, hypothesis, phenomenon, reasoning, model, or the like
 * A taxonomy of variations can be sketched according to five different criteria, or taxa:
 * Variations across time.
 * Variations across individuals.
 * Variations across characteristics.
 * Counterfactual and control group variations.
 * Observational vs. interventional variations.

The epistemology of causal reasoning

 * A rationale of causality as measure of variation.
 * The content of the epistemology of causal methods: when we reason about causal relations, we reason about variations/differences (in/between groups etc.). We search for causation in variation
 * What is the relations between epistemology and methodology?
 * Causal assessment (what causes what) has two objectives: explanation (i.e. knowledge) and action/prediction (knowing where to intervene)
 * Causation can be singular in time & place vs. empirical generalizations
 * What do you ask (for the objects of causation, the causal relation, the linguistic meaning of causality) and what notion guides us through causal reasoning?
 * In the lab, you do have manipulation of a variable but this manipulation is not present in the statistics as well
 * The manipulationist view: y = ßx + z → change in x is causally related to change in y but the manipulationist view prioritizes the experimentally induced variation (making people smoke) instead of naturally occurring variation (differences in smoking behavior)
 * Counterfactual reasoning as a helpful tool to generate hypotheses and select variables (e.g. the covariates)
 * An epistemological notion of causality represents the scheme of causal reasoning, i.e. it is a rationale
 * A rationale is the principle or notion underlying some opinion, action, hypothesis, phenomenon, reasoning, model, or the like.
 * If X is a cause of Y, we have in mind that a change in X produces a change in Y and not merely that a change in X is followed by or associated with a change in Y.
 * The difficulty is that we cannot test whether the change in X actually produces the change in Y, unless we postulate a causal mechanism responsible for that beforehand. Producing refers to an ontological process, whereas here we are interested in the epistemological principle—i.e. in the rationale—that allows causal inferences in causal modeling  although a cause produces an effect, a change or ‘’variation’’ in the causal variable produces a change or variation in the effect variable
 * The study of change is thus the study of factors which produce change.
 * When we say that X is a cause of Y, we mean that a variation in Y has been detected, and that we have good reasons to say that X, particularly, a change in X, is responsible for, i.e. produced in Blalock’s words, the change in Y
 * Probabilistic theories of causality focus on the difference between conditional probability P(E|C) (i.e. the probability that E occurs, given that C occurred) and marginal probability P(E) (i.e. the unconditional probability that E occurs)  probabilistic theories of causality, by using statistical relevance as a necessary ingredient, establish causal relations by making variational claims, that is the variation in the conditional probability of the effect is due to a variation in the marginal probability of the cause. Whilst variation is a necessary condition to infer causality regularity isn’t. Probabilistic theories of causality might require regularity to allow generalisations, but not to infer causality tout court.
 * To claim that we expect a cause to affect the marginal probability of the effect is not equivalent to saying that from independence we can straightforwardly conclude to the absence of a causal path
 * The basic idea of Structural Equation Modelling is that in a system of equations we can test whether variables are interrelated through a set of linear relationships, by examining the variances and covariances of variables  write the covariance of any pair of observed variables in terms of path coefficients and covariances. The path coefficient quantifies the (direct) causal effect of X on Y; given the numerical value of the path coefficient β, the equation claims that a unit increase in X would result in β units increase of Y. In other words, β quantifies the variation on Y accompanied by the variation on X
 * Variations in X are accompanied by variations in Y, not that values of Y regularly follow values of X
 * r2 measures the goodness of fit, not the validity of the any underlying causal model. This means that the guarantee that we hit upon a correct causal structure has to be assessed on other grounds
 * Covariance structure models attempt to explain the relationships among a set of observed variables (in the measurement model) in terms of a generally smaller number of unobserved variables (in the structural model).
 * In particular, we are instead interested in a more specific probability of an effect B, say P(B|A ∩ C ∩ D ∩ E), where C, D and E are all relevant factors, for instance gender, religious background, marital status of parents, etc. The crucial point is that if conditioning on a further factor, say F, does not change the previous conditional probability, this means that F is not a relevant factor and hence should not be considered in the explanation. So all factors entering the S-R model are statistically relevant, i.e. responsible for variations, in the probability of the fact to be explained
 * Average causal relations tell only part of the causal story → thus see double-frontier approach: the causal impact of a certain characteristic is evaluated, roughly, depending on the concentration of the set of observations between the two frontiers. Simply put, the goal is not to estimate an average causal effect; instead, we are interested in the distance between the maximum frontier and the minimum frontier—i.e. how the health indicator varies between extreme values, and on how this distance varies for different levels of the explanatory variables. The shorter the distance between extreme values, the stronger the (causal) influence of the characteristic under analysis on the health indicator
 * The rationale of variation conceptually precedes the condition of stability of the putative average causal effect
 * The study of cause involves the detection of change in a dependent variable produced by change in a independent variable. The key point is to establish what variations are causal, not what regularities are causal
 * A factor is a cause of a certain disease when alterations in the frequency or intensity of this factor, without concomitant alterations in any other factor, are followed by changes in the frequency of occurrence of the disease, after the passage of a certain time period.
 * Hume says that, in spite of the impossibility of providing rational foundations for the existence of objects, space, or causal relations, to believe in their existence is a ‘built in’ habit of human nature. Imagination allows us to order complex ideas according to (i) resemblance, (ii) contiguity in space and time, and (iii) causality. Of the three, causation is the only principle that takes us beyond the evidence of our memory and senses. It establishes a link or connection between past and present experiences with events that we predict or explain, so that all reasoning concerning matters of fact seems to be founded on the relation of cause and effect. The crucial step in Hume’s argument is significantly different from the rationale Russo proposes. Her claim is that we look for variations, not for regularities.
 * Hume’s regularity view immediately rules out singular causation, i.e. causal relations that occur only once and that therefore do not instantiate regularities
 * An explanandum-partition is imposed upon the initial reference class and cells in the explanans-partition must be (objectively) homogeneous, or otherwise it would mean that a relevant explanatory factor has been neglected.

Pearl

 * The basic principles of causality:
 * Causation = encoding of behavior under interventions
 * Interventions = surgeries on mechanisms
 * Mechanisms = stable functional relationship = equations + graphs

Generalizability and external validity

 * With generalizations we are usually referring to empirical generalizations (as opposed to reasoning about things)
 * Usually, the deductive-nomothetic model assumes that if we can explain anything, it is by means of laws
 * Construct validity: are we sure that under different definitions of the construct, we find the same thing?
 * The difference between internal and external validity is disputed since (internal validity) I always extrapolate when I deduce something from the experiment (external validity)
 * Social regularities are phenomenal rather than governing; social explanations depend on the discovery of causal mechanisms underlying various social processes. The causal properties of social institutions, and the micro-mechanisms that underlie them, give rise to phenomenal laws, and these are the chief regularities identified by social scientists—not governing regularities
 * Regularities are not fundamental; we are always well-served by seeking an account of the causal mechanisms that produce them, and we will better understand the scope, reliability, and variance of such regularities when we have a true theory of the underlying causal mechanisms. This also allows us to account for the failures of the generalization we make based on a correlation
 * Debate about the foundations of the social sciences among philosophers has largely divided between empiricist and interpretivist positions. Philosophers of social science proceeding from an empiricist perspective have generally taken the view that generalizations are essential; whereas interpretivists have maintained instead that singular interpretation of agent’s meanings is essential. Some social scientists write as though the scientific credentials of their disciplines rise or fall on the strength of the law-like generalizations and regularities that they are able to identify. The task of social science research is to discover the laws that govern social processes. And if a given level of analysis and description fails to produce such laws, then we need to probe more deeply until we discover the underlying order. At the other extreme, some social scientists write as though generalizations have nothing at all to do with social knowledge. All knowledge is “local knowledge”: historical, culturally specific, unique, particular, singular. On this approach, there are no interesting regularities among social phenomena, and causal explanation is an inappropriate model of explanation for the social sciences
 * Regularities derive from features of individual agency in the context of specific social arrangements
 * We must therefore pay more attention to the specifics of the social and individual-level mechanisms that produce the regularities as well as the exceptions.
 * What is most worthwhile within the positivist program is epistemological, not metaphysical. The insistence on appropriate empirical controls on knowledge, the broad distinction between observation and theory, the emphasis on coherence and deductive closure—these are all abidingly important features of scientific knowledge
 * Causal realism: discovery of mechanisms and processes that derive from agents and institutions, and that in turn produce regularities. There are real social relations among social factors (structures, institutions, groups, norms, and salient social characteristics like race or gender). We can give a rigorous interpretation to claims like ‘racial discrimination causes health disparities in the United States’
 * Causal relations depend on the existence of real social‐causal mechanisms linking cause to effect. Moreover, it is defensible to attribute a causal relation to a pair of factors even in the absence of a correlation between them, if we can provide evidence supporting the claim that there are specific mechanisms connecting them. So mechanisms are more fundamental than regularities
 * The successes of the natural sciences have given natural scientists confidence that natural systems operate in accordance with a strict set of laws, that these laws may be given precise mathematical formulation, that they derive from the underlying real properties of constituent physical entities, and, finally, that these facts entail that the future behavior of physical systems is in principle (though perhaps not in practice) predictable → this gave rise to a paradigm of scientific explanation: to explain a phenomenon is to derive the explanandum from a set of general laws and a description of the initial conditions of the system
 * Explanation of a phenomenon or regularity involves identifying the causal processes and causal relations that underlie this phenomenon or regularity
 * Regularity is derivative, not constitutive, of the causal power
 * Successful causal analysis permits us to arrive at statements of social regularities, based on an understanding of the underlying processes that give rise to them
 * A law-like regularity is a universal generalization about empirical phenomena. It is one that conveys necessity. It is a generalization that supports counterfactual judgments. It is a regularity that is grounded in the real causal properties of the entities in question
 * Social regularities emerge rather than govern. The governing regularities are regularities of individual agency: the principles of rational choice theory or the findings of motivational psychology. Social regularities are strictly consequent, not governing. Social regularities are strictly consequent, not governing. They obtain because of the lower-level regularities; they have no independent force
 * There are phenomenal regularities among social phenomena, and these can be discerned through familiar forms of empirical investigation; but they do not serve an important explanatory function within the social sciences
 * Are there social kinds (e.g. riot, revolution, class, religion)? A natural kind is a set of entities which share a common causal structure, and whose behavior can therefore be predicted on the basis of the laws that govern the behavior of such entities. Social concepts function as ideal types or cluster concepts, permitting us to classify a range of diverse phenomena under a single concept. An ideal-type concept is a complex description of a group of social phenomena that emphasizes some features and abstracts from others
 * Since the entities that fall under such concepts do not share a homogeneous causal structure, we cannot infer that instances of the concept will behave in the typical way. Thus market economies have many properties in common. However, particular market economies also have causally significant differences
 * The value in making use of cluster concepts and ideal type concepts in the social sciences is that it permits us to group social entities together in ways that emphasize their common features. This serves to suggest hypotheses about the dynamic properties of such entities. But at the same time, the fact that there are wide differences in important causal features among the entities that fall under a given concept, means that we cannot simply project the future behavior of the entity on the basis of the general features that it shares with other instances
 * Explanations in social science typically involve efforts to uncover the causal properties of social entities and processes.
 * Causal mechanisms are more fundamental than regularities of association between causal variables.
 * It is possible that A has the causal power to bring about B in some fields and not others, with the result that it is practically impossible to observe the corresponding regularity, given data limitations. Causal stories involving complex causal diagrams, complex sets of INUS conditions, probabilistic causation, and incomplete causal fields give rise to situations in which two things may be true: A is a cause of B (in that A is an ineliminable part of the underlying causal diagram or INUS conditions), and there is no observable correlation between A and B
 * Social entities have causal influence, and these causal capacities are to be explained in terms of the structuring of incentives and opportunities for agents. The causal powers or capacities of a social entity inhere in its power to affect individuals’ behavior through incentives, preference-formation, belief-acquisition, or powers and opportunities.
 * Social entities can exert their influence in several possible ways:
 * They can alter the incentives presented to individuals
 * They can alter the preferences of individuals
 * They can alter the beliefs of individuals (constraints on knowledge; ideology )
 * They can alter the powers or opportunities available to individuals
 * Institutions have effects on individual behavior (incentives, constraints, indoctrination, preference formation), which in turn produce aggregate social outcomes
 * Certain institutions have specific causal powers with respect to given social outcomes as a consequence of the common constitution and circumstances of individuals. The Fed has the causal power to dampen inflation, in that it can tighten the money supply; this creates an individual disincentive to purchase; this leads to reduced demand for goods; and this lessens the upward pressure on prices. This causal power is entirely derivative, however, upon facts about typical consumers. The Fed has the power to alter the environment of choice for consumers; the result of this new environment is a pattern of consumption in which demand is shifted downward
 * Social entities possess causal powers only in a weak and derivative sense: they possess characteristics that affect individuals’ behavior in simple, widespread ways. Given features of the common constitution and circumstances of individuals, such alterations at the social level produce regularities of behavior at the individual level that eventuate in new social circumstances
 * For example, the causal power of the state derives entirely from the ways in which the institutions of the state assign incentives, powers, and opportunities to various individuals
 * The causal powers of a thing rather give rise to whatever regularities are observed; and the discovery of regularities is only one out of a number of methods by which we can identify causal relations and powers. Regularities are symptomatic rather than criterial of causal powers and relations.
 * Social causal ascriptions depend on regularities; but these are not generally social regularities, but rather lawlike characteristics of agents (e.g. the central axioms of rational choice theory). The rock-bottom causal stories—the governing regularities for the social sciences—are stories about the characteristics of typical human agents. The causal powers of a particular social institution—a conscription system, a revenue system, a system of democratic legislation—derive from the incentives, powers, and knowledge that these institutions provide for participants. So social explanation does not rest on discovering regularities, or deriving outcomes from statements of regularities. Instead, analysis of the underlying causal mechanisms, and particularly the microfoundations, is central
 * The fact that social causal hypotheses are more than usually burdened by extensive ceteris paribus conditions cautions us to be skeptical about the predictability of social phenomena
 * There are three avenues through which scientific predictions are generated:
 * 1) Predictions based on simple induction/phenomenal regularity (low income country → high infant mortality; thus, we assume the as-yet unexamined low income country to also have high infant mortality)
 * 2) Predictions based on a theory/abstract models of the governing regularities of the (social sub-)system in question (but there are arguably no such regularities among social phenomena)
 * 3) Predictions in novel circumstances on the basis of an analysis of the causal mechanisms/pathways that we can identify (i.e. institutional-logic explanation) in the circumstance, along with a model that permits us to attempt to estimate the aggregate effects of these causal mechanisms (i.e. work out the institutional logic implicit in a given set of social arrangements, e.g. economic models designed to capture certain social mechanisms)
 * A common source of failures of prediction in the social sciences stems from the fact that causal hypotheses and models are generally subject to ceteris paribus conditions. Predictions and counterfactual assertions are advanced conditioned by the assumption that no other exogenous causal factors intervene; that is, the assertive content of the hypothesis is that the social processes under analysis will unfold in the described manner absent intervening causal factors. To the extent that a given causal hypothesis has only identified some of the causal conditions included in the true underlying causal diagram, it is foreseeable that predictions based on the hypothesis will often go wrong. Putting the point in another way, to the extent that a given causal analysis does not provide a complete representation of the full causal field, its predictions may be expected to fail on occasion.
 * Predictions based on such analysis must be understood as representing tendencies rather than probable outcomes
 * Thus, there are no governing social regularities underlying social phenomena. There are governing regularities of sorts, but they are not social (rather, they are regularities representing features of rational agency). And there are social regularities, but they are phenomenal. Therefore the social world is not a system of interrelated variables, concerning which we might aim to discover the state laws. It is rather a complex of processes subject to various causal influences, conveyed by individual agency, onto diverse and rarely predictable outcomes
 * In a nutshell
 * 1) there are social regularities, but they are weak and not the central component of social explanations
 * 2) that the explanatory work of social inquiry commonly takes the form of a search for causal relations and causal powers
 * 3) that causal relations among social phenomena derive their necessity through features of structured individual agency, and nothing else.
 * Extrapolation is a type of argument by analogy
 * Analogies involve judgments that one thing is similar to another, and hence that information about the one tells you something about the other.
 * Refer to mutually informative similarities of this sort as “analogies.”
 * Call a set of interconnected, regularly operating causal relationships that generate one or more regularities between (observable or unobservable) events a "mechanism."
 * Any adequate account of extrapolation must answer a pair of challenges, namely, the problem of difference and the extrapolator’s circle
 * The problem of extrapolation concerns in principle any attempt to generalize from an epistemically privileged system (experimental mechanism, animal model, etc.) to a less privileged target of interest (a nonexperimental situation, human system, etc.).
 * The most common and important instances of extrapolation regard the generalization of causal claims.
 * The problem of difference is that some causally relevant differences between the model and target are inevitable in most interesting examples of extrapolation
 * The extrapolator’s circle rests on the observation that evidence is needed to show that the model is relevantly similar to the target and hence a good basis for extrapolation (given the limits of what we know about the target). In other words, it needs to be explained how we could know that the models and the target are similar in causally relevant respects without already knowing the causal relationship in the target.


 * Comparative process tracing (or, “forward chaining”/”backtracking”): the idea, roughly, that an inference from laboratory conditions to a nonlaboratory (target) system is validated by cross-checking the crucial nodes of the relevant causal mechanisms.
 * In order to discover the fine-grained structure of a mechanism, one may start from one of its end points (one of the initial causes or one of the final effects) and then step-by-step reconstruct the path that connects it with the other elements of the mechanism via the intermediate nodes. This is, in broad terms, the method of process tracing as applied in the service of mechanism discovery
 * A good model should possess the same or very similar mechanisms as the target.
 * One answer to the problem of difference is that the closeness of match required between the model and target depends on the specificity of the causal claim that one wishes to extrapolate. The greater the specificity of the causal claim, the closer the match between model and target must be.
 * An answer to the extrapolator’s circle might be to have background knowledge according to which causally relevant disanalogies are likely to be found at some stages of the mechanism and not others. Alternatively, a difference in a mechanism matters to the outcome only if it has an impact on subsequent steps along the way. Hence, comparisons of model and target mechanisms will be more efficient if they focus on mechanism activities and components that are downstream in the sense of being more direct causes of the outcome
 * The basis of an analogy: whatever features constitute the relevant (i.e. similarities that matter to the cause-effect-relation) similarity of the model and target
 * Mechanisms are not the only possible basis for an analogy. In some cases, the relevant similarity might be largely functional (e.g. the same user-interface in various technological devices)
 * A good argument by analogy should make reasonably clear what the basis is and provide evidence that this basis is present in the model and target.
 * The most direct approach for establishing similarity of mechanisms is a step-by-step comparison of components and interactions among those components.
 * Similar causes as well as similar effects can provide evidence that the basis is present in both model and target. Similar effects of model and target provide evidence of a particular basis only if there is reason to think that those effects are consequences of that basis and not something else. Similarities of the sort relevant to analogical reasoning differ from mere coincidences in that they provide information about one another. Coincidences, by contrast, are similar outcomes resulting from unrelated or independent processes.
 * What evidence, then, would justify drawing a line between B(m) and B(t) in the chain graph and thereby asserting a basis for analogy?
 * Both of these alternatives (mere coincidence and a totally different basis) depend on the possibility that there exist causes of E1 and E2 in addition to the winner’s curse. If we knew or could somehow show that such alternative causes do not exist, then we could be confident that the chain graph depicting our analogy is correct
 * A distinctive marker is a telltale indicator of a particular cause, as fingerprints are indicators of the manual contact of a particular person. A distinctive marker is an effect that could only have resulted from the hypothesized basis of analogy. Thus, if E1 and E2 are distinctive markers of the basis B, then we can be practically certain that B is present in the model and target when both have these effects. But on what basis could we claim that an effect, or set of effects, is a distinctive marker?
 * Prior knowledge of a distinctive marker can be used to provide evidence of an analogy, and prior knowledge of an analogy can be used to discover that an effect is a distinctive marker of a particular cause.
 * One reason to establish an analogy is to enable the model to serve as grounds for further inferences about the target.
 * Effects may be indicators only for a key stage or component of the mechanism rather than an indicator of the whole thing.
 * So far, we have assumed that we knew about the effects of the basis but were unsure about the analogy between the model and target. Hence, we tried to establish the analogy by comparing effects. Another possibility is that we have strong antecedent reasons for the hypothesized analogy between model and target but are uncertain of the effects of the basis.
 * A common origin of the model and target can constitute evidence for an analogy.

Analogical reasoning

 * Positive analogy: the properties or relations they share
 * Negative analogy: the properties or relations they don’t share
 * Neutral analogy: the properties of which it is not known yet whether they belong to the positive or the negative analogy
 * Criteria for evaluating analogies:

Laws of society

 * Distinguish between the accidental and nomic necessities
 * Nomic having the general force of natural law: generally valid
 * Talking about social laws is talking about social reality (is it law-like?)
 * Hempel's deductive-nomological model: law + initial conditions, and then deduce the event/conclusion

Hempel
 * A law is a law because we make some reference of context in which the law holds (inherently contingent)
 * Conceive of a law as a universal hypothesis, a regularity of the following type: In every case where an event of a specified kind C occurs at a certain place and time, an event of a specified kind E will occur at a place and time which is related in a specified manner to the place and time of the occurrence of the first event.
 * "Hence, "therefore," "consequently," "because," "naturally," "obviously," etc., are often indicative of the tacit presupposition of some general law: they are used to tie up the initial conditions with the event to be explained; but that the latter was "naturally " to be expected as "a consequence " of the stated conditions follows only if suitable general laws are presupposed.

Scriven
 * A scientific theory is typically a system of propositions which organizes the evidence internally and in relation to other propositions of the system which concern certain (possibly hypothetical) entities or states;

Kincaid
 * The question “What is a law of nature?” is not one question but several
 * What kind of thing in the world is a law?
 * What are the kinds of statements that pick out laws?
 * Which particular alleged statements of laws actually do so?
 * Laws might be universal statements (atoms are XYZ) while at the same time not being universal regularities (which is a statement about the relation of events)
 * As things in the world, there might be laws that govern social phenomena. That claim could be true independently of social science’s ability to find and state such laws.
 * A force is causal in that it influences something. It is a factor in that it need not be the only influence present.
 * We go about deciding if there are laws/and which by whatever means we can use to decide if there are causal factors influencing social phenomena
 * The law at most tells us what regularities we would see if gravity were the only force; it does not tell us when or if that is the case and thus it does not generally entail any specific regularities on its own
 * Laws tell us reliably what would happen if things were different than they are (they support counterfactuals) in a way that statements picking out causal factors do not. Laws allow us to reliably predict what unobserved events will look like, but statements citing causal factors may not.
 * Laws are universal in that they do not refer to particular entities. Laws must state precise quantitative relationships; statements picking out causal factors need not.
 * There are laws in the natural sciences that do not refer to causes at all. Snell’s law, for example, tells us the relation between the angle of incidence and the angle of reflection for a wave. Causes are not mentioned, only functional relationships.
 * Laws in the social sciences claim to identify causal factors and make no commitment by themselves to what other causal factors there might be and how they might combine. Ceteris paribus is about applying the laws, not the laws themselves
 * Explanations in the social sciences are interpretations, not the citing of causes. However, it might be that some social phenomena – for example, aggregate economic activity – are best explained by identifying causal factors and that other phenomena – for example, symbolic rituals – are best explained in terms of meanings
 * It is possible that each individual act might be uncaused and yet that there be causes of the aggregate behavior of individuals.

McIntyre, 1993 - Complexity and Social Scientific Laws
 * Social phenomena are not complex as such, but only as described and defined at a given level of inquiry. A subject matter is defined by the questions that we ask about the phenomena we see. At any level the phenomena are too complex for us to find laws, for complexity just is a function of the level of inquiry that we are using.
 * What we single out as wholes, or where we draw the "partition boundary", will be determined by the consideration whether we can thus isolate recurrent patterns of coherent structures of a distinct kind which we do in fact encounter in the world in which we live.
 * This is an ontological issue: We cannot treat a subject matter as if it marks off a natural kind in the world, and we come upon it fully formed. At the ontological level, the subject matter is just 'matter'; it is only when we begin to ask questions about it that a 'subject' comes forward
 * It is only at the level of description, explanation, and inquiry which a particular type of investigation warrants that any distinctions begin to arise

Hayek
 * The complexity of human phenomena is not inherent, but derivative, in that it is dependent on the nature of our interests.
 * Statistics deals with the problem of large numbers essentially by eliminating complexity and deliberately treating the individual elements which it counts as if they were not systematically connected. It avoids the problem of complexity by substituting for the information on the individual elements information on the frequency with which their different properties occur in classes of such elements, and it deliberately disregards the fact that the relative position of the different elements in a structure may matter. In other words, it proceeds on the assumption that information on the numerical frequencies of the different elements of a collective is enough to explain the phenomena and that no information is required on the manner in which the elements are related.
 * The theory of evolution by natural selection (or Marx?) describes a kind of process (or mechanism) which is independent of the particular circumstances in which it has taken place on earth, which is equally applicable to a course of events in very different circumstances, and which might result in the production of an entirely different set of organisms. The validity of this general proposition is not dependent on the truth of the particular applications which were first made of it. The theory as such, as is true of all theories, describes merely a range of possibilities.

Explanation

 * Explanation as valid symbolic linking in the human mind between events and the conditions which produce them
 * Different models of explanation, e.g. deductive-nomological, statistical relevance
 * Little says: yes you can explain (i.e. he is a causal realist) but you don't need laws for explanation - you can use mechanisms
 * An explanation is an answer to a particular why-question
 * To explain an outcome is to demonstrate what conditions combined to bring it about–what caused the outcome in the circumstances, or caused it to be more likely to occur
 * The exception to this feature of causation is the rare set of cases where an outcome is ‘overdetermined’–that is, cases in which there are multiple factors present, each of which would bring about the outcome in isolation.
 * Social causes (as opposed to natural) involve the actions of individuals within the context of social institutions and the actions of others
 * Actions are purposive performances by individual agents within social and natural constraints; institutions are sets of rules embodied in the beliefs, values, and behaviours of groups of individuals.
 * Generally speaking, a cause is a condition that either necessitates or renders more probable its effect, in a given environment of conditions
 * An adequate explanation is a true description of underlying causal factors sufficient to bring about the phenomenon in question
 * The explanation will be constituted by an answer linking the explanandum to a set of possible explanans.
 * Role of context in explanation. A fruitful way of thinking about explanations is that they are answers to questions. Work on the logic of answers and questions suggests that any specific question and the answer to it must be spelled out by contextual factors.
 * Levels of explanation:
 * Historical: Why did the phenomenon occur in the first place?
 * Social/Political: Which social, political, economic processes have made it persist?
 * Comparative: Why did this phenomenon not occur in other places and at other times?
 * Causal: Which causal factors put the mechanisms in place, made it work the way it did, and locked it in subsequently?
 * Unlike the simplistic positivistic conception of science as elaborating event regularities, the process of uncovering and explaining significant causal structures and mechanisms, including geo-historically rooted and dynamic totalities, will usually be a painstaking, laborious, and time consuming, transformative activity, one that gives rise to results that will always be partial and contingent (and usually contested).
 * Explanation presupposes a description of the phenomenon to be explained. Phenomena as such are not explained. It is only phenomena as covered by a description which are capable of explanation, and then, when we speak of explaining them, it must always be with reference to that description. So an explanation of a phenomenon must be relativized to a description of that phenomenon. The form and the very possibility of explaining an event is, to a large extent, dependent on the description given of that event because different descriptions of the same historical occurrence call for different explanations of ‘what happened’.
 * Specific explanations belong to different level of reconstruction of the object of inquiry
 * It is not the sources which define the questions asked by a discipline, but the questions which determine the sources

---> check Russo's http://www.slideshare.net/titalla/russo-unam2

EOW

 * Level of abstraction is in no ways synonomous with, or even indicative of, a systems level focus. all writing is an abstraction.
 * The level usually refers to the unit of analysis you are talking about...micro level is usually focused on agents and you can be more concrete and specific in analysing policymaking outcomes ...meso level generally means the focus is on institutions and how actors navigate within those institutional boundaries...
 * The level of abstraction is distinctive from the micro and macro ....micro and macro refer to the UNIT OF ANALYSIS (person is micro - firm is meso - nation is macro, to take one example). The level of abstraction refers to the DEGREE OF COMPLEXITY involved (concrete events as opposed to more stylized representations of events). In principle it is entirely possible to speak about the person (micro unit) in a thoroughly abstract fashion, while speaking about the nation (macro unit) in a thoroughly concrete fashion.

Mechanisms

 * Is it the mechanism that is complex or simply our description/model of it? Does our level of (complexity of) explanation depend on the audience?
 * If the output is wrong, it doesn't follow that the input was so, too
 * Advantages of looking at mechanisms:
 * Extrapolating from past (e.g. financial crises) does not work for an uncertain future - prediction is impossible but if I understand what the general mechanisms for crisis are, I can intervene
 * Take the evidence on the link between "reduced calorie intake → longevity". I might infer that I should eat less but looking at the mechanism might show that the crucial variable here was "weight loss/absence of obesity", which might also be achievable by other means → mechanisms allow possibly less costly interventions with the same effect

Definitions

 * Mechanisms are effect-generating processes. To say that mechanisms are effect-generating processes does not imply that mechanisms invariably generate empirically observable effects. It is always possible that the presence of one mechanism can block the effects of another. This does not deny that mechanisms can be viewed as effect-producing processes but simply that the effects never appear as discrete events in our empirical observations.
 * To identify a mechanism is to give an account of the way it produces specific kinds of effects.
 * Finding mechanisms can help with causal inference
 * Nothing is a mechanism as such, only a mechanism for a particular phenomenon
 * The first thing to specify is what mechanism is under analysis and in what context. An isolated description of activities of entities is possible, but what we need for informative mechanistic explanations is role-functions, which presuppose the identification of a context. In social science, the same factor often performs different roles, or functions in different contexts.
 * A mechanism for a behavior is a complex system that produces that behavior by the interaction of a number of parts, where the interactions between parts can be characterized by direct, invariant, change-relating generalizations.
 * A simple definition of mechanisms: a delimited class of events that alter relations among specified sets of elements in identical or closely similar ways over a variety of situations. And processes are concatenations of mechanisms: regular sequences of such mechanisms that produce similar (generally more complex and contingent) transformations of these elements
 * Mechanisms are entities and activities organized such that they are productive of regular changes from start or set-up to finish or termination conditions. (Machamer, Darden and Craver) → mechanisms are active.
 * Machamer, Darden & Craver (MDC) define mechanisms as: mechanisms can be said to consist of entities (with their properties) and the activities that these entities engage in, either by themselves or in concert with other entities. These activities bring about change, and the type of change brought about depends upon the properties and activities of the entities and the relations between them. A mechanism, thus defined, refers to a constellation of entities and activities that are organized such that they regularly bring about a particular type of outcome, and we explain an observed outcome by referring to the mechanism by which such outcomes are regularly brought about
 * When you have a mechanism, you have at least one causal relation. MDC think that the activities in mechanisms are little specific kinds of local causes.
 * Mechanisms have (components/factors that fulfill) functions. The function of an entity is the role it plays in the overall behaviour of a containing system - a (complex) mechanism involves different factors that explain how people self-rate their health. Each factor in this mechanism has a role, or function. For instance, the function of alcohol consumption is to relieve stress, and this works, in the sense of having the capacity to relieve stress
 * Giving the function of components of mechanisms, where the surroundings set the role of the component.
 * What is needed to understand functions of mechanisms more precisely is isolated descriptions in some cases, and role-functions where a containing system is specified
 * A mechanism is a structure performing a function in virtue of its component parts, component operations, and their organization. The orchestrated functioning of the mechanism is responsible for one or more phenomena
 * A causal mechanism is (i) a particular configuration of conditions and processes that (ii) always or normally leads from one set of conditions to an outcome (iii) through the properties and powers of the events and entities in the domain of concern
 * Mechanisms help us decide what causes what—i.e. they help with causal inference. Mechanisms are concerned with what links or connects putative cause-and-effect variables or events, going beyond mere correlation, they fall within the school of thinking concerned with productive causality
 * It is the case that, whenever the conditions of the mechanism are satisfied, the result regularly ensues
 * Mechanisms as continuous causal processes and componential analysis
 * Mechanism in the componential sense thinks that explanation proceeds in explaining a complex whole by invoking the elements comprising the whole and their interaction.

Identifying mechanisms

 * It is necessary to have a hypothesis of the mechanisms that link the variables before we can arrive at a justified estimate of the relative importance of the causal variables in bringing about the outcome (i.e. are busy things, constantly doing something to create or sustain the phenomenon)
 * Finding mechanisms means specifying the mechanisms by which change is brought about in social processes. Central to an adequate explanatory theory is the specification of the mechanism that is hypothesized to underlie a given set of observations
 * Mechanism discovery proceeds in three steps:
 * 1) Describe the phenomenon or phenomena;
 * 2) Find the parts of the mechanism, and describe what the parts do;
 * 3) Find out and describe the organization of parts by which they produce, in the sense of bring about, the phenomenon
 * At this level of abstraction, entities are the variables having a causal role (e.g. alcohol consumption, psychological distress or physical health) and the activities are the (statistical) interactions between these variables (e.g., self-rated health is correlated with these variables). At this level of abstraction, entities and activities are captured by the variables in the model and the (statistical) interactions between these variables.
 * In quantitative social science statistical analyses, and, more specifically, recursive decompositions are used to model the structure, that is the organization, of the mechanisms at hand. Recursive decomposition refers to the process whereby any complex informational event at one level of description can be specified more fully at a lower level of description by decomposing the event into a number of components and processes that specify the relations among these components
 * Social mechanisms can be identified, described and isolated, insofar as the key entities and their activities and interactions are specified. This includes identifying the function of both actors and their actions, other entities and their activities, and ultimately of the complete mechanisms for a given societal structure
 * In practice, the process of mechanism discovery helps you understand the phenomenon, and it might lead to significant redescription of the phenomenon itself. Looking for the organization might lead you to an entity you hadn’t previously noticed.
 * We can directly identify the features of purposive action within given structures that make the mechanism work. Human actions and refrainings are the ‘stuff’ of social causation, and features of human agency underwrite the ‘necessity’ of social mechanism
 * Mechanisms implicate regularities but these regularities are low-level and may not be observable at macro-level social behavior (e.g. because of the mixing of several causal processes and the possibility of countervailing mechanisms in play)
 * Theories of the middle range: accounts of the real social processes that take place above the level of isolated individual action but below the level of full theories of whole social systems.
 * Horizontal vs. vertical mechanisms: a continuous causal process involves specifying the intervening steps between a given cause and its ultimate effect, a horizontal mechanism. Identifying the components of a complex whole is giving a vertical mechanism – explaining the behavior or causal capacities of a complex whole by identifying component elements and their relations.
 * Mechanisms at different levels of detail: we always have a causal process picked out under a description that can be at various levels of detail.
 * We can be most confident in statements of lawlike regularities when we have an account of the mechanisms that underlie them.
 * Law-based necessity in the form of a social mechanism.
 * Causal realists maintain that we can only assert that there is a causal relationship between X and Y if we can offer a credible hypothesis about the sort of underlying mechanism that might connect X to the occurrence of Y
 * Things and events have causal capacities: in virtue of the properties they possess, they have the power to bring about other events or states
 * Identifying causal relations requires substantive theories of the causal powers (capacities, in her language) that govern the entities in question. Causal relations cannot be directly inferred from facts about association among variables.
 * Attribution of social causation depends unavoidably on the formulation of good, middle-level theories about the real causal properties of various social forces and entities–the social mechanisms that convey social causation
 * Identify crucial causal mechanisms that recur in a wide variety of contention, but produce different aggregate outcomes depending on the initial conditions, combinations, and sequences in which they occur
 * Don't confuse a mechanism with a family-resemblance/umbrella term that captures a number of different instances of collective behaviour and agency
 * The level at which we find real causal connections in the social world is the level of the socially situated and socially constituted individual in interaction with other individuals.
 * Kinds of social mechanisms:
 * Agent-centered explanations
 * Rational‐intentional mechanisms: Why do empires establish a policy of rotating senior military officials? Because emperors want to avoid the creation of warlords
 * Imitation mechanisms: Why did phenomenon X become so popular? Because it was successful for some, and others copied the offense in the hope that they too would be successful, too
 * Conspiracy mechanisms: Why did A introduce B? because C had a powerful interest in it
 * Aggregative mechanisms: aggregate consequences of individual‐level strategies
 * Social-influence explanations
 * Mentality mechanisms: behaviour is changed by changing beliefs and attitudes).
 * Social network mechanisms (information and norms proliferate through concrete sets of social relationships among individuals; why was the Soviet military system less adaptive in combat than the Israeli military system? Because information flow among officers and troops was more rapid and more bidirectional in the latter than the former.
 * Evolutionary mechanisms: Why does the level of firm efficiency tend to rise over time? Because the net efficiency of a firm is the product of many small factors. These small factors sometimes change, with an effect on the efficiency of the firm. Low efficiency firms tend ultimately to lose market share and decline into bankruptcy. Surviving firms will have features that produce higher efficiency
 * System-level features of the environment of social change
 * Filtering mechanisms: Why are passengers on commercial aircraft better educated than the general population? Because most airline passengers are business travellers, and high‐level and mid‐level business employees tend to have a higher level of education than the general population.
 * Critical mass mechanisms: A new social networking site experiences slow growth for the first eighteen months of operation until it reaches N users; it then takes off with rapid growth for the next eighteen months. We attempt to explain this change by arguing that N is a critical mass of users, stimulating much more rapid growth in the future.
 * Path‐dependency mechanisms: → momentum of prior social choices; Why do we still use the very inefficient QWERTY keyboard arrangement that was devised in 1874? Because this arrangement, designed to keep typists from typing faster than the mechanical keyboard would permit, was so deeply embodied in the typing skills of a large population and the existing typewriter inventory by 1940 that no other keyboard arrangement could be introduced without incurring massive marketing and training costs
 * One example of a social mechanisms are transportation systems: extension of railway network stimulating the growth of new towns; a direct airline link between A & B causes more rapid spread of a disease; breakdown of the administration of the rail system leads to logistics bottlenecks and military defeat of the French army in the Franco‐Prussian War → A rail system provides convenient transportation among a number of places, while providing no service at all between other pairs of locations. So a rail system certainly has direct effects on social behaviour; it structures the activities of the several million residents of a major city by making some places of residence, work, shopping, and entertainment substantially more accessible than other places. In brief, a rail system has definite social effects. It creates opportunities and constraints that affect the ways in which individuals arrange their lives and plan their daily activities
 * So a rail network has structural and causal characteristics at multiple levels. The physical network itself has structural characteristics (nodes, rates of travel, volume of flow of passengers and freight). This can be represented statically by the network of tracks and intersections that exist; dynamically, we can imagine a ‘live’ map of the system representing the coordinated surging of multiple trains throughout the system, throughout the course of the day. The railroad organization has a bureaucratic structure–represented abstractly by the organizational chart of the company, but embodied in the internal processes of training, supervision, and recruitment that manage the activities of thousands of employees. And the social and technical ensemble that these constitute in turn creates an important structure within the social landscape, in that these physical and human structures determine the opportunities and constraints that exist for individuals to pursue their goals and purposes
 * Transportation is a robust family of causal mechanisms that mediate many important social processes and outcomes. So transportation is a causal mechanism whose microfoundations are especially visible, and whose causal consequences are often very large.
 * Quantitative sociological reasoning is not analogous to epidemiological reasoning, for this reason: there is a substantially greater possibility of multiple causal pathways and conditions in the case of the social world, leading to the result that discovery of gross correlations between factors is unlikely to correspond to unique causal mechanisms and pathways leading to the observed outcome.
 * Entities frequently interact with each other in mechanisms
 * We can not be satisfied with disciplinary accounts of mechanisms. We have to worry about how different sciences manage to build a single mechanistic explanation of a phenomenon like memory.
 * That a mechanism is identified correctly, in fact, does not exclude that other mechanisms might be present and escape cognizance by the social scientist. And if the other mechanisms happen to exert causal influence in the opposite direction, the overall effect of their joint operation would be nullified.

Context

 * We understand the function of mechanism components by their role in the surrounding mechanism
 * Context affects what is considered a part of the mechanism, and what is not - only some parts are relevant, think of the relevant ones as working parts.
 * If maniuplating the component affects the phenomenon, and manipulating the phenomenon affects the part, they are mutually manipulable, and we have found a working part of the mechanism.

Organization

 * Organization is the final phase of building mechanistic explanations
 * The same entities arranged in different ways might do something different
 * Most generally, organization is whatever relations between the entities and activities discovered produce the phenomenon of interest: when activities and entities each do something and do something together to produce the phenomenon
 * This captures simple spatial and temporal relations, and extends all the way to complex forms of organization such as feedback, homeostasis, and limit cycles
 * Organization is whatever the entities and activities do and do together to produce the phenomenon
 * Mechanisms can act to change their own organization
 * Spell out organization as referring to the practices and norms (internal rules and modes of functioning) that govern the relationships between agents
 * Structure is supposed to give stability to the mechanisms. This is an important feature because in social contexts mechanisms are very fragile and contextual. So to know how stable a social structure is means precisely to know the boundaries of the applicability and validity of a study.
 * We might not know what will happen on the basis of the entities and activities because we don’t know yet about the organization. In particular we need quantitative dynamic models of the organization to know what the system overall will do
 * Entities and activities can be nested within different levels of the phenomenon

Methodological localism/individualism

 * Structuralist theories (capitalism causes XYZ) which treat social constructs as macro-entities with their own causal power do not give us an answer to the HOW. In other words, we want to know something about the lower-level mechanisms through which large social factors impact upon behaviour, thereby producing a change in social outcomes. We want to know quite a bit about the ‘microfoundations’ of social causation
 * Social behaviours are carried out by individuals, and individuals are influenced only by factors that directly impinge upon them (present and past). In other words, one's current political judgments and preferences are caused or influenced by a past and current set of experiences and contexts. So the individual is socially influenced and formed at every stage. But here is the important point: every bit of that social influence is mediated by locally experienced actions and behaviours of other socially formed individuals. ‘Catholicism’, ‘Chicago culture’, and ‘union movement’ have no independent reality over and above the behaviours and actions of people who embody those social labels.
 * Thus, concrete social mechanisms are embedded in the actions of social actors
 * The foundation for the explanation of social action and outcome is the local, socially located and socially constructed individual person. The individual is socially constructed, in that her modes of behaviour, thought, and reasoning are created through a specific set of prior social interactions. And her actions are socially situated, in the sense that they are responsive to the institutional setting in which she chooses to act. Purposive individuals, embodied with powers and constraints, pursue their goals in specific institutional settings; and regularities of social outcome often result.
 * Methodological localism affirms that there are large social structures and facts that influence social outcomes. But it insists that these structures are only possible insofar as they are embodied in the actions and states of socially constructed individuals. The ‘molecule’ of all social life is the socially constructed and socially situated individual, who lives, acts, and develops within a set of local social relationships, institutions, norms, and rules.
 * A theory of social mechanisms: There is such a thing as social causation. Institutions, structures, demographic features, and widespread social arrangements have specific causal effects on the societies in which they exist. The mechanisms that convey these causal powers exist in a social ontology of socially situated and socially constituted individuals, acting and refraining in response to their own motivations and beliefs and the rules, conventions, and constraints that exist around them.
 * Two approaches to social mechanisms
 * Agent-based perspectives: aggregate the results of individual-level choice into macro-level outcomes (micromotives and macrobehavior; pays primary attention to the motives and reasonings of agents within a given set of constraints)
 * Social-influence theories: identify the factors that work behind the back of agents to influence their choices(identify socially salient influences such as race, gender, educational status, and to provide detailed accounts of how these factors influence or constrain individual trajectories–thereby affecting sociological outcomes; attention to the broad social factors that influence individual agency)

Issues

 * Absence: we seem to cite absences as causes. For example, the absence of oxygen in Alice’s bloodstream causing her death. Since biological mechanisms are contextual, always described in terms of the normal operation of the body, the absences of normally available factors are naturally described as causes. But how can an absence be involved in a mechanism that links cause and effect?
 * Mechanisms, as causal explanations, are often described merely qualitatively - This means it can be hard to identify causally relevant properties
 * Some explanations are NOT describing a mechanism. They might miss at least one element of a mechanistic explanation:
 * Phenomenal description: precise description/characterization of the phenomenon does not equal mechanistic explanation
 * Entities without activities
 * Activities without entities
 * No organization

Microfoundation

 * A microfoundational approach to social causation: the causal properties of social entities derive from the structured circumstances of agency of the individuals who make up social entities—institutions, organizations, states, economies, and the like
 * The microfoundations debate deals with the question of whether holist causal claims need microfoundations: Should holist causal claims be supplemented by accounts of underlying mechanisms at the level of individuals. Holist causal claims are ones in which both the cause and the effect are described in holist terms.

Holism vs. individualism

 * All theses of methodological individualism and holism express a view as to what is the proper focus of explanation. By implication, they involve some view of what an explanation is.
 * Ontological holists contend that social phenomena exist sui generis or over and above individuals, whereas ontological individualists deny this
 * What it takes for social phenomena to exist sui generis:
 * The causal overriding criterion: Social phenomena such as social organizations exist over and above individuals insofar as they have causal powers that are independent of, and override, the causal powers of individuals
 * The translation criterion: Social phenomena such as social organizations exist over and above individuals insofar as terms, like “nation” and “school,” that refer to these phenomena cannot be translated into statements about individuals.
 * The composition criterion: Social phenomena such as social organizations exist over and above individuals insofar as they are not merely composed of ensembles of individuals.
 * The determination criterion: Social phenomena such as social organizations exist over and above individuals insofar as individuals do not non-causally determine what kinds of organizations, properties, and the like, are being instantiated.
 * The agency criterion: Social phenomena such as social organizations exist over and above individuals insofar as they qualify as group agents that have attitudes supervenient upon the attitudes of individuals.
 * Today, the majority of theorists tend to be concerned with the composition and determination criteria
 * In discussions of the ontological status of social phenomena, the term “explanation” is sometimes used to refer to ontological analyses of how individuals must be related, what properties they must have, and so on, in order to constitute social phenomena of various sorts. Explanations of this sort map out non-causal or synchronic relationships between social phenomena and individuals. They should be distinguished from explanations which map out the causal or diachronic relations between events or states involving social phenomena and individuals
 * The debate about dispensability revolves around the question of whether individualist (or individual level) or holist (or social level) explanations may, and should, be dispensed with within the social sciences
 * Popular types of methodological individualism explanations: rational choice, appeal to individuals' dispositions etc.
 * Popular types of holistic explanations: accounts that point to the statistical properties of groups, explanations on how social organizations bring about various effects.
 * Three commonly mentioned forms of explanations are
 * Functional explanations (favored by holists): explain the continued existence of various social phenomena by reference to their function or effect, in some society.
 * Intentional explanations (favored by individualists): explain actions by appeal to individuals’ reasons or motivations for their actions
 * Straightforward causal explanations

Objectivity

 * What is the relation between objectivity, truth and validity?
 * Validity is approximating truth
 * Correspondence theory of truth
 * In the social sciences, most of our entities are not clear-cut
 * How is objectivity related to completeness and laws?
 * Objective probability as mind-independent, subjective probabilities (Bayesian) as mind-dependent (as dependent on the epistemic agent)