Sociosystemics and Statistics

Igor Mandel [email protected], Telmar, Inc. I

Why had the promise of the nineteenth century been dashed? Why had much of the twentieth century turned into an age of horror or, as some would say, evil? The social sciences, which claimed such questions as their province, could not provide the answer. Nor was this surprising: they were part, and a very important part, of the problem. Economics, sociology, psychology and other inexact sciences – scarcely sciences at all in the light of modern experience – had constructed the juggernaut of social engineering, which had crushed beneath it so many lives and so much wealth…

Paul Johnson, The Modern Times, 1992, p.776 (my highlight here and anywhere – I.M.)

New York INFORMS January 19, 2011 Introduction An old politically incorrect joke

Two beggars are sitting across from each other on the stairway of a Christian cathedral. One of the beggars wears a chain with a large holy cross on his chest, and the other wears a chain with a large Star of David.

As Mass lets out, many give money to the Christian as they walk down the stairs, but no one gives money to the Jew.

The priest follows at the end of the group. As he heads down the stairs, he stops by the Jew and says in a friendly tone, “You know, this is a place of worship for Christians; it’s not very good for you to be here. You see, you are sitting empty-handed. Why don’t you try somewhere else?”

“Well,” the Jew answers, “you are right, I should think about it.”

The priest leaves, and the Jew addresses the Christian: “You see, Shlomo, this goy is trying to teach us how to do our business!” Introduction This joke touches several topics highly relevant to modern social sciences and to the current presentation:

1. The beggars exploit the fact that people leaving the church are more psychologically inclined to donate than in other situations (situational marketing)

2. They deliberately use the “decoy” approach (the “Jew” plays a role of decoy for the “Christian’s” success), which is proven to be very important in the process of decision making (Ariely, 2009)

3. Their approach works only because there is strong nationalism there (role of social factors in economics is poorly recognized in neoclassical economics theory - Smelser and Swedberg, 2005) 4. The beggars, in turn, demonstrate a very tight ethnical connection, for they have to have a high level of mutual confidence to succeed - all proceeds are collected just in one person’s hands (a “homogeneous middlemen” theory – Landa, 2008) 5. Their real relationships are hidden from the public’s sight, which makes business a success (an unobserved but critical part of social reality)

6. The priest plays the role of a naive liberal or a researcher who “wants to do better”, but he doesn’t bother to investigate the real moving forces in the given situation (this is a typical position for many)

7. A sudden revelation of a real mechanism creates two effects: shock and laughter (a very good illustration of the thesis that humor and truth are closely related – I. Kant and others) Introduction This analysis shows that important questions of social life can be approached from almost any perspective. This is so because we are agents, spectators, and researchers of it at the same time and know enough from our own experience. It creates unique problems unfamiliar to other sciences – a researcher, looking at the society, is the product of this very society and as such is subject to all its and prejudices.

Is there such a thing as an “objective social science”? This was my first thought leading me in the direction of sociosystemics.

Another was an attempt to find some basic reference points in the ocean of social knowledge just for my own comfort (with the hope that it may be constructive for others as well). This “déjà vu” phenomenon has always irritated me: I remember a idea, but not its source. Can it be helped?

The third was that a person’s ability to simultaneously pay attention to many important affecting factors is very limited, and science should supposedly assist in that in a maximally human way. But in fact it doesn’t. Therefore, I was trying to understand how to make science more “user friendly”. When writing, I’ve always felt strong discomfort thinking that I’ve missed something very important just because I don’t know where it is – a feeling well familiar to any writer, I guess.

Finally, I’ve always felt that huge ideological and methodological differences between people can mainly be explained by either their unwillingness, or incapability (or both) to learn each other’s opinions or theories. Of course, I realized that the very unwillingness is usually explained by material rather than other motives. Yet, I felt the need to “contribute” to this never solved issue by proposing at least the possibility to learn the opinions of others in a more objective fashion.

I will give the definition of sociosystemics later; it intends to find some unifying frame for all social sciences, and for that reason, inevitably touches a very wide range of topics. A one hour presentation does not allow me to describe each of them deeply enough; it rather raises questions and poses problems than offers answers and solutions. Much more supportive material will be presented in the articles that follow. The content of the presentation at large is as follows:

1. What is wrong with the current status of social sciences 2. How well statistics, as a universal tool, contributes to social knowledge integration 3. Main components of sociosystemics 1. What is wrong with the current status of social sciences? There are two types of reasons why one cannot consider the status of social sciences satisfactory – it’s not because the science “doesn’t know everything” (any science doesn’t), but because it often offers just surrogates instead of the real scientific findings. A. Failures of policies supposedly equipped with scientific backing The use of Gant charts in construction industry in 1970s failed in the USSR because they needed everyday 1 adjustments, which made no sense, yet I saw the same picture couple of months ago in an American company. Afghan and Iraq wars have changed their geopolitical purposes for the USA in the last 9 years and do not serve 2 anti-terrorist politics anymore, yet they are still continued in ambiguity of goals (Friedman 2010). The USA has the highest health care expenses in the world, but has a high mortality rate for critical diseases, 3 slide 6.

4 There is an abundance of theories about the real estate market, but its crisis has not been predicted or avoided. For decades, this country has heavily invested in education, yet it has one of the lowest levels of mathematical 5 knowledge among school students in all OECD countries, slide 7.

6 Financial modeling uses the finest minds, but failed to predict the latest crisis and make sense of it. Huge amount of money has been spent on climate change research, yet there is lack of evidence for both Global 7 Warming and its human nature coupled with data manipulation scandal (Climategate) There is a common need for better measurement, yet there is strong resistance to it. J. Stalin didn't like results of 1937 Census showing a huge population loss after collectivization, so he canceled it. Modern advertising 8 agencies don't like that estimates of reach and frequency have changed because of better methodology. Multiculturalism policy has been promoted for decades, yet it has failed as declared by chancellor A. Merkel in 9 Germany in 2010 http://www.bbc.co.uk/news/world-europe-11559451 There are corporate scandals with Enron, MCI, B. Madoff, yet world leading auditing companies that were not 10 able for whatever reasons to reveal the fraud retain good reputation .

11 Chase bank keeps sending me offers to open a checking account, yet I have for years (in spite of all data mining) 1. What is wrong with the current status of social sciences?

A. External failures of policies supposedly equipped with scientific backing High level of health expenses doesn’t guarantee good health as a result Pictures like these show several effects: Health expenses - Cancer mortality 1. Aggregate data are not good measures Source: UN data (USA - 97th place out of 190) http://unstats.un.org/unsd/demographic/products/dyb/dyb2007/Table01.xls for complex processes, but alternative ones often do not exist 350 2. Ineffective spending structure, especially 300 250 in the USA, becomes obvious 200 3. Life style and genetics play key role, 150 USA

but they are clearly understudied 100 age standardized age 50

Health expenses - Cardiovacular mortality, Morocco Cancer mortality, per 100, Source: UN data (USA - 26th place out of 190) - http://unstats.un.org/unsd/demographic/products/dyb/dyb2007/Table01.xls - 1,000 2,000 3,000 4,000 5,000 6,000 7,000 Health expenses, $/person, 2005 900 800 700 600 500 400 300 USA

200 100,000, standardized age Cardiovacular mortality per Cardiovacular 100 Chile Japan - Guatemala - 2,000 4,000 6,000 8,000 Health expenses, $/person, 2005 1. What is wrong with the current status of social sciences? A. External failures of policies supposedly equipped with scientific backing Level of mathematical knowledge among school students OECD conducted a massive study of the level of knowledge in 55 countries around the world asking students of the same age sets of practically identical questions Distribution of 15 years old school students by mathematics scores. Source: OECD data, 2007 in mathematics, science, and literacy. Here are http://w w w .oecd.org/dataoecd/31/0/39704446.xls tab.6a some results for mathematics. 50.0 USA’s students are among the worst in OECD Brasil 45.0 (i.e. most developed) countries and worse than 40.0 many others. It has many far reaching 35.0 Indonesia Korea consequences in modern economics where technology 30.0 and innovation plays an increasingly important role. 25.0

But it’s the outcome of the decades of development, category 20.0 15.0 USA backed by advanced scientific recommendations Finland and massive financing – maybe, the highest in the 10.0

Percent of students in each in students Percent of 5.0 world! 0.0 Number of students with worst scores per one student 1 2 3 4 5 6 7 w ith best scores for OECD countries Scores (1 - no knowledge, 7 - extremely good knowledge) Turkey 21.0 Mexico (68, 456) If the first example with mortality 16.0 USA demonstrated ineffective spending

11.0 structure and inability to pursue a Estonia Canada healthy life style, this one shows direct 6.0 failure of the long-term federal and state education policy

1.0 Index value, persons Indexvalue, 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 (4.0) Country number 1. What is wrong with the current status of social sciences? B. Internal problems within social sciences There are many internal problems, either common for all social sciences, or specific for some of them. The contradiction between the need in complex expert judgment and technical impossibility to provide it in a reasonable way is one of the biggest challenges of our information era. Internet, paradoxically, just greatly contributes to it due to its phenomenal mix of unfiltered spam and quality information. Sociosystemics is intended to contribute to filling the gap, and the whole idea of it has emerged for that reason. B1. Problems common for all social sciences

Exponential growth of information and linear growth of knowledge (Rescher, 2006), and as a consequence, bigger 1 difficulties to extract this linear essence from ever growing exponential volume Fragmentation and specialization; lack of mutually penetrating works, in spite of the integrative character of 2 social life 3 Lack of common language among specialists in different fields Concentration on problems interesting for scientific community itself rather than for society as a whole, which 4 greatly affects emerging of new sub-fields of science, slide 9 5 Materialistic, career, and corporate interests of scientists as any other people (Snooks, 1998, p.173-174) 6 Ideological coloring of almost all social studies 7 General cognitive biases 8 Specific scientific biases, slide 11 Belief in statistics as in a universal tool and ignoring statistical internal problems, which are much more severe 9 than they seem, Part 2 Lack of accepted approaches to test the correctness of different statements, which allows a very high level of 10 unfounded speculations and not necessarily sharp differences in opinions, slide 11 1. What is wrong with the current status of social sciences?

B1. Problems common for all social sciences There are three main motivations for scientific activity:

1. Demand from the society, when it intends to solve some recognized problems posed by governments or corporations (military needs that stimulated the big part of physics in the first half of the 20th century; political needs that boosted space studies; demand from pharmaceutical companies for new drugs, and so on). Social sciences, in general, experience more demand of that type from the business side (operational research, marketing, etc.) rather than from the government, which usually “knows what to do”.

2. Demand from the scientific community. Internal interests of scientists are of two types: materialistic, based on the need to support themselves and group of dependent people (for which the grant system works generally as the main source), and idealistic, which is motivated by purely scientific interests. The dictum “Science is the best way to satisfy personal curiosity at the state’s expense” describes it perfectly.

3. Emotional and intellectual need in self-actualization is a specific but very important part of idealistic motivation which expresses the deepest interrelated “beauty instinct” and “knowledge instinct” (Perlovsky, 2007). It may not need any compensation; it drives people by itself. It has almost the same mechanism as the drive for art and poetry, and it most likely stimulates the same areas in the brain. The famous Einstein’s saying "Dostoevsky gives me more than any scientist, more than Gauss” explains the idea. If it works for physics, it works even more for social sciences. Social thinkers from antiquity were inspired by artists, poets, and writers, and they reflected it in their oeuvre. As a result, a huge part of the social texts is in fact semi-fiction, which cannot be tested and verified. It represents one of the main problems in social sciences classification and formalization (part 3). 1. What is wrong with the current status of social sciences? B1. Problems common for all social sciences

Selection of scientific cognitive biases from the larger list http://en.wikipedia.org/wiki/List_of_cognitive_biases 1 Tendency to ignore available statistical data in favor of particulars. 2 blind spot Tendency not to compensate for one's own cognitive biases 3 Choice-supportive bias Tendency to remember one's choices as better than they actually were. 4 Tendency to search/interpret information in a way that confirms one's preconceptions Tendency to test hypotheses exclusively through direct testing, in contrast to tests of 5 possible alternative hypotheses. Enhancement or diminishing of a weight of other measurement when compared with 6 Contrast effect a recently observed contrasting object. Experimenter's or Tendency for experimenters to believe data that agree with their expectations for the 7 Expectation bias outcome of an experiment and to do the opposite with contradicting data Tendency to place too much importance on one aspect of an event; causes error in 8 Focusing effect accurately predicting the utility of a future outcome. Using an approach or description of the situation or issue that is too narrow. Also 9 Framing framing effect – drawing different conclusions based on how data are presented. 10 Information bias Tendency to seek information even when it cannot affect action. 11 Mere exposure effect Tendency to express undue liking for things merely due to familiarity with them. 12 Normalcy bias Refusal to plan for, or react to a disaster which has never happened before. 13 Tendency to judge harmful actions as worse than equally harmful inactions. Tendency to judge a decision by its eventual outcome instead of based on the quality 14 of the decision at the time it was made. 15 Semmelweis reflex Tendency to reject new evidence that contradicts an established paradigm. 16 Tendency to like things to stay relatively the same. Formation of beliefs and making decisions according to what is pleasing to imagine 17 Wishful thinking instead of appealling to evidence or rationality. 1. What is wrong with the current status of social sciences? B1. Problems common for all social sciences Differences in fundamental philosophical questions (based on survey of more than 3000 philosophers, http://philpapers.org/surveys/results.pl?affil=Target+faculty&areas0=0&areas_max=1&grain=fine) Topics Alternative positions Comments (I.M.) God theism atheism 15% 73% Free will compatibilism libertarianism no free will C. defines 'free will' in a way that allows it to co-exist with 59% 14% 12% determinism S.R.: the world described by science is the real world, Science scientific realism anti-realism independent of us. A-R: non-observable entities (electrons) do not 75% 12% exist or just "instruments" for control Truth correspondence deflationary epistemic C.: T. corresponds to reality. D.: a statement "something is T." is the same as statement itself. E.: T. as a derivative of belief, 51% 25% 7% acceptance, verification, etc. Laws of nature Humean non-Humean H.: there is no physical necessity in laws of nature (regularities 25% 57% instead). N-H.: Necessitarianism Knowledge empiricism rationalism E.: knowledge from evidence gathered via sense experience. R.: 35% 28% relies upon reason, can incorporate innate knowledge. Meta-ethics moral realism anti-realism M.R.: ethical sentences express propositions, some of which are 56% 28% true objectively. MA-R.: there are no moral facts or properties Moral motivation internalism externalism 37% 30%

Normative ethics deontology consequentialism virtues D.: moral is based on rules or duties. C.: based on expected 26% 24% 18% consequences. V.E. - based on the person's virtue or intentions C.: individuals are shaped by the cultures of their Politics communitarianism egalitarianism libertarianism communities.E.: all humans are equal in fundamental worth or moral status. L.: redistribution of power from the state to 14% 35% 10% voluntary associations of free individuals 1. What is wrong with the current status of social sciences?

B2. Problems specific for separate social sciences - economics I will consider here only economical theory problems without going into similar aspects of other social sciences. It is the most developed and by far the most formalized social science, yet it has many deep fundamental unresolved drawbacks, of which only some are listed here. The main formalism of the dominant neoclassical economics theory, reflected in hundreds of textbooks and main stream publications, is based on the physical paradigm of pre-quantum physics (Mirowski, 1989). It assumes rational (“natural”) behavior of agents, individual utility maximization, perfect information about the environment, tendency of systems to be in equilibrium, and so on.

The last 3-4 decades witnessed wide and deep attacks on that idealized picture, which I would classify tentatively into four interrelated groups: a) Critique from within, when many economists started to challenge and modify different aspects of the theory, yet remained within the “mainstream”. It’s enough to mention concepts of bounded rationality and imperfect information, prospect theory of decision making, experimental economics, and names such as G. Becker, W. Vickery, J. Stiglitz, D. Kahneman, V. Smith, T. Shelling, and others. b) Critique from sociological point of view claims that the formal description of economics is incomplete, doesn’t account for many social aspects, and brings many arguments for that; a new branch of economic sociology has been developed (Smelser and R.Swedberg , 2005); partly the neo-Austrian school is positioned here, too (Salerno, 2010) c) Critique from psychological and biological points of view, especially sharp in behavioral economics (Ariely, 2009), convincingly demonstrates that main postulates, especially those of rational and consistent behavior, are just wrong in practice. d) Critique from the point of view of other sciences, especially physics and science of complexity, most clearly expressed in creation of econophysics (McCauley, 2004), which blames economics for ignoring physical and systematic principles in general. (Mandel and Kuznetsov, 2009). 1. What is wrong with the current status of social sciences? B2. Problems specific for separate social sciences - economics

All these critiques seem very sound. A systematic discussion of many of these challenges is collected in “The economics anti-textbook” (Hill and Myatt, 2010), and I don’t need to repeat it here. I would rather formulate what I consider the most important problems within economical science from my own prospective.

1. Economical models are usually extremely complicated and formal, but they are normally tested (if they are) by using econometrics techniques, which are in fact some modification of statistics with all troubles coming with it (Part 2). These testing procedures as a rule have nothing to do with the complexity of the original model (say, a model yields strong non-linearity, but testing uses linear procedures, etc.).

2. In my experience, I’ve never seen the use of any economic models in any situation; moreover, the evidence often says something completely against the theory (slide 16), and empirical considerations always win. No manager I’ve known (myself included) has ever used any models but simple statistical ones.

3. The most important point of massive critique undermining the basic premises of the neoclassical paradigm is that one cannot just say (as many do), that “in principle”, ceteris paribus, theory works. In fact it doesn’t. Mathematical constructions in economics, being very interesting mental tools to explore some plausible ideas, are not to be used in reality as they are - a specific form of intellectual self-actualization having nothing to do with real life (see slide 10). When I see a “model of war”, where the main concern is about obtaining Nash equilibrium (A. Jacobsson 2009), I cannot think otherwise. So many economical theoretical works are exactly of that type… The problem is – no one can say that particular model just “approximates” reality, because there is no way to estimate the level of approximation. 1. What is wrong with the current status of social sciences? B2. Problems specific for separate social sciences - economics Several examples can clarify my previous statements, which may look too harsh.

1. Empirical violation of the Generalized Axiom of Revealed Preferences, GARP. The axiom postulates some logical features of preferences between bundles of goods. For example, if one prefers a bundle with given price p and quantity x, he/she would not also prefer another bundle with a worse combination of p and x. Offered for the first time in weaker form by P. Samuelson in 1938 and formalized later by many, this axiom plays a key role in economics theory and is (in a bit different form of Strong ARP) “a necessary and sufficient condition for data to be consistent with utility maximization” (Varian, 2006, p.106). As H. Varian concluded in his review of a problem, “We anticipate that in the future, revealed preference analysis will make a significant contribution to empirical economics as well.” (p.116) .

And it did, but in its own way: series of empirical experimental studies showed that, depending on different settings, GARP was violated by 10-75% of people, usually more than 50% (List and Millimet 2008). Instead of being treated as an axiom, it should be treated as an empirical fact with a certain frequency. What theory should be built around it?

2. Effect of zero price. Experiments show that when a product becomes free, its preference irrationally increases compared to the situation with the negligibly low price; the difference between 1 cent and 0 cents creates a huge gap in preferences (Ariely 2010). It shows that the price functions are not smooth, which doesn’t fit standard theory.

3. Convincing evidence of the inefficiency of investments in mutual funds, yet continuing practice of investing there (slides 25-27). 1. What is wrong with the current status of social sciences? B2. Problems specific for separate social sciences - economics Another numerical illustration of discrepancy between economics theory and observed facts out of many reported in literature. “One of the most important findings in economics is the Law of Demand: consumers demand more of a good the lower its price, holding constant tastes, the prices of other goods, and other factors that influence the amount they consume (Perloff ,2008, p.34). In one study, I calculated correlations between the two key variables – number of customers and price – for a chain of 949 restaurants all over America. Each correlation was calculated for data collected for about 3 years, or 154 weeks. Histograms of distributions of these coefficients are shown below. One can see that, regardless of the way of calculating (straight, with lag or with lead), about half of the correlations have positive sign, contrary to “Law of Demand”. It leaves a lot of room for speculations why it happens and why theory does not fit the data. I just stress the fact that the observational problem exists even within this seemingly absolutely basic economics law.

Correlations "Price Index - Number of Correlations "Price Index with lag 1 - Correlations Price "Index - Number of customers", 45% positives Number of customers", 54% positives customers with lag 1", 40% positives

160 160 160 140 140 140 120 120 120 100 100 100 80 80 80 60 60 60 40 40 40 20 20 20 0 0 0 -0.68475 -0.278845 0.22853625 0.7359175 -0.66887 -0.268925 0.13102 0.530965 0.93091 -0.6422 -0.25656 0.03267 0.41831 0.80395

Dinamic correlation for104 weeks of operation Dinamic correlation for 104 weeks of operation Dinamic correlation for 104 weeks of operation 1. What is wrong with the current status of social sciences? B2. Problems specific for separate social sciences - economics

Another good example of economics-political discrepancy between theory and practice could be derived from the chart below (copied from Zeihan, 2010). It shows that the politically and economically grounded idea of the European Union creation after its establishment in 2000 started to give expected results, which, in particular, produced the whole unification of bonds yields (reflecting confidence level in a country’s economy). However, when the crisis happened in 2008, the real big difference in economies started to take off, and the same pattern as it was before emerged, especially for Greece, where inefficiency was and is permanent, and for Ireland becoming dependent on financial industry, which it hadn’t had before. So, the theory turned out to be working properly only under some conditions and unexpectedly failed under others. That means that the theory was not good enough.

So much could be said about similar problems in sociology, psychology, political science, etc.– but I stop here. 2. How does statistics contribute to social knowledge integration? Statistics is a dominating tool routinely used in any sociological, econometrics, or psychological study, but its implementation encounters many more uncertainties than is usually assumed or described in literature. A chart from (Mandel, 2008) explains it in some detail. Business (domain) uncertainties are related to unclear goals, interests, data, and domain expertise: the goals of a study are often not precisely formulated; business interests make studies biased, sometimes in a severe way; domain expertise is usually vague and cannot help much in formulating of the statistical problem; data collected very often reflect what is easily available, but not what is really needed. S Statistical uncertainties include all business ones transposed into Statistical uncertainties Goals statistical realm – goals, interests, and domain expertise of statisticians (as Data opposite to their clients); data-related Interests issues from statistical point of view. But two of the three types of uncertainties are just statistical. Fundamental ones represent the disagreement of specialists on the Built True model principal aspects of statistical inquiry; model technical ones are about algorithmic Domain implementation of statistical methods, expertise and probabilistic ones are about relations between “empirical observations and Fundamental Business general parameters”. One of the biggest Uncertainties drawbacks of statistics is that only Technical Probabilistic probabilistic uncertainties are the main concern of this science, while they represent typically only a small portion of the real uncertainty in any concrete situation. 2. How does statistics contribute into social knowledge integration?

There is a vast amount of literature about fundamental and technical statistical problems (Berk, 2003; McCloskey and Ziliak, 2008; Moore, 2001; and others), which I tried to summarize and extend in (Mandel, 2008;2009). I see the main lines of fundamental disagreements between specialists and schools and respectively deep Epistemological discrepancies as follows:

1. Statistics as a way of estimating unknown parameters – statistics as data analysis

2. Objectivist (usually frequentist) – subjectivist (usually Bayesian) interpretations

3. Descriptive (inferential, learning) analysis– causal analysis

4. Axiomatic probabilistic (A. Kolmogorov) – collectivistic probabilistic (R. von Mises) approaches to foundational principles

Fundamental problems statisticians face regardless of chosen paradigm:

1. Mathematical requirements for data generating mechanisms and actual data generating mechanisms

2. Individual nature of any single observation and mass character of statistical estimates

3. Requirement to work in homogeneous situations and impossibility to determine ones

Inclination of a researcher to one or several of these lines of thinking implies his/her further following some others – but very often they are mixed in one study.

Each of these lines of disagreement is a huge topic which I cannot discuss in depth. Instead, I just give some comments which highlight certain respective problems without any attempt of systematic description. 2. How does statistics contribute to social knowledge integration?

1. Problem of causality. In terms of applicability, any models which do not guarantee causality are rather useless. If a statistician, say, makes a regression equation but cannot really guarantee that found coefficients represent some kind of causal relationship, the client will not even understand why the model was done. For that reason, all industry models look and are interpreted like causal ones – yet, of course, they are not. A special discipline has been developed (Pearl, 2009; Rubin, 2006) for that purposes, but it has some serious drawbacks.

A. One of the main ideas there is a concept of potential outcomes, PO (J. Neyman, 1923; D. Rubin, 1973). PO is a value of a variable which would appear if the “treatment variable” (effect of which is studied) were different. For example, if a patient took a pill and his temperature was 98 degrees, what would his temperature be, if he did not take a pill? The respective techniques received an intensive development in the last decade. But its very logic is very often flawed, since “what if “ question (reasonably denied in history, with which statistics is tightly linked) may be just not applicable (like “if he were a woman, then… everything else would change”).

B. The main algorithm to estimate effects in PO approach is “propensity scores (PS) matching”, where PS are obtained as a solution of logistic regression of a treatment variable against all other variables. Each treated patient is assigned (using different techniques, like nearest neighbor) to non-treated groups by PS and their average value of variable of interest is compared with that of treated patients to estimate a causal difference between the two. It creates, however, a paradoxical situation. If the goodness of fit of this logistic regression is very high (i.e. regression distinguishes treatment from non-treatment group very well), then a treated person has zero probability to belong to a non-treated group, and any assignment will be in fact random. But if the goodness of fit is zero, each treated person can belong to a non-treated group with the same high probability, i.e. assignment will be random again. But if so, why and where in the middle between these extremes does PSM work? I didn’t see discussion of this simple argument in special literature (but many other problems with PSM have been detected).

C. This approach cannot be applied to non-categorical variables (for which other technique of Structural Equation Modeling, SEM, is used). Recently it’s been shown, however, that PO could be considered as a particular case of SEM (Pearl, 2009), (big achievement; i.e. we are talking about two different but compatible languages). But SEM, as a type of regression has its own internal problems as well (slide 21).

D. Causal modeling of any type considers just most simple relations between variables, which are far from real causal relations.

The bottom line: non-causal methods are just surrogates; the causal ones are very far from being relevant. That’s the situation for now. L. Zadeh expresses deep pessimism about the very possibility of causal modeling (Zadeh, 2003) 2. How does statistics contribute to social knowledge integration? 2. Regression problem. By far the most popular in social sciences method of linear regression has internal inconsistencies. In series of experiments (Mandel,2008; calculations are done by V. Kamensky) I used generated data set with 10 variables and known coefficients. Coefficients were than estimated by regression for all individual variables, their pairs, triples, etc.; the averaged values are on the graph.

X1 means that the coefficient was 1, etc. The vertical axis shows the average percent deviation of the estimated coefficient from the real value; the horizontal axis – the number of predictors used. For example, the value 1.35 for variable X5 against 3 shows that the average estimation of the coefficient for X5 out of all triples possible was higher than the real value of 5 by 135%.

In general, the experiment shows that for the data with and without multicollinearity: a) the estimated values may be far away from the real values of the coefficients; b) the estimated values may diverge from the real values when the number of variables used is approaching the actual number of predictors originated in the outcome (some times divergence is much stronger) These results indicate that even if a mechanism of data generation is exactly as it is supposed to be in a regression model, it cannot always identify the real structure of the data generation even in simple cases. This fact alone undermines the blind use of regression as a universal tool. But there are not alternatives either. 2. How does statistics contributes to social knowledge integration? 3. Problem of Bayesian statistics – changing a personal belief with new facts coming. A real life story (http://www.russia-ic.com/news/show/8539/; http://www.ej.ru/?a=note&id=10599; ttp://amberbridge.org:8008/newstext?id=2132&lang=eng) 1. A teacher of Russian in the German embassy school, Benjamin Hobert (BH) drove his car with the speed of 66mph in a zone with a 36mph speed limit in Moscow and hit two students on the crosswalk on Nov. 30, 2008. One guy died immediately, another was dragged by the car for about 60 yards and died afterward.

2. BH pretended he knew neither Russian nor English, refused to take the drug test and showed his diplomatic passport, after which he was released by the police. Two days later, he left Russia and never contacted anyone since then.

3. He had been stopped several times earlier for speeding and other traffic violations (always denying the drug test, using diplomatic immunity); the last police report was sent to the embassy one month before the accident.

4. In 2009, BH was sentenced in the FRG in closed court hearings, without witnesses, etc. for one year in prison (conditionally), 5,000 Euros fine, and a one month driver license suspension. No compensation to the victims’ parents had been paid, no reaction to Russia’s requests about extradition followed.

5. In Germany, for accidents resulting in death the punishment could be up to 10 years in jail.

Statistical formulation of the problem My prior belief H: Germany judges its citizens equally regardless of the country where crime was committed (=Germany has no double standards) P(H) = 0.8; P(~H) = 0.2. A question: how will my belief change when I find out about a super soft verdict? Assumption 1 (presumably made based on statistical data available): a probability that verdict for Analogous crime will be soft P(Soft|H)=0.1 Assumption 2: if hypothesis H is not right (i.e. Germany does have double standards) than P(Soft|~H) = 0.6 2. How does statistics contribute to social knowledge integration?

3. Problem of Bayesian statistics – changing a personal belief with new facts coming.

Bayes rule: P(H|Soft) = P(Soft|H)P(H)/{P(Soft|H)P(H)+P(Soft|~H)P(~H) = 0.4

According to that calculation, my prior belief in fairness became 50% less probable (08/0.4) due to this one observed case. Assuming that another case of that type comes to my attention, the respective probability will be 0.1, i.e. 8 times drop.

How did I get my prior belief in fairness 0.8? I may use two types of information: one “from general education”, hearing randomly that Germany is a free country, etc., and another – from statistical data. These data, logically, may contain a set of observations about this very fairness: if, for instance, I know, that in 80% of cases when similar crime was committed outside of Germany, the verdicts were the same as if it was committed in Germany,

I may make an estimate P(H)=0.8. But than BH case should have no more weight than any others comprising these 80%, and all Bayes procedure described becomes meaningless! To block this type of thinking, in Bayesian statistics it is assumed, that P(H) has no statistical ground, it came from “general education” only. Fine! But what if somewhere in this “education” still sits the same statistics – in indirect form – and affects P(H)?

Another problem is this: how may one reasonably distinguish between assumptions 1 and 2? The only way to make some estimates (if not make them again subjective, as with P(H), which sounds too much) is to measure some frequencies. But there is only one Germany, either fair or not!

There are intense discussions of these and other issues, culminating somehow in proposal of “hybrid statistical tests”, satisfying “everyone” (Berger 2003), but it just shows the problem is not to be easily resolved, it has indeed deep existential roots. The dualism in facts interpretation and other paradoxes do not let me take Bayesian approach wholeheartedly. 2. How does statistics contribute to social knowledge integration?

Statistics’ main doubtless success is the area of supervised learning, where causal aspect of the problem is not very important, but these results are not guaranteed to be useful for management (Mandel 2009). Statistics is a science about mass random processes built around some mathematical requirements concerning these processes. But social life goes in such a way that randomness does not always obey these requirements for several fundamental reasons: a) Randomness of two types: 1. uncertainty the outcome of rolling the dice, comparatively repeatable and sometimes reversible, which statistics successfully (within limits partly discussed above) deals with; 2. events at bifurcation points (Prigogine, 1997), non-repeatable and irreversible, where statistics is meaningless. Both are permanently present in social life; b) huge importance of non-random structures, like state or corporations laws or regulations, geographical and political boundaries and so on; c) principal heterogeneity of playing agents of any type: individuals or organizations making decisions incorporable by its effect to millions of other decisions; d) self-organizational and power based mechanisms in society, which create, on the one hand, stabilizing forces (like pseudo-equilibrium on markets), and, on the other hand, destabilizing events like financial and political crises, rebels, wars, and revolutions. All events of that type are beyond statistical realm but critically important for society; e) interdependence of social agents, especially those making key decisions, undermines basic statistical models holding independence and “one data generator – one data set” assumptions as key ones. f) Homogeneous data source, the main condition for reasonable frequentists conclusions, is rather an exception than a rule (data mining just proved it on rich material – see also slide 55)

Social science hasn’t found yet the reasonable approach combining these aspects together. Quantitative studies are overwhelmed with statistical models, which don’t take into account “non- random aspect”, while qualitative studies do not go beyond simple descriptions or comparisons. One of the tasks of Sociosystemics is to establish some language dealing with this controversial nature of social processes. 3. Main components of sociosystemics

The main question to be eventually answered by sociosystemic: How can one provide maximal available relevant information for making social decision in reasonable and friendly form?

One may say that science is already busy with this problem and there is nothing new in it, and this will be partly right. But this “part” would be fairly small, recalling disarray in social sciences and in supporting them Statistics, as I’ve tried to demonstrate. Sociosystemics is considered both a science of answering such a question and, finally, as a computer answering system SocioS based on this science. Below there are comments about the features of SocioS; slide 28 tells more about scientific components needed to make such a system.

1. “Maximum available relevant” assumes no less than gathering of what is really known about a given problem in social science as a whole (social journalism may be included). This problem is clearly very far from solution. It involves understanding of what the problem is, specific search and data aggregation with strong filtering, bridging millions of different sources, and so on.

2. “Social decision” may be of any kind, from investment recommendation to the choice of a wife. Respectively, the proposed science should know how to deal with absolutely different requests in a unified way, which is a huge problem in itself. It also means that the system should be decentralized, since it is impossible to imagine answering of any questions from one center.

3. “Reasonable” assumes that filtered information is transformed into “knowledge”, so to say, in an intelligible manner , where its correctness (trustworthiness) and meaningfulness are estimated.

4. “Decision“ assumes choice among the options and associated risks, costs and benefits. So, the system ideally should make these estimates, which means that it works not only with analysis, but also with forecasting.

5. “Friendly” means, that everybody can use the system in such a way that he/she receives just a motivated response at first, but then is able to follow the system’s logic up to some (maybe, bottom) level, i.e. the system should be transparent and clear in its structure – it is the only way to make people believe that it really works instead of making up results.

Now I may give a shorter definition: Sociosystemic is a science about creation of SocioS. 3. Main components of sociosystemics Let assume now, that software is ready and SocioS’ talks with a certain inquisitive user (say, statistics professor ). Professor suddenly found himself with $20,000 to invest for about 15-20 years; he joyfully inserted this information into SocioS and typed a sacramental American question:

0. I don’t want to put money into particular stocks. In which mutual funds should I invest?

Than a dialog follows (SocioS’ answers are yellow; all technical aspects, like interface, hypertext, etc. are, of course, ignored).

1.You should not invest in mutual funds; rather invest in common index (Trustworthiness=TW is 0.90) 2. Why? Some funds perform much better 3. In the long run, for 15-20 years, it is not true (TW 0.90) 4. How can I make sure you are right? 5. There are 340 studies and primary sources confirming it. Do you want a list conclusions from the most convincing ones? 6. OK, yes, but not too many, just for a quick reference 7. Mutual funds managers very rarely outperform the market (TW 91). First 3 out of 120 arguments. A. Out of 2073 Morningstar Mutual Funds operating for ate least 10 years 16.7% beat SP500; 10.6% beat the Value-Weighted NYSE http://www.som.yale.edu/Faculty/zc25/Investments/Lect-16.ppt (TW 0.98) B. “Looking at the empirical evidence from nearly half a century, it seems safe to conclude that very few portfolio managers display the characteristics of truly successful or unsuccessful activity when adjusting for costs and chance.” (Ekholm 2010, 3 ) (TW 0.94) C. “For fund investors the simulation results are disheartening. When α is estimated on net returns to investors, the cross-section of precision-adjusted α estimates, t(α), suggests that few active funds produce benchmark adjusted expected returns that cover their costs. Thus, if many managers have sufficient skill to cover costs, they are hidden by the mass of managers with insufficient skill. On a practical level, our results on long-term performance say that true α in net returns to investors is negative for most if not all active funds, including funds with strongly positive α estimates for their entire histories.” (Fama and French, 2009, 3) (TW 0.88) 3. Main components of sociosystemics

Example: SocioS’ answers to investment question 8. Why is the trustworthiness of the last argument lower than that of others? 9. Additional studies may be needed for clarification of authors’ conclusions 10. Are there some contradictory results? 11. Yes; three such studies are rebutted in the article, but it still decreases a TW. List other results? 12. No, just tell me what statistical methods they applied in their work 13. Regression, times series, bootstrap. Do you want more details? 14.Yes, can I look at how they applied bootstrap? 15. It is not a local issue; we would recommend you should see the whole article. Do you want it? 16. No. OK, it is strange, anyway. There should be very good managers on the market. Are managers stimulated enough to do their job? 17. Do you want to formulate another question? 18. Yes 19. Formulate a question in a clear manner, please. 20. OK. Are funds’ managers stimulated for outperforming the market? 21. I don’t have a direct answer to that question. Would you like an indirect conclusion? 22. Yes 23. Most likely they are not (TW 78) 24. Interesting. Give my just one argument for that. 25. Top managers are mainly paid in the form of options of their companies stocks (TW 87). First out of 10 arguments. 26. Could you show some data about it? 27. Here is a chart showing the ratio of CEO wages to the average workers salary (TW 88) (Hill and Myatt 2010, 192) First out of 5 sources. 3. Main components of sociosystemics

Example: SocioS’ answers to investment question 28. Is it specifically for mutual funds top managers? 29. No, it’s for all CEOs 30. OK, never mind. But why does it prove that managers are not interested to outperform the market? 31. I don’t understand your question. Repeat it in a clear manner, please. 32. I’m returning to your answer 23. Give me another argument there. 33. Managers are just not capable to outperform the market and are satisfied with the current incentives system which often guarantees a high pay anyway. TW 60 (SocioS argument) 34. A third argument? 35. “Errors of omission consist of not doing something that should have been done… a security-seeking manager's optimal strategy is to avoid such errors by doing as little as possible, including nothing” (Ackoff and Addison, 2006, 35). TW 57 36. OK, enough. The risk to invest into mutual fund is much higher than just to SP500 index. 37. Do you want to ask another question? 38. No, thanks. You saved me a day of study. 39. You are welcome; I saved people years and millions of dollars. 3. Main components of sociosystemics In order to support a dialog like that SocioS should have quite unique combination of different features, some substantial, and some more technical. In broad terms, I may say that Sociosystemics is proposed as a science about coherent reorganization of social knowledge into workable automatic or semi-automatic tool for solving goal oriented related problems.

The main tasks and stages of this science are:

A) Creation of a methodologically coherent and ideologically free minimalist theory of the main components of social reality, with as clear as possible distinction between objectively determined and measured facts, features and relations, and subjective opinions (theories) about them;

B) Creation of an appropriate language for describing this reality, which includes commonly accepted vocabulary, grammar, and traditional use of language;

C) Retrieving of the relevant pieces of information in a form of meaningful chain of arguments triggered by requests related to particular goals;

D) Classification of accumulated knowledge about reality into clearly determinable, separable, and interrelated pieces, some kind of “social systematic”, analogous (but not equal) to biological one, which by design overcomes disciplinary boundaries;

E) Estimation of the trustworthiness of the derived conclusions and recommendations based on propagation of specific scores through the networks.

F) Methodology for Self-Organized Agent Based Models (SOABM) for any particular problem, adjusting its structure and parameters based on the retrieved knowledge and its trustworthiness and the level of relevance.

The following is a brief description of possible ways to obtain these purposes. I will mainly focus on problem A) and touch others just in brief. 3. Main components of sociosystemics

Creation of a methodologically coherent and ideologically free picture of a social reality means development of a “minimalist social theory” which describes a “skeleton” of any society in such a way that it does not contradict any (or at least major) existing theories. It should ultimately cover the main relationships between all (major) agents in a society, which allows, after “feeding” proper parameters, imitating of different aspects of this society activity in a practically acceptable way. Such a huge and seemingly impossible task is indeed very complicated, but many building blocks are ready or could be made, if efforts are applied in that direction. Main ingredients of this “minimalist theory” could be as follows: 1. Society is comprised of individual and group agents. Each group (family, corporation, state, etc.) is a specifically organized set of individuals and does not possess any features irrelevant to the features of these individuals. “Group will” doesn’t exist per se, i.e. “principal – agent problem” is always persistent and needs to be taken into account.

2. Each agent pursues some set of goals, which are derivatives of individuals’ (not groups) desires. These primary desires should be analyzed and put into proper place to understand goals. The “homo economicus” myth should be denied; economical motives play very important but often not decisive role and intersect, sometimes unpredictably, with other motives, goals and desires (see positions of Durkheim, Polanyi, Bourdier and others in Smelser and R.Swedberg, p. 15-20). Constant failures to find a strong correlation between happiness and wealth (Christof, 2010) just confirm this.

3. Individuals have different strengths of these desires (a question learned not enough). The list of desires (I guess, incomplete – for instance, it doesn’t have such a basic thing as “desire for survival”) in a table on slide 30 gives an impression of how psychologists think about it. The fact that economics is interested mainly in “Saving”, and political sciences in “Power” does not change the real complexity of the list. If, say, desire for tranquility is dominant for a person or a group, no economic stimulus will prevail. Measuring strength of desires and their distribution among people in different societies is one of the most important task for sociosystemics, which is hardly started yet either in sociology or psychology.

4. Agents have different features and are involved in different relationships with other agents, including nonhuman ones (like mechanism, natural components, etc.) These relations are multifaceted and are subject to determination and classification. The language of “things, features and relations” (Uemov, 1963) is the most appropriate for “minimalist” description, for it doesn’t require any universal concepts of human nature, etc. which are usually questionable. 3. Main components of sociosystemics A. Minimalist picture of social reality List of basic desires (Reiss, 2004) combined with universal values (Schwartz, 1994) Intrinsic Universal Desire name Desire (Motive) feeling values Desire to improve society (including altruism, 1 Idealism justice) Compassion Universalism 2 Family Desire to raise own children Love Benevolence

3 Social contact Desire for peer companionship (desire to play) Fun Hedonism, benevolence Desire to influence (including leadership; 4 Power related to mastery) Efficacy Power, achievement Desire to get even (including desire to compete, 5 Vengeance to win) Vindication Conformity 6 Honor Desire to obey a traditional moral code Loyalty Tradition 7 Order Desire to organize (including desire for ritual) Stability Security 8 Independence Desire to be autonomous Freedom Self-directing 9 Tranquility Desire to avoid anxiety Safe, relaxed Excitement (opposite) 10 Curiosity Desire for knowledge Wonder 11 Physical exercise Desire to exercise muscles Vitality 12 Romance Desire for sex (including courting) Lust Desire for social standing (including desire for 13 Status attention) Self-importance Avoidance of 14 Eating Desire to eat hunger 15 Acceptance Desire for approval Self-confidence 16 Saving Desire to collect, value of frugality Ownership 3. Main components of sociosystemics A. Minimalist picture of social reality

5. Agents’ features include those that are genetically inherited, practically unchanged during a life time (Freese and Shostak, 2009; Yang et. al., 2005) and the ones that are acquired (changeable, flexible, dynamic). Some features, like intellectual ability, psychological type, sexual inclinations, and others might have mixed nature. The same applied to group agents, where inheritance is, generally, changeable (as far as the notion of a group is changeable), yet often could be quite transparent (like family traditions, principles of inheritance within a kingdom, etc.). Features affect persons’s and groups’ behavior in all aspects.

6. The measurement of some features might be a very hot topic. One of the most famous examples is publishing of “The Bell Curve” (Herrnstein and Murray, 1994), when mere generalization of hundreds of studies of IQ triggered a storm of liberal criticism. The authors’ only guilt though was that it was convincingly (in their opinion) demonstrated, that different races have different average IQ. Why is this so surprising, if they have many other traits different as well (Matsumoto, 1993; Downey, 2008)? But the whole thick book of opposing opinions (Jacoby and Glauberman, 1995) was issued just in one year. It contained many reasonable arguments but still did not enter into the core of the original arguments of their opponents. The huge meta-study thirteen years later (Lynn, 2008) not only confirmed the main findings of “The Bell Curve”, but brought new statistics from around the globe. I’m writing it here not to defend the theory, but just to stress out, how wrongly understood “political correctness” plays a destructive role in the search for truth. One of the major sociosystemic goals is to create a mechanism blocking such situations.

7. The number of features to be somehow measured on agents is enormous, or, maybe, infinite from practical point of view. The USA Census, say, collects at least a hundred of variables about each person; if you sum up all marketing, sociological and psychological studies, it will include thousands of variables, but all that richness is in disarray. One of the tasks for sociosystemics is to create as rich as possible picture of humanity, not in a sense of one gigantic data base (although it is also a possibility), but by aggregating in an intelligent manner of diverse information collected from thousands of studies. This is to answer approximately the following questions: if in one study in Arizona it is shown that children drink Pepsi twice as much as adults, will this ratio hold in Maine? In France? What if in another study in Mexico they drink exactly as adults? Special techniques of data integration are data fusion already used in statistics, but I’m talking about a more general problem. 3. Main components of sociosystemics

A. Minimalist picture of social reality 8. In order to achieve their goals agents enter into relationships with each other. Some examples of important relationships are shown in a table. I didn’t find a more or less People - people People - corporation exhaustive list of these (it may not exist), Kinship of different kinds Working in but it should be gradually created. Particular Friendship Working for but very important type of relationships Sexual Awareness is its absence, i.e. independence of agents. Love Economical Compassion Loyalty 9. Each of these relationships forms a network. Hate Consuming products of Social networks have long been a subject of Subordination of different study in sociology (Wasserman and Faust, types Interest

2006), but the approach received a big boost Peer relations Corporation - corporation just in the last decade (Newman, Barabási, Watts, 2006) Cooperation Selling products due to the introduction of the new concepts Suspicion Selling service (small world and scale free networks, in Envy Awareness addition to the traditional random graphs), Jealousy Interest but mainly because of availability of the Sharing opinions Cooperation huge datasets about emails, telephone calls, Business partnership Competition etc. However, I do not know of systematic Economical People - political party studies where networking of a person was Awareness Membership explored under the angle of several relationships Respect Moral support like these listed in a a table. Similarly to the Dependence Financial support features data problems, it waits its development, Interest Recruiting but it is a much harder task. Competition 3. Main components of sociosystemics A. Minimalist picture of social reality The idea of connectedness between different nets could be illustrated by the example of two very important networks and their relations.

1. Network of social knowledge (relations between concepts) 2. Network of people (awareness or friendships, reflected in the different thickness of the arrows)

It shows, on the one hand, that people may effectively disseminate different ideas between themselves (on which idea the WWW is working and all hope of the booming social marketing is based), but also that they may easily block or misinterpret the concepts, because different people capture, say, just one concept from the tight cluster and actively pass it further. I guess, the nature of the greatest misconceptions in human history, affecting millions of lives, is related to this type of ideas propagation from network 1 to network 2 with blocking and distortions on a way. One of the purposes of sociosystemics should be to elaborate on this kind of relationships - a job hardly started in modern science, as far as I know. An example - the phenomenon of secondary dependence: since some volume of knowledge (usually unknown) is always shared between agents, in what sense one may consider them independent in other aspects (a key statistical assumption)? 3. Main components of sociosystemics A. Minimalist picture of social reality 10. Set of relationships form a background for agents’ actions. One may say an action is a materialization of a goal in a multidimensional space of relationships (there are some purposeless or unconscious actions, which I will not consider here). Only through actions agents change the world, which should be a subject for social science. Action is the result of some decision, which goes according to the scheme. Here: GR (Goal Related) System is part of reality containing main forces affecting a goal’s achievement; GP (Goal Related Perceived) system – part of GRS limited by initial personal predispositions; GM (Goal Related Mental) system – image of GPS in the mind of decision maker, a distorted subset of GRS. Scheme of making individual decision If a person has no predispositions (paradigms, prejudices), then GPS=GRS, but it is very rare that GMS GRS. GRS is different by definition and design from the system in common sense of General Systems Theory, since it doesn’t include idea of integrity or commonality. But it is most relevant to what people need to know when they solve the problems – so, GRS, GPS, and GMS are the main subjects for sociosystemics. All decisions involve calculations of costs, benefits, and risks associated with selection of the given option or indecisiveness. 3. Main components of sociosystemics A. Minimalist picture of social reality 11. The key component of the graph on slide 34 is a concept of goal in its relation to reality where the goal-oriented action should be performed - thus intentional aspect is critical here. It represents a serious difference from influential in sociology Actor- Network Theory (ANT: Latour, 2005; Cerulo 2009 and others), where all agents - actors (both human and nonhuman – another common part) are considered as such if they occupy some place in network needed for “making thing done”. It includes components like “mental stage of actors”, their knowledge, and other immaterial things, together with all material but “passive” agents like mechanisms, etc. In other words, in ANT, “network” is practically equal to GRS, but merging “mental and physical” together does not allow to clearly differentiate between known and unknown, as done in “minimalist theory”. Distinction between everything beyond the action (predispositions and all decision- related mental processes) and actions per se seem much more constructive from many points of view.

12. To address the question of “integrity of network”, critical for ANT, one should separate two types of GRS: goal-critical and goal-desirable systems. In the 1st system, the absence of any component implies that the goal will not be achieved (like broken engine in a car would not allow driving, if the goal was to drive). The second assumes, that given component is desirable, but not critical for the goal (like broken window in a car does not prevent it from driving, but makes it less comfortable). To stress the point, in typical statistical models that distinction is almost never made – all features (components) are assumed just as desirable, while it is vital that some of them are critical. It relates to the problems of causality and its mechanism again – so, I stop here. 3. Main components of sociosystemics A. Minimalist picture of social reality It seems that described concepts represent a minimal sufficient set of characteristics needed for analysis of basic social categories. I don’t specify any particular things, which followed from the existing social or economic theories, I don’t speculate about the main forces driving society, and I don’t claim I know historic laws governing our lives (Snooks, 1998). I just list the elements, the interplay between which represents particular social situation:

1. Individual and group agents, human and nonhuman.

2. Human basic desires (motives) having different strength for different agents

3. Agents goals predetermined by agents desires

4. Constraints for goal achievement represented by other agents

5. Agents’ features of different types somehow correlated between themselves

6. Relationships of many types and different strengths between agents affected by their features and goals, which created networks for each type of relations

7. Desires, goals, constraints, features, relationships

8. Special type of connections between knowledge networks and human networks

9. Decision as a mental act working with distorted images of reality and aggregating cost, benefits, and risks 3. Main components of sociosystemics A. Minimalist picture of social reality 10. Interplay between real goal-oriented system and two mental systems-images, which are different only by level of predisposition of a person

11. Human agents’ actions as materialization of their goals in given relationships as consequence of decisions made - the only force in the social world what matters. The results of actions of other agents at the moment of the given action create conditions for this action

12. Actions of group agents always depend on actions of individual agents comprising them

13. All components, from desires to actions, could be partly or completely hidden from other agents

14. Pure randomness on each level above, to be distinguished from the effects of hidden actions.

The described picture allows to use different models of human nature – for example, any traits in (Jensen and Meckling, 1994), who consider REMM (Resourceful, Evaluative, Maximizing Model) as the best compared to economic, sociological, psychological, and political models of man, because it picks some key aspects of all others. In turn, I don’t focus on any optimization mechanism - it is just a non-common part of a picture. These concepts should be exploited and filled with meaning with the development of sociosystemics. Many of them are already under deep scientific scrutiny, but should be included in a general scheme.

A very important feature of the described minimalist picture is that it potentially allows to describe in unified terms all or almost all accumulated social knowledge. In many sections of economics theory practically the same terms are already in use: game or decision theories explore featureless (!) agents with interests, goals, payoffs, relations (conflict, cooperation), risks, and so on. Other branches do not use this terminology, but could easily make necessary adaptation. Say, in psychology agents are usually individuals, with many complicated features (including abnormal); relationships between them may have multiple forms; predispositions play very important role and so on. In sociology I mentioned ANT; in political science dominate very complex agents with conflicting goals; history is full of agents with colorful features, etc. 3. Main components of sociosystemics

This reduction allows to solve (or greatly facilitate solving) three fundamental problems:

1. Unify a language space of social sciences for further treatment, making interdisciplinary bridges much stronger and longer in any direction (if the same agent or the same features are meet in several sciences, they may be picked up based on that) – problems B and D on slide 28.

2. Systemize different social topics not by their formal similarity (slides 42-43), but by their relational identity, which is much more important. This systematization, in turn, would initiate another use of statistics (slides 38-41) – problem E.

3. Create a common ground for parameterization and calibration of Self Organized Agent Base Models, allowing them to collect information from different sources automatically or semi- automatically – problem F. As a prelude to solving this very complicated problem, the minimalist background would allow to make just “normal ABM” within this framework in a systematic manner. Agents relations as opposite to features correlations 1. One of the key features of minimalist theory is its emphasis on the importance of relations between agents as opposite to correlations between features, which is so predominant in statistics. One may say that, as the bottom line, all meaningful features correlations are to be causally explained by agent’s relations (saving for strong random interventions) – but not directly. Say, observed positive correlation between education and income (slide 39) is the result of many interactions between people having different educational level with their employers, etc. Under different conditions (thus different relations between agents) this correlation may, of course, change its sign drastically – in Cambodia under Red Khmers in 1975-79 educated people (if survived) had much less pay (in the form of food) than their illiterate young guards. Systematic analysis of things like that will ultimately reveal still mysterious “correlations” and decompose them into meaningful patterns. Search for these patterns should be in fact the main purpose of social science – but in current situation, in absence of clear classification of agents, relationships, etc. it doesn’t work this way. 3. Main components of sociosystemics A. Minimalist picture of social reality 2. In general, the interplay between “relations between agents – correlation between agents’ features” is completely understudied due to its complexity. But it’s time to face it. One may say, that overwhelming obsession with agents features, which remains the main premise in statistics (the whole multidimensional analysis, data mining, and so on), does not stimulate progress in this direction, and only recently developed ABM approach can and should change a paradigm. However, within ABM community another extreme flourishes: when calibration of AB model happens, it is usually concerned with some key variables (target ones) rather than with the whole observed picture, correlations between features including.

3. The general logic of relations between these two aspects of reality is approximately like that. Let’s say, we observed some correlation between two variables, like on a graph – it’s a more or less

Correlation betw een education and income, r=0.63 expected positive correlation between education and income. What does it really mean, why does it appear this 120000 way? After receiving some level of education, a person goes to the labor market and looking for a relevant 100000 position. At this time, several mechanisms may work: 80000 1. Typical mechanisms (free labor market) 60000 2. Extra-promotion mechanisms (a need in specialists in 40000

graduation,$1,000 areas where graduated professionals will not go, etc.),

20000 promoting uneducated people to highly paid jobs Income per year after 5 years of years after5 Incomeyear per 3. Blocking mechanisms (health or family reasons; decline 0 in demand for a given profession, etc.), preventing 0 1 2 3 4 5 6 Education (1- low est) educated people from entering a labor market

The correlation scatter plot then aggregates all these things in one picture, leaving no opportunity for statistical methods to distinguish between these different mechanisms. 3. Main components of sociosystemics A. Minimalist picture of social reality 4. A possible objection to this last statement is that one should add new variables like “blocking factors”, etc., measure them and then run a more complicated model (what statisticians usually do). But it doesn’t work for several reasons.

“Education” in this example is not a target variable to be explained by adding new variables (let assume, the “Income” is). If one would do it for each independent variable, the model would immediately become just unaccountable. b. Even if it were – it is not what is really needed. One cannot replace an entire mechanism by one or several variables, since it also includes its internal structure. It is like one cannot just say that a car is to be described by a set of variables like “engine”, “ transmission”, etc. without knowing the exact way in which they are linked, i.e. relations. c. Agents playing a role in these mechanisms are not only those in the original setting, i.e. not just individuals on a scatter plot, but job agencies, employers, labor departments, and so on , and each can contribute into the location of each dot on a graph. d. It shows that the key problem is not to group individuals by their similarity to each other on a graph or in multidimensional space (what cluster analysis or data mining do), but by mechanisms which brought them into these positions, like shown on the graph. And it is here where the real causal problem of statistics is lying - not in SEM or other regression based procedures.

5. Only parameters of the mechanisms as causal drivers make sense for the real management of the process, not parameters of regression, which usually have no sense, except for some situations. If one knows that increase of education level on one level yields on average increase in income in $6,300, it doesn’t mean that “forcing people” to increase their education will indeed produce such an increase in income (see other reasons in (Mandel 2009) 3. Main components of sociosystemics A. Minimalist picture of social reality The reason is that regression coefficient mixes up consequences of work of several mechanisms. The correct solution would be to have coefficients for each mechanism separately plus to have information about comparative size of these mechanisms in a society. Then management may operate either with coefficients, or with proportions (which may also be manageable) and get the sought results. But traditional statistics cannot go there. The term “homogeneous“, very popular in statistics (yet very often misused even in this meaning) and describing similar values of the variables on objects, does not penetrate the core of the problem, which is identity of forming mechanisms, not their outcomes. This term should be replaced by “having the same cause” or “generated by the same mechanism”. The whole pathos of predictive modeling in data mining leads to missing of this key distinction.

Empirical data aggregation and systematization

1. Detection of causes of social events is the main purpose of social sciences which are not to be masked by any other considerations. Whatever methodology is used for this, it needs empirical support, both as a basis for causes derivation and as a criteria for testing the findings. There are several types of empirical data, each of which has its specific problems. A short summary is presented in a table on slide 42. I can’t discuss all issues there (a lot of literature exists about it), but stop to talk only about the two outlined cells.

2. A reporting problem in 2.1 is about the following situation: the usual report of an experimental study tells us, that some effect was observed “in average”, just to prove the main point. But one of the most interesting questions is the exact proportion of people who do not support the main effect. For example, D. Ariely reports that when respondents are in arousal – they change their opinions about some subjects (in this case – related with sexual behavior) significantly (Ariely 2010, 136- 138). He does it by showing average scores in normal and arousal conditions and their difference. But what is important is how many people didn’t change their scores, i.e. what is a share of psychologically stable or even rigid people (while it was, I guess, measured, but not published). As a result, many experimental studies lose their applicability just because of this misreporting. If reported (and measured) systematically in a correct way, experimental results would represent a unique picture of different human conditions, both on a level of features frequencies, their correlations among themselves, and relations between agents – i.e. what is ultimately required for sociosystemics development on its initial stage. 3. Main components of sociosystemics A. Minimalist picture of social reality Types of empirical data and their use in causal analysis Advantages for causal Problems from causal analysis Types of data Examples analysis view point

Actions are always more A What people really do important than intentions Government statistics, Most reliable and rich Data represent the outcome of many 1 In real life observational studies data causes, which is very hard to untangle In experimental setting Never guarantees universality; should (artificial situation; isolating The closest way to be closer to material science than 2 factors) experimental physics physics Participants know it is an Actions under researchers' Many reporting and other problems, experiment, but usually supervision (randomized Often reveals causes when including a need in causal procedures 2.1 don't know its purpose trials, laboratory settings) organized correctly to randomization Actions in artificially created but naturally Participants do not know it looking conditions (taking A best approach for 2.2 is an experiment cookies for free in a lobby) particular situation The participants are not controlled What people do in extreme Natural experiment (a situations, like after real life under specific hurricane Katrina, or in Possibility for mass circumstances, allowing to absence of woman (Roehner actions, not obtainable in Usually these are rare events, not 3 isolate some factors) 2007), etc. other experiments covering wide spectrum of social life Simplicity, low price, Saying is different from doing; B What people say they do Polls, questionnaires possibility for repetition conditions of poll and life are different Direct measurement of neuro reaction to Very limited experience; expensive; What people really stimulus, "pure truth" in universality is unclear; thinking is not C think Neuromarketing "answers" equal doing 3. Main components of sociosystemics A. Minimalist picture of social reality 3. The key problem of experimental results is the lack of clarity in answering the following question: how universal they are, could they be generalized for the wider population (or other agents community) or not. If, for example, in a study of attribution error the participants who ask questions are perceived much more knowledgeable than those who answer, although the roles were assigned randomly (Myers 2005, 90), then – does it mean that this effect holds for all situations and for all groups of population (there is a popular joke that the most studied group in the world is college students – they provide respondents for vast majority of studies due to availability at any time and for free)? If not, then where are the boundaries of the finding? If yes, then where is evidence? These questions did not find a convincing answer. The same, of course, is correct for data like B and C – if, say, one study discovers that 60% of all Americans believe that they may reunite with their pets in heaven (Cerulo 2010, 541), should I assume that in an average crowd of 100 people 60 would believe it, saving for random error? Or does it apply to a particular group only? The advantage of different data of type B is that researchers try to represent certain population as good as they can, but usually contrast between available number of respondents and huge variety in demographic profiles makes trades offs hard to achieve.

4. Another problem with experimental data is that in social sciences it is a strongly underdeveloped approach. It is not because it is hard to make experiment and so on (the usual arguments), but because it is still focused on one- or two-dimensional “feature-feature” situations and very rarely (as far as I know) on something different. If physical experiments, either lead by theory, or triggering a new theory, finally produce results which are to be expressed in the form of exact relationships between agents involved (literally in formulas), in social life it never happened and will never happen for the reasons explained earlier. That’s why these experiments should rather estimate the boundaries between zones where different relationships exist (as a function of changing conditions or particular features) than be relevant to some presumed formulas. In that sense, social sciences are much closer to, say, material sciences. As far as pure metal becomes “not pure” – exact physical formulas really disappear and yield place to empirical findings about boundaries between different zones where different crystal structure (i.e. relationships) exist. It is illustrated in a chart. 3. Main components of sociosystemics A. Minimalist picture of social reality Iron-carbon constitutional diagram Source: http://www.davistownmuseum.org/PDFs/Pub42_Glossary_ Appendix1_Iron%20Carbon%20Diagrams.pdf A grah shows which transformations experience pure iron when two parameters – content of carbon and temperature – are changing. As one can see, about twenty different structures appear, which received special names due to their importance for metallurgy. Creation of this scheme took decades of experiments. Each zone has different crystal structure, i.e. different type of relations between agents (atoms).

Statistically, it is like a simple “scatter plot”, but in fact it is something which I do not remember seeing in social literature (phase transitions may appear in theory see concept of mediaphysics Kuznetsov and Mandel 2007, Mandel and Kuznetcov 2009). Did someone try to change people’s conditions carefully enough to observe similar patterns? Who studied, for example, how friendship between people is changing when income and age are growing? Or how are relations between troops and officers during the war changing when duration and winning attitude are changing? How these diagrams are changed for different other “additions”. (like this one with addition of zinc)? These and many similar questions await systematic studies. 3. Main components of sociosystemics B-D. Language for description of social reality Language problem in its entirety may be the most important and complicated problem in social sciences for it critically touches everything else; I would just touch upon just some most relevant to sociosystemics aspects of it. 1. The vagueness of terminology used in social sciences is notorious and I would not discuss it at length here. It may be one of the main reasons why P. Johnson called them “scarcely sciences at all”. Deceptive and catching words like “commonwealth”, “national pride”, “people’s democracy”, etc. have been used to cover horrible crimes. On a lesser scale, highly ambiguous terms are constantly used in scientific texts, which makes their understanding and especially comparison impossible, while it often hides the trivial concepts beneath (a postmodernist discourse, such as Derrida’s, gives especially bright examples of that). It represents a sharp contrast to other sciences, where terminology is much more strongly established. Some might praise vagueness (van Deempter, 2010), but for social sciences it is a real disaster, as I’ve always felt (Mandel, 1975).

2. There are several deep reasons for that, such as natural ambiguity of common language; historical tradition of the free unconstrained discussion; magnitudes of styles and schools; tight links between social science and daily social life and politics (where terminology is especially ambiguous); the benefits ambiguity often has, especially for forecasting, etc. But one reason is of a special importance: from the times of Plato, social texts are deeply related to other “humanities”, such as art, literature, religion, moral teaching, and so on. Vagueness may almost disappear in super-mathematical theoretical economics, and it may blossom in literature or art criticism, but it never goes away. Ambiguity, hyperboles, metaphors, fuzziness, and multiple meaning are very welcome in these types of activities, but it inevitably crosses the boundary into social science language, giving its authors the feeling of “emotional self-actualization” (slide 9). It seems clear that the problem of the creation of strong, algorithmic-like scientific language, as dreamed in the 30s by logical positivists, will never be solved.

3. But, on the other hand, the commonly accepted language is an absolutely necessary component of social science integration. Without it, a vast majority of studies would remain unsearchable, unclassified, and thus unused, which is exactly true in the current situation. From the point of view of sociosystemics, common language is needed for solving two fundamental problems: A. Classification of social knowledge, the only condition unde which it becomes workable. It includes describing of similar situations, where supposedly the same mechanisms and statistical regularities take place -one of the key conditions for statistical algorithms to work and for analogy to be useful (slide 28-c; 3.A) B. Tracing chains of arguments to support the validity of the final conclusion (slide 28-d,e; 3.C) What could be done for this language unification? 3. Main components of sociosystemics

B-D. Language for description of social reality 4. There are two principal ways of solving the language unification problem: one is “from people to language”: finding some language consensus among specialists (analogous to L. Zamenhof’s idea of Esperanto); and another - “from language to people”, which is creation of algorithms capable of making the current language workable for modeling needs, like Natural Language Processing (NLP) theory and applications. From people to language 5. Scientific Subject Classification. The simplest (relatively) thing to do in the first direction is to classify different scientific topics (subjects) in a logical hierarchical structure. Such classifications already exist in mathematics (MCS), in physic, astronomy (PACS - Physics and Astronomy Classification Scheme), and computer science (ACM Classification), but not in social sciences in general. As far as I know, such a system is developed only for economics - Journal of Economic Literature (JEL) Classification System with about 800 topics in it (compare with 9200 in PASC!)- but not for sociology, psychology, and so on. The traditional system of key words accepted in social science only multiplies the intrinsic vagueness of the parlance. I’m pretty sure, that if respective associations, INFORMS including, improved such a practice and come up with agreeable social sciences subjects classification, it would be a big step ahead. Then it would be a matter of editors enforcing the usage of the accepted classification by the authors, for authors have to assign their work to certain groups, as they do now in mathematics and physics.

6. Subject Classification is necessary but far from sufficient step at least for three reasons: a) Authors’ arbitrariness when they classify their work (a universal problem); b) Originally vague terminological base from which classification of subjects can be ever made in principle (a problem specific for social sciences). In that sense classification should differ from JEL in one very important aspect – it should count for big number of synonyms and synonymous constrictions which will direct statements to the same group; c) Even if classification is perfect and not ambiguous as MCS, it does classify just topics, but not concepts within the papers or books. This is something to be dealt differently. 3. Main components of sociosystemics B-D. Language for description of social reality 7. Understanding and description of the concept of scientific material is often a big problem both for authors and for readers. It is not clear from an abstract, it could be not clear from the main text as well. It could be a great step ahead if some uniform way of summarizing the results of a study by authors were adapted by scientific community instead or as a supplement to the usual practice of abstracting. It could include the following elements of Content Protocol (far from complete list): a) Statements which the author considers as established in her work (goals obtained), in a formal way as relationship between agents or their features b) Evidence she presented for establishing each statement, separating conventional (referral); logical (conversational); theoretical (deductive, mathematical); statistical (empirical, numerical, observational); experimental; and computational (imitational) types of evidence; c) References to data sources used, if any, for each evidential argument; d) References to other studies conforming or conflicting with each argument.

8. An example of simple Content Protocol is given on slide 48. It demonstrates just some possible options and could be advanced in many directions; it could be done not only by authors themselves, but, in the future, automatically. The whole idea is: it should describe an essence of elementary “pieces of knowledge” (or memes, so to speak) obtained in the given material, using a language, maximally close to the one of minimalist theory (roughly described on slide 28). For doing that in a uniform way several things should be performed (besides the main one – convincing authors to follow the recommendations): a) Classification of names of relationships where all possible synonyms are included (like cause, dominance, conflict, cooperation, war, love, and so on) b) Standardization of logical structures of statements and evidence representations c) Unification of references (like to refer to the book/article as a whole, or to the paragraphs, or to the statements in the source Shifting social texts representation into this formalized direction will facilitate the development of any unifying tools tremendously. 3. Main components of sociosystemics B-D. Language for description of social reality Content Protocol (based on abstract only; outlined are terms used in definitions)

Abstract (Jacobsson 2009, 467): This paper demonstrates how the analysis can differ dramatically between two common static modeling approaches to conflict. The first approach uses a one- period setup and associates positive arms investments with conflict. The second approach has two periods, where arming decisions are taken in the first period, and the decision on whether to go to war is taken separately in the second. Building on Features of the latter approach, I introduce a repeated game protocol with myopic players. agents/ Under these circumstances countries may end up in cycles of war and peace. This Types of Relations Relations result offers a novel explanation for a common pattern in history. evidence type name Countries may end up in cycles of war and peace Statement 1 depending on county leaders decisions Agents War; piece; Country - cyclical Countries Not described country alternation Countries leaders Myopia IF arming decisions are taken by country leaders in period of arm investment AND decision on whether to go to war is taken by countries leaders Theoretical Evidence separately afterwards THAN statement is correct mathematical Reference to data source No Reference to confirming sources Yes (list)

Reference to opposing sources Yes (list) 3. Main components of sociosystemics B-D. Language for description of social reality

From language to people

9. This way of language unification means the whole huge area of computational linguistics, Natural Language Processing (NLP), and text mining, which might have been the most dynamic area of science in general for the last two decades. Progress there is evident, but complexity of the subject still feeds pessimism about its final results, or rather about the time when they will be achieved. The problem of translation, for example, which was considered as most adequate for computers back in the 1950s is not solved yet in a satisfactory form, in spite of great deal of enthusiasm. In sociosystemics context, the most important achievements are expected in the area of text summarization, i.e. condensed description of the main ideas of the content (analogous of Statements and Evidence in Content Protocol), but it is also very far from acceptable solution except, maybe, for some very specific situations. The development in the area of NLP, tightly related with Semantic Web, Ontology Network and other web initiative, changes the landscape every month, and I’m far from being a specialist in that area. I’d rather call myself an “interested user”. From that point of view, I will just make couple of comments which seem especially important as contributions into sociosystemics.

10. One is related to terminology unification and/or creation of the minimally required dictionary/ grammar to be used everywhere (not for human relations, of course, but for internal usage). On the one hand, there are intensive studies in semantic and emotional primes by A. Werzbicka and her colleagues started in 1970 (Werzbicka 1999) , which already resulted in lists of equally understandable in all languages minimal numbers of words to be used for any communicative purposes, but on the other hand, in the influential recent handbook for NLP (Clark, et. al. 2010) this direction is not even mentioned. It means that it hasn’t become part of computational linguistics, and I can only be surprised why, since in my eyes it looks very promising for the described earlier purposes. 3. Main components of sociosystemics B-D. Language for description of social reality

11. Concept of relations between agents is addressed in NLP on a most granular level – it tries to understand relations between initial elements of text within its syntax or (with less success) semantic structure. it allows to make decomposition of any sentence. But what is still unclear to me (at least I didn’t find it in a literature, software, or conversations with linguists – although I make a disclaimer about my poor literacy in the subject again) is how NLP handles many sets of chains of relations within a document or corpora. I expect NLP to answer several questions: a) If A is cause of B, B leads to C, C is the same as F, F has generated G (assuming that all these outlined relations have been detected by NLP algorithm as such), then can we conclude from this algorithm that A affects G? b) If in the previous example there are synonyms for all subjects A-G within the text and other sets of synonyms for subjects and relations outside the text, then how will it be treated in NLP procedures? The purpose is – if in some other texts (corpora) some synonym structure identical to the one found here already exists, it should be reflected and saved for further treatment. c) Is it possible to follow all chains like that in corpora and make an exhaustive summary which includes all conclusions at least of one type – related to causal relationships? d) How can all that relate to techniques of propagation of production (inferential) rules widely used in expert systems (Giarratano and Riley 2005)? It seems underdeveloped in NLP as well. e) Is it possible to generalize so many conclusions and select the most important ones based on a certain criteria? If yes, it could be the real solution of text summarization problem on the relational level adequate to sociosystemics purposes for making automatic Content Protocols.

One of the main problems to be solved within “sociosystemics’ language sub-branch”, as I see it, is the problem of finding the reasonable compromise between extreme granularity of NLP techniques and required generality of the Content Protocols or similar structures. 3. Main components of sociosystemics B-D. Language for description of social reality

12. The difference between “from people to language” and “from language to people” paradigms is that the first depends on the free will of myriads of researchers and different institutions (like associations, publishing houses, etc.), forcing authors to follow some standards (I haven’t seen much success in that direction), and the second depends on the extreme complexity of the natural language itself. An advantage of the first approach is its comparative simplicity and better clarity; a disadvantage is that it might take many years for it to become a reality, even if some actions start today. An advantage of the NLP is that it already has enormous material to work with (which doesn’t have a standardization approach); the problem (which may never be be completely solved) is an internal (and eternal) ambiguity of natural language. These two approaches may and should somehow emerge, yet just on some problems. From sociosystemics point of view, their main goal is to describe social reality in terms compatible with its minimalist theoretical picture, i.e. create conditions when language is a sufficient tool for systemizing knowledge, while theory provides descriptions of the real life situations using the same language. It would allow to trace human knowledge through the chains of events and their description simultaneously in such a way, that question in SocioS like “Is democracy the best form of government ?” triggers two chains – one in a realm of evidence (studies, statistics, data), and another – in a realm of thought (concepts, ideas, models), and they can be read in the same language of relations : Is democracy the best form of government? Conclusions (Statements)

Social life (empirical evidence) Standardized Relations Social thought

Freedom studies in USA Language processing Hegel Schopenhauer Berlin

Feminist studies UN data Wall Street database Spenser Mill

International democracy studies Models of voting Plato Aristotle 3. Main components of sociosystemics E. Estimation of statements Trustworthiness 1.The question about truth is one of the most fundamental and cannot be solved once and forever. The recent “Truth guide” (Blackburn 2005) in fact demonstrates that any truth concept is subject to philosophical debunking. However, troubles with strong logically perfect definitions do not mean that some approximation cannot be designed. The need in it for sociosysetmics is crucial: different statements in sociological literature are countless, and tracing lines of arguments some of which are just false does not make any sense because it undermines the whole idea of reasonable science. Most likely, the following hypothesis is correct: for each sociological statement some others might be found, both strongly supporting it and strongly denying it. In situation like that some criteria of trustworthiness are just a necessity.

2. The table on slide 53 shows the main concepts of truth in the way they have been formed over the centuries. They cover practically all possibilities and can give guidance in where to find the sought “trustworthiness indicator” as shown in the dialog on slides 25-27. The most natural for practicing scientists correspondence theories assume some kind of relevance between any statements and observed facts (note that just half of the philosophers share that vision –slide 11!). The main cause of sociosystemics is, of course, also in that realm (slide 52 and others) but it doesn’t mean other aspects are to be neglected. The coherence and negative pragmatic theories bear huge heuristic potential and cannot be ignored . Consensus and social constructivism theories, in my opinion, are not worth to follow, but since “truth” generated in such a way (i.e. forced by consensus or political power) exists and may prevail in some areas of social thought, it should be detected and counter-tested by other approaches. Positive pragmatic theory, also very popular in science and especially in (bad) managerial practices , has limited value – something that worked well is not guaranteed to work well in the future. But, again, since empirically one has to collect any supportive data only from the past, one cannot ignore accumulated positive experience as well.

It shows that all principles should be somehow used in a united way and support each other in particular situations. How can this be done? 3. Main components of sociosystemics E. Estimation of statements trustworthiness 3. For clear reasons axiomatic theory of social life (and respectively acceptable ways to test correctness of theorems – putting aside internal troubles in mathematics like Gedel’s theorem) is impossible. However, some set of undisputed (universal) facts about social objects, applicable to all of them, could be still formulated, but yet it will be very poor in content, like “all people need food and air for living”, etc. All other facts of any nature would be by definition partial, applicable for some agents and not for others. It creates a necessary and sufficient requirement for saying that all socially important statements are conditional (i.e. there is no absolute truth).

Prominent Truth theory Content proponents Correspondence True beliefs and true statements correspond to the actual state of Plato, Aristotle, 1 theories affairs common sense Truth it is primarily a property of whole systems, and can be Spinoza, Leibniz, Coherence ascribed to individual propositions only according to their coherence Hegel, some logical 2 theories with the whole. positivists Social Truth is constructed by social processes, is historically and culturally 3 constructivism specific, it is in part shaped through the power struggles Vico, Hegel, Marx Truth is whatever is agreed upon, or might come to be agreed upon, by some specified group. Such a group might include all human Habermas, Rescher, 4 Consensus theory beings, or a subset of them. Kuhn Pragmatic theory Truth is verified and confirmed by the results of putting one's Peirce, James, 5 (positive) concepts into practice. Dewey Pragmatic theory What works may or may not be true, but what fails cannot be true Popper, Hocking, 6 (negative) because the truth always works Feynman, Deflationary Different theories link linguistic aspects of true propositions to other Frege, Tarsky, 7 theories statements (like "true is a redundant concept", etc.) Norwich, Ramsay 3. Main components of sociosystemics E. Estimation of statements trustworthiness 4. So, the next after universal facts are almost universal statements, such as “every baby has emotional attachment to the mother”, or “every baby older than 3 months follows strangers with his/her eyes “, etc. Many facts like that are presented in fundamental summary of human ethology (Eibl-Eibesfeldt 1989). Here “everybody” might be 99.99% of all human beings but still not all. Practically features like that play a background role in any social situation; it means that they are not significant statistically (since practically do not vary), but rich in content – if for some reasons some of these basic things didn’t materialize, there would usually be heavy consequences. In other words, it should always be in sight for behavior’s explanation (analogy can be drawn with psychoanalysis trying to discover the child’s trauma, but with obvious differences).

5. Starting with this level, all statements have statistical nature, in a sense that they cannot be attributed to all people but to some fractions of them. If we recall the table of desires (slide 30), even such a statement as “everybody wants to survive” will not be true due to suicides, desires to go to war, etc. Respectively, some frequency distribution should always be attached to each statement; provided that those distributions are one of the tasks of sociosystemics, as was already discussed for a number of occasions. Let’s assume, we indeed have a set of these distributions and are able to retrieve information about them when needed. Then any statements like “all Jews are rich” or “night time in New York is very dangerous” will be dismissed (or receive very low level of trustworthiness) automatically after comparison with these distributions. It alone would stop propagation of stereotypes and prejudices in social discourse and eliminate a major part of journalists’ rhetoric from consideration.

6. The better one knows the details of these distributions of features and relations, the higher the confidence in actions performed and in knowledge accumulated. I may suggest the following frequency-resistance principle: the higher the frequency of certain feature (relation) in a group of agents, the higher resistance, Ceteris Paribas, this group will demonstrate, if this feature (relation) is to be endangered; the opposite (an increase in approval, if a feature’s frequency is increasing ) is not guaranteed. The illustrations are: everybody will protest, if one tries to cut off access to the air; a smaller part of population would resist cutting off religion freedoms in a secular society than in a religious country, and so on. But extension of religious freedom in religious country would not be necessarily met with enthusiasm (similarly to law of diminishing return).

It means that, if frequencies of different features and relations are reliably established, it gives a powerful tool for estimating possible consequences of any actions, which is very often neglected. 3. Main components of sociosystemics E. Estimation of statements trustworthiness 7. Estimating frequencies in general is not a trivial thing. The biggest and almost unresolved problem is to which “general population” they are applied – lack of clarity in this generates many problems in statistics (slide 18). Here is an example. The US Census bureau annually makes a sample of about 3,000,000 respondents (so called PUMS study within American Community Survey), which is 1 % of the US population. I calculated a Pearson correlation between two variables: level of education of a head of household (measured in 22 points rank scale) and level of household’s value, $1,000. The correlation coefficient is 0.071, with confidence intervals 0.0693 and 0.0727 with 99% confidence. However, the minimal value of the coefficient within particular state is 0.017 and maximal is 0.160, i.e. they lie far from the boundaries of the American total. Keeping in mind huge sample size for each state, states’ intervals would not even overlap with those above. Should we assume that geography drastically affects such correlations? Should we still believe in correctness of the confidence intervals? Since blind trust in “statistical confidence” is a key feature of thousand of studies, this example might be an eyes opener for many researchers.

8. A special class of statements, where frequencies are not directly applied, but which have a very big if not universal meaning, is a set of different adages, or sayings, or rules, or aphorisms, or “laws” produced throughout millennia and often bearing much more sense than any particular studies, although they are not supported by statistics or modeling, but by wisdom and common sense. In business and management areas one may find Parkinson’s laws, Murphy’s laws, Ackoff’s (Ackoff and Addison 2006), or Augustine’s (Augustine 1997) laws, etc. I would assign rather high trustworthiness to these statements, but, of course, it should be specified much better.

9. Another very important case is the estimation of ordinary statements, which relate two or more “universal” notions, assuming “all” or “every” . “Happy families are all alike; every unhappy family is unhappy in its own way.“ - this L. Tolstoy’s sentence gives two examples at once. Taking “seriously” – this phrase is false, because not “all families are alike”, etc. But social discourse is full of such phrases. One may say, that frequency estimation is the only way to make statements like that scientifically valid, a way to transform metaphors into knowledge and by that to extract any solid sense in these if any. It is maybe where the fragile boundary between art and science is.

9. A classical way to test validity of theorems derived from axioms is to check correctness of all steps, each of which should follow the accepted (logical) rules. Modified to fuzzy social statements, this approach should also be applied, where trustworthiness of particular statement is considered as a function of trustworthiness of all other statements supporting this one and trustworthiness of the relations (deduction rules) which were used to come up to this statement on a way. 3. Main components of sociosystemics E. Estimation of statements trustworthiness The difficulty here, however, is not only that the original statements are fuzzy, but the derivation rules (relations) themselves could be quite ambiguous. Two most developed languages for solving similar problems – fuzzy logic and Bayesian networks – could be used for that purpose (with some reservations – slides 21, 22), but it is not all. From the point of view of coherence theory, if a given statement finds confirmation or contradicts (which is in fact a negative pragmatism) some other established (i.e. having high trustworthiness) piece of knowledge – the trust for this statement - should be corrected respectively. It creates a wide frame of reference for testing each statement, which allows to pursue the balanced way and avoid extremes by design, unless it is just correct.

10. To illustrate the role of the coherence checking, let’s consider the following statement: A. The US government knew in advance that attacks were planned on or around September 11, 2001, and it consciously failed to act.

This statement may find rather strong correspondence support, like: B. “There were intelligence reports about planning an attack, but they were neglected”; C. “Defense system around Pentagon did not work to prevent a plane’s attack, which is impossible without an internal command”

It also has huge consensus support: D. 49% of New York City residents in 2004, 36 % of Americans in 2006, believed in it; 61% of those surveyed by Gallup Poll in 9 Muslim countries in 2002 thought that Muslims had nothing to do with the attacks (Sunstein and Vermeule 2008, 2, 16). Opinions like that and many similar ones are usually (and somehow vaguely) considered as conspiracy theory and, as cited and as other studies show, it is extremely hard to fight them.

Is there a way to avoid assigning high trustworthiness scores to statements like A? One cannot easily disprove facts like B and C, especially because they are related to confidential information (although it is possible, of course). Facts like D cannot be disproved (!) due to their nature and don’t need to be (that’s why consensus approach has law credibility). But coherence approach should help a big deal. Let’s assume the statement is correct. It follows from it that many other statements should be false or have very low trustworthiness level, such as:

E. Government’s first priority is security for its people F. Top US officials are not criminals H. It is impossible to keep in secret for many years any facts about such a complicated plot 3. Main components of sociosystemics E. Estimation of statements trustworthiness

While it is sufficient for a normal person to disprove A, statements E-H and alike show that coherence logic is very hard to apply automatically – much harder than correspondence logic where lines of arguments may be followed from the chain of relations and synonyms. The main problem is that E-H statements may even be absent from the system (while they are embedded in human minds), since they assume some status of universality or just being. No one would ask any questions about government or top officials in such a way that statements E-H would have been triggered, it’s like to list “all features of the government” just in case (slides 29, 31). It poses a serious problem which sociosystemics should address and which I cannot discuss further here.

8. Another big issue is how to calculate some trustworthiness indicator (as shown in the dialog on slides 25-27) for each statement, both inserted into the system from outside and derived by the system, in the situation where different approaches give contradictory results, like an example in 7 above. Again, I will not go into detail here, but can just suggest that procedures similar to these used in estimating utility functions (aggregating different indicators in one by weighting) are not adequate. The global TW function, keeping all technical difficulties aside, is one where several contradictory things coincide:

 TW of a statement may have different status in different truth theories (as mentioned); TW should somehow depend on credibility of a source (like statement in “Nature” should weigh more than one in a local university journal), but there are many contrary examples TW should depend on the author’s reputation, but some not very credible person may still produce absolutely right statements TW should not depend on the society where the statement is made, but in fact it is treated completely different in different, especially opposing cultures. Maybe, future indicators of the TW should just ignore these and other social factors. 3. Main components of sociosystemics F. TRIZ and Self-Organized Agent Based Models (SOABM)

1. The whole discussion so far has been about knowledge description/classification and conclusions extraction/evaluation. Doing that, I inevitably had to go far away from the initially formulated problem, which was to give an answer to the question or to solve a particular goal-oriented problem. Now, let’s assume that all the preparatory work is done, i.e. SocioS has recognized the problem and aggregated all relevant to it information into referential groups of arguments. Then three scenarios are possible: a) To give some qualitative answer to the question analogous to the one described in the dialog on slides 27-27. For doing that arguments analysis is sufficient; b) To give a qualitative answer in a situation when non-trivial decision is needed and contradictions between different requirements take place. For that TRIZ seems the best methodology; c) To give a quantitative answer, if a problem needs that. For this ABM methodology is the best tool, followed by certain optimization techniques.

2. TRIZ is a Russian abbreviation in Latin letters, which became internationally recognized, for Theory of Solving Inventive Problems. It was developed by G. Altshuller during 1940s-90s (Altshuller 1996) and was successfully used by inventors all over the world. Many leading companies use it as a necessarily tool for technical innovations – say, Samsung has more than 100 TRIZ specialists (Souchkov 2007), etc. In the last decade, it also became adapted for business innovations (Mann and Domb 1999) and is used for that purpose with growing popularity (http://www.triz-journal.com/). But it still remains virtually unknown in social sciences, in spite of its great epistemological potential and proven very high efficiency. I will not describe TRIZ here, but just touch upon its potential place in sociosystemics structure. 3. Main components of sociosystemics F. TRIZ and Self-Organized Agent Based Models (SOABM) 3. The main idea of TRIZ is very dialectical and in that sense is as old as dialectics itself: solving any problem is in fact overcoming certain 1 Segmentation 21 Skipping contradictions. If one needs to "Blessing in Disguise" or increase power of a car it will 2 Taking out 22 "Turn Lemons into Lemonade" ultimately result in too high gas 3 Local quality 23 Feedback consumption, and so on. The great 4 Asymmetry 24 'Intermediary' discovery of G. Altshuller after studying of thousands of patents was 5 Merging 25 Self-service that inventors in fact use just forty 6 Universality 26 Copying principles for resolving all these 7 "Nested Doll" 27 Cheap Short-Living Objects contradictions. He was able to 8 Anti-Weight 28 Mechanics Substitution formulate them in general terms and 9 Preliminary Anti-Action 29 Pneumatics and Hydraulics aggregate in the so called TRIZ 40 10 Preliminary Action 30 Flexible Shells and Thin Films matrix. It shows how improving of some conditions results in worsening some 11 Beforehand Cushioning 31 Porous Materials others and which ones out of these 40 12 Equipotentiality 32 Color Changes principles are capable of overcoming 13 'The Other Way Round' 33 Homogeneity the given contradiction. A list of 14 Spheroidality - Curvature 34 Discarding and Recovering principles in its engineering original 15 Dynamics 35 Parameter Changes version is presented here, but how Partial or Excessive (Mann and Domb 1999) state for the business 16 Actions 36 Phase Transitions innovations, “we did not discover any 17 Another Dimension 37 Thermal Expansion ideas or innovations which caused us to believe there might be a 41st” 18 Mechanical vibration 38 Strong Oxidants 19 Periodic Action 39 Inert Atmosphere

20 Continuity of Useful Action 40 Composite Structures 3. Main components of sociosystemics F. TRIZ and Self-Organized Agent Based Models (SOABM)

4. These principles received further elaborated explications, both in engineering and in business. For example, Principle 2 “Taking out” is commented by D. Mann and E. Domb as follows:

“Separate an interfering part or property from an object, or single out the only necessary part (or property) of an object. Breakdown barriers between departments (Point No.9 of Deming’s Fourteen Points). Eliminate exhortations (Point No.10 of Deming’s Fourteen Points)… Separate development and production activities - skunkworks, tiger-teams, etc. The optimum committee has no members (Augustine’s Law #31)”

Two things are important here: first, the commentary links a general principle to much more concrete business terminology and established concepts; second, it links Principle’s recommendations to some well known in management sciences “laws” such as Deming’s 14 points (Deming, 1986), or 52 Augustine’s laws (Augustine 1997) (see others on slide 55). It allows, in principle, to integrate different relations used in these “laws” and concepts into a scheme of search for the best solutions and answer questions in SocioS where innovation is assumed (and not only that).

5. But another, even more important implication of TRIZ for sociosystemics, is that TRIZ gives a successful, and, maybe, the only example of how to operate with very wide sets of unrelated and non-formal issues and yet to obtain a strong, formalized and extremely constructive result. It provides hope, that similar unifications might be achieved also in sociosystemics: the list of important relationships; the unified forms of description of each scientific material; the standards of texts classification; classification of features, and correlations based on frequencies, all of which were discussed in this presentation. TRIZ is a tool for use by humans; but being included in the statistical and linguistic structure of sociosystemics it may ultimately acquire some semi-automatic features.

6. The last clear importance of TRIZ (or some similar techniques) is that it goes beyond the past into future, i.e. extends the statistical realm of the observed facts into creative “new world”. 3. Main components of sociosystemics F. TRIZ and Self-Organized Agent Based Models (SOABM) 6. From what have been said so far the following conclusion can be drawn: the only universal way to model social reality is to imitate it on the computer as precisely as possible; and for doing that apparatus of ABM is most adequate. In fact, it is the only possibility to aggregate different aspects of the minimalist social theory, since no analytical approach can provide all the necessary flexibility, robustness, and universality. There is no need to go into details of ABM modeling, since there is a vast and growing literature (see recent survey and development in Shoham and Leyton-Brown 2009) about it. I would just emphasize some aspects of it which will ultimately made ABM self- organized when embedded in the whole sociosystemics structure. It is better to illustrate it with a chart flow which combines all elements of sociosystemics (slide 62).

7. From this chart, it is clear that ABM are to be used when numerical solution is needed, i.e. if direct verbal answer and innovative qualitative solution by TRIZ are not enough to solve the problem. Please note, that previous steps (answering and TRIZ) are not lost; information from them is used in ABM section. Parameterization of ABM directly from the goal oriented system (a nucleus of SocioS after the system is triggered by a problem or a question) is what makes it self- organized, for it plugs ABM into the whole network of knowledge and leaves only some very specific local parameters initializations to the user. In other words, any ABM should not contradict with previously collected knowledge, but, instead, feed it back after successful implementation (changing components of the network and trustworthiness estimation). ………….. The flow chart shows all components of Sociosystemics and misses many details which I tried to explain in this presentation. It demonstrates how big the problem is and how many unclear aspects of it still exist. But on the other hand, it gives hope that, being recognized and properly weighed, these problems should find the best solutions, whether under sociosystemics name or not. Otherwise, the future of social sciences is rather gloomy.

Development of sociosystemics will need time, money, resources, and talents. I leave these problems beyond the scope of this talk…

Acknowledgements: I’ very grateful to Dr. M. Cherkes and Dr. I. Lipkovich for their help and fruitful discussions. Thank you ! 3. Main components of sociosystemics Sociosystemics Components (SosioS Prototype) Description of Social Reality Problem/Question Treatment Solution Feeding Parameterization model Numerical Agents, features, of ABM solution Natural language processing empirical relations, No

Enough? facts, frequencies Systemic Goal Yes Rethinking

Trustworthiness relational oriented Feeding Innovative qualitative of statements knowledge systems TRIZ solution estimation database

No

Scientific discourse unification Enough? Dialog Yes Concepts, ideas, models, theories, theoretical relations, Answers Answer statements, scientific documents, derivation Semantic Web from chain of arguments References R. Ackoff and H. Addison. A little book of f-LAWS. 13 common sins of management, Triarchy Press, 2006 G. Altshuller1996. And Suddenly The Inventor Appeared TRIZ, the Theory of Inventive Problem Solving. 2nd Edn., Technical Innovation Centre, Inc., Worcester, D.Ariely Predictably Irrational. Harper Perrenial, 2010 W.Arthur Increasing Returns and Path Dependence in the Economy.Ann Arbor: University of Michigan Press, 1994. N. Augustine Augustine's Laws AIAA, 1997 J. Berger. Could Fisher, Jeffreys and Neyman have agreed on testing? Statistical Science. Volume 18, Issue 1 (2003), 1-32 R.Berk, Regression Analysis: A Constructive Critique, Sage Publications, 2003. S.Blackburn, Truth. A guide, Oxford University Press, 2005. K. Cerulo Nonhumans in Social Interaction Annu. Rev. Sociol. 2009. 35:531–52 B.Christoph The Relation Between Life Satisfaction and the Material Situation: A Re-Evaluation Using Alternative Measures Soc Indic Res (2010) 98:475–499 A. Clark, C. Fox, and S. Lappin, Eds.,The Handbook of Computational Linguistics and Natural Language Processing, Wiley-Blackwell, 2010 CNN Team. Retracted autism study an 'elaborate fraud,' British journal finds, Jan.5, 2011, http://www.cnn.com/2011/HEALTH/01/05/autism.vaccines/index.html?hpt=T1 W.Deming (1986). Out of the Crisis. MIT Press D.Downey Black/White difference in school performance: the oppositional culture explanation. Annu. Rev. Sociol., 2008, 34, 107-126 I. Eibl-Eibesfeldt (1989) Human Ethology, AldinTransaction, USA, 2008 A.Ekholm Portfolio Returns and Manager Activity SSRN Working Papers, September 2010, http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1302329 J.Freese and S. Shostak Genetics and Social Inquiry Annu. Rev. Sociol. 2009. 35:107–28 G.Friedman 9/11 and the 9-Year War. STRATFOR, September 8, 2010, http://www.stratfor.com R. Greene. The 48 Laws of Power. Penguin, 2000 R.Herrnstein and C. Murray, Bell Curve: Intelligence and Class Structure in American Life, Free Press, 1994. D.Hillard, S. Purpura, J. Wilkerson Computer-Assisted Topic Classification for Mixed-Methods Social Science Research Journal of Information Technology & Politics, Vol. 4(4) 2007, 31-46 R. Jacoby and N.Glauberman (Eds.) The Bell Curve Debate. Times Books, 1995 A. Jacobsson War and peace - cyclical phenomena? Public Choice 2009, 141: 467-480 M.Jensen and W.Meckling, The nature of man Journal of Applied Corporate Finance, Summer 1994, V. 7, 2, 4 - 19. P. Johnson Modern times. The World from the Twenties to the Nineties. Harper Perennial, 1992 E.Khalil Natural selection and rational decision: two concepts of optimization J Evol Econ (2009) 19:417–435 D.Kuznetsov and I.Mandel Statistical physics of media processes: Mediaphysics Physica A 377 (2007) 253–268 B. Latour Reassembling the Social: An Introduction to Actor-Network-Theory, Oxford University Press 2005 I. Lim Sing Sheng and T. Kok-Soo Eco-Efficient Product Design Using Theory of Inventive Problem Solving (TRIZ) Principles J. List and D. Millimet, 2008 The Market: Catalyst for Rationality and Filter of Irrationality. The B.E. Journal of economic Analysis & Policy, V. 8, Issue 1, Article 47 T. Lukasiewicz and U. Straccia Managing uncertainty and vagueness in description logics for the Semantic Web Web Semantics: Science, Services and Agents on the World Wide Web Volume 6, Issue 4, November 2008, 291-308 R. Lynn The Global Bell Curve. Race, IQ, and Inequality Worldwide. Washington Summit Publishers, 2008 I. Mandel Statistical modeling and business expertise, or where is the truth? Model Assisted Statistics and Applications 3 (2008) 3–20 I. Mandel and D.Kuznetsov Statistical and physical paradigms in the social sciences Model Assisted Statistics and Applications 4 (2009) 39–62 39 I.Mandel Problem of definitions unification, Philosophical Sciences 11 (1975), 180–184. I.Mandel Statistical modeling in a wider frame: A marketing primer Model Assisted Statistics and Applications 4 (2009) 215–237 D. Mann & E. Domb, “40 Inventive (Management) Principles With Examples”, The Online TRIZ Journal, September, 1999. http://www.triz-journal.com/archives/1999/09/a/index.htm D.Matsumoto Ethnic Differences in Affect Intensity, Emotion Judgments, Display Rule Attitudes, and Self-Reported Emotional Expression in an American Sample, Motivation and Emotion, VoL 17, No. 2, 1993 J.McCauley, Dynamics of Markets: Econophysics and Finance, Cambridge University Press, 2004. D.McCloskey and S. Ziliak, The Cult of Statistical Significance: How the Standard Error Costs Us Jobs, Justice, and Lives, The University of Michigan Press, 2008. P.Mirowski More Heat than Light: Economics as Social Physics, Physics as Nature's Economics Cambridge University Press, 1989 D.Moore, Statistics. Concepts and controversies, W. H. Freeman and Company, 5th edition, 2001.

M. Newman, A. Barabási, D. Watts (Editors) The structure and dynamics of networks. Princeton Universuty Press, 2006 M.Newman, Power laws, Pareto distributions and Zipf’s law, Contemporary Physics 46 (2005), 323–351. J.Pearl, Statistics and Causal Inference: A Review (together with following discussion and a rejoinder), Test 12(2) (2003), 281–345. J.Perloff Microeconomics: Theory and Applications with Calculus, Pearson education, 2008 L.Perlovsky (2007). Neural Dynamic Logic of Consciousness: the Knowledge Instinct. In Eds. L.I. Perlovsky, R. Kozma, Neurodynamics of High Cognitive Functions, Springer. Preprint. I.Prigogine, The end of Certainty, Time, Chaos and the New Laws of Nature, The Free Press, New York, 1997. N.Rescher Epistemetrics, Cambridge University Press, 2006

B. Roehner Driving forces in physical, biological and socio-economic phenomena Cambridge University Press, 2007 D.Rubin, Matched Sampling for Causal Effects, Cambridge University Press, 2006. J.Salerno Menger’s causal-realist analysis in modern economics Rev Austrian Econ (2010) 23:1–16

S.Schwartz, (1994). Are there universal aspects in the structure and contents of human values? Journal of Social Issues, 50, 19-45. Y. Shoham and Kevin Leyton-Brown, Multiagent Systems: Algorithmic, Game-Theoretic, and Logical Foundations. Cambridge University Press, 2009 N.Smelser and R.Swedberg (Eds), 2005,The Handbook of Economic Sociology, Princeton University Press V. Souchkov Breakthrough thinking with TRIZ for business and management: an overview, 2007, http://www.newshoestoday.com/library/cms/2007-08-17TRIZforBusinessAndManagement.pdf C. Sunstein, A.Vermeule Conspiracy Theories 2008, Harvard Public Law Working Paper No. 08-03 http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1084585 A.Uemov Veshchi, svojstva i otnosheniya. Nauka, Moskva, 1963 (Things, features, and relations, Moscow, 1963) R.von Mises Probability, statistics, and Truth. Dover Publications, 1957 S. Wasserman and K. Faust Social Network Analysis. Cambridge University press, 2006 A. Wierzbicka Emotional Universals. Language Design 2, 1999, 23-69 Y.Yang, Raine, A., Lencz, T., Bihrle, S., Lacasse, L., and Colletti, P. (2005). Prefrontal structural abnormalities in liars. British Journal of Psychiatry 187 320-325 L.Zadeh, Causality is indefinable, www.cs.berkeley.edu/˜nikraves/zadeh/Zadeh2.doc, 2003. P.Zeihan Europe: The New Plan. STRATFOR, December 21, 2010, http://www.stratfor.com