MY EDITORIAL ON YOU TUBE
After having devoted some of my personal energy to reviewing other people’s science (see Second-hand observations), I return to my own science, i.e. to my book on the civilizational role of cities. Reviewing that manuscript in the field of energy management gave me some inspiration. I realized that the core message I wanted to convey in the book was that human societies have collective intelligence, that intelligence manifests itself in typical, recurrent patterns, and cities are one of those patterns, where creating a demographic anomaly allows creating new social roles for a growing population, and assuring functional partition between two types of settlements: agricultural land for producing food, on the one hand, and urban land for producing new social roles and new technologies, on the other hand. Moreover, cities are the base of markets, and of the market-based economy. The whole social system based on the development of skills and technologies, so as to produce tradable surpluses, that whole system is precisely based on the role of cities. Maybe there are other social structures to obtain the same result, but we haven’t figured them out yet. With the creation of cities, we developed a pattern of further development, where apparently fundamental markets interweave with the apparently futile ones, and the whole system facilitates technological change and the creation of new social roles. The social system based on cities is like a social vortex, largely powering its own momentum and sucking new people into it.
That overview of my thinking brings me one more time to the issue of collective intelligence and to the big methodological question: to what extent can neural networks be used to simulate collective intelligence in human societies? I know, I know, this is some n-th time I return to that thing. Still, it is important, both methodologically and fundamentally. There is a whole big stream of research, including my own, where neural networks are used as mathematical tool for validating theoretical model. I can see that neural networks tend to replace the classical methods, such as GARCH, ARIMA or the good old Ordinary Least Squares. All that stuff works with the same basic data, i.e. with residual errors which inevitably appear as we try to apply our grand ideas to the brutal reality of life. Still, the way that neural networks process those errors is fundamentally different from stochastic models. The latter just cut through the entire set of data with one line (or curve, for that matter), which minimizes the sum total of residual errors, all in one go. Neural networks are more patient, and they minimize error by error, case by case. Neural networks learn.
The point is that when I use a neural network to validate a theoretical model in social sciences, I should substantiate the claim that the network represents the way of learning in the given society. The best theory of learning which I have found so far is the Interface Theory of Perception (Hoffman et al. 2015[1]; Fields et al. 2018[2]; see also I followed my suspects home). I rephrase it shortly and I try to put it against (or next to) my own methodology.
When an aggregate socio-economic variable, such as e.g. GDP per capita or energy consumption per capita, changes over time, it allows assuming a society doing something differently as time passes. In other words, those aggregate variables are manifestations of collective decisions and collective action. Question: how are those collective decisions being taken and how are they being turned into action? Some sort of a null assumption is that we have no way to guess anything about that process. Still, I think I can make a slightly stronger assumption, namely that we collectively know what we are doing, we just know it imperfectly. Therefore, when I observe a variable such as GDP per capita, or the average number of hours worked per person per year, change over years, I can assume it manifests a collectively intelligent adaptation: we do something together, we contemplate the outcomes, we say ‘Blast! It is not exactly what we meant. Folks! Get ready! We adapt! That rate of secondary education has to change! We are running a civilisation here, aren’t we?’, and we engage into another set of decisions and actions.
Collective decisions and collective action mean that people argue and diverge in what they say they intend to do, in what they really do, and in what they claim they have just done. We diverge from each other and we lie to each other on the top of it, and we lie to ourselves, and yet that whole business of civilisation seems to be working. We have a set N = {se1, se2, …, sen} of n social entities (people, basically, or various agglomerations thereof), and they all cheat, lie, and egoistically get after it, in the presence of a set R = {r1, r2, …, rm} of m external stressors (viruses, presidents, wars, bad crops etc.). Mind you, as long as n > 1, i.e. as long as there are many social entities, probably at least one of them is doing things sufficiently well to survive in the presence of m stressors.
We have those n social entities trying to get by in the presence of m external stressors, and one could wonder how that lot can learn anything? I subtly change the shade of the question ‘how?’ into ‘how can we know that?’. How can we know that a social entity has learnt anything in the presence of external stressors? Learning can be understood from two perspectives: subjective internal impression of having learnt something, on the one hand, and objective, externally observable fact of having acquired new skills. When I prepare the same sauce 30 times, 20 times it is completely spoilt, 9 times it sort of approaches the ideal, and 1 time, the final one, I have the impression I nailed it. I have the impression I have learnt something, however it does not mean other people think the same.
I need a gauge to assess learning. One possible method, more specifically the one used in artificial neural networks, consists in checking how close my social entities are to a pre-defined desired outcome. In more elaborate artificial neural networks, the pre-defined component might be just the method of setting the desired outcome. That outcome can be simply survival or something more, such as achieving a certain amount of something valuable.
Good, so I have those n social entities, and the stressor they act under is the pressure to achieve a desired outcome, i.e. to obtain a certain payoff. The social entity sei which gets the closest to that outcome, or which, in other words, collects the greatest payoff, can be considered as the most successful. Life is reproduction. People die, and new people are born. Governments change. On the long run our set N = {se1, se2, …, sen} of n social entities is interesting to the extent that it reproduces into another set Nk of n(k) social entities. Social change can be viewed as an (almost) ever-lasting chain of sets, each with social entities inside: N1 transforms into N2, which turns into N3 etc.
I think I have just nailed an important point (involuntarily, to be clear). When I study any kind of social change, I can make one of the two alternative assumptions: continuity of social entities versus their generational reproduction. Social structures can be seen such as I have just described it: as changing sets of social entities. Under that angle, the 38 million people in my native Poland today are a different set of people from the roughly 36 million who were around when I was 10, i.e. in 1978. It does not necessarily mean that each person present in 1978 died and has been replaced by someone else; I am pretty sure I didn’t die. However, some people died, some new people have come to the fore, some people changed significantly etc. On the whole, the set N2020 is different from the set N1978. There is a different angle for looking at the same reality: people in Poland, 2020, are the same big social entity as the one in Poland, 1978, and it is just the internal structure of that entity that has changed.
What is the practical (well, theoretical) difference between those two angles of approach to the same theatre of social change, i.e. consecutive sets of small entities as opposed to consecutive states of one big entity? When I simulate social change as a sequence of sets, where individual components can change their properties, a long sequence of that type is like a journey of discovery. Each consecutive set Nk comes out of learning that occurred in its predecessor Nk-1. The transformation of one set into another certainly follows some constraints, yet a long sequence of such transformations is like a long hike up a hill: we have to take turns around boulders and ravines, we have to choose between paths of different difficulty, and usually an easier path is a less steep one, thus a longer and slower one. This type of change, social or biological, is known as adaptive walk in rugged landscape in Kaufman & Levin 1987[3]. Mathematically, it is a Markov chain, i.e. a chain of states, where the properties of each consecutive state are determined just by the properties of the previous state as well as by the algebra of transformation from one state to another (the so-called σ-αλγεβρα, oops! Excuse me, I wanted to say σ-algebra).
When I take the other approach to a social structure, i.e. when I study it as one big, perennial social entity which undergoes structural change inside, that change is something like different shapes of the same thing. I noticed strong marks of such an approach in that scientific paper entitled ‘Evolutionary Analysis of a Four-dimensional Energy- Economy- Environment Dynamic System’, which I was reviewing recently on the request of the International Journal of Energy Sector Management (ISSN1750-6220). In that paper, a complex system of relations between economy, energy and society is represented as four gradients of change in, respectively, volume of pollution x, real economic output y, environmental quality z and energy reduction constraints w. I know it is a bit abstract, at this point, yet I want to make an effort and explain it. Imagine an irregular quadrilateral, i.e. a rectangle with broad intellectual horizons. Four angles, four edges, yet the angles do not have to be right and the edges do not have to be parallel in pairs. Just four of each lot. The length of each edge corresponds to the gradient of change in one of those 4 variables: x, y, z, and w. Any substantial change in that system is a change in lengths of those 4 edges, and, as it is a closed polygon, it entails a change in angles between edges.
As I am trying to grasp fundamental differences between those two views upon social change, namely sequence of sets as opposed to an internally changing perennial entity, I think the difference is mostly epistemological. As a matter of fact, I don’t know s**t about reality as it is, and, let’s be honest, neither do you, my dear readers. We just make many possible Matrixes out of the real stuff and settle for the one that offers the greatest rewards. This is the stance adopted in the Interface Theory of Perception (Hoffman et al. 2015[4]; Fields et al. 2018[5]), as well as in classical Western empiricism (see William James’s ‘Essays in Radical Empiricism’, 1912). This holds for social reality as well as for anything else. When I see social change, I see most of all change in itself, and only secondarily, in my spare moments, I can try to figure out what exactly is that thing that changes. This is science, or philosophy, depends on the exact method I adopt, and this is hard, and time-consuming. Most of the times, I just use a ready-made explanation, conveyed in my culture, that what is changing is, for example, the market or the constitutional order, or the set of cultural stereotypes. Still, at the bottom line, those labels are just labels. What I am really experiencing, is change in itself.
When I assume that social change is a Markov chain of sets made of small social entities, I study social change as change in itself, i.e. as the say σ-algebra of that chain. I do not pretend to know exactly what is happening, I just observe and give the account of the passage from one state to another. Conversely, when I assume that social change is structural recombination inside a big, perennial social structure, I pretend to know the limits and the shape of that big structure. This is a strong assumption, probably an overstated one.
Now, I connect the dots. I am linking my thinking about cities with my thinking about collective intelligence, and all that I serve in a sauce peppered with the possibility to use artificial neural networks for studying the whole business. I can study the civilizational role of cities under those two angles, i.e. as a Markov chain of states which I barely understand, yet which I can observe, on the one hand, or as internal reshuffling inside a finite, big social entity called ‘civilisation’, with nicely outlined contours. I am honest: I am not even going to pretend I can outline the precise contours of the civilisation we live in. With all the s**t going out there, i.e. leftist extremists in Germany erecting an illegal statue of Lenine, in response to #BlackLivesMatter in United States, and neo-Nazi extremists from The Base organization receiving orders from a leader who is an American with an Italian name, currently living in Russia: man, I do not feel up to trace the external contours of that thing.
I know that I can and want study the phenomenon of cities as change in itself, and I assume that the change I see is an aggregation of local changes in small social entities sei. As those small sei’s change locally, their local transformations translate and manifest as the passage from the aggregate set Nk = {se1, se2, …, sen} into another set Nk+1 = {se1, se2, …, sen}. The next hurdle to jump over is the connection between sets of the type Nk = {se1, se2, …, sen} and aggregate socio-economic variables commonly used as so-called statistics. Those ‘statistics’ tend to have one of the 4 possible, mathematical forms: averages, totals, frequencies, or rates of change. When they are averages, e.g. GDP per capita, they are expected values of something. When the come as aggregate totals, e.g. aggregate headcount of population, they stand for the size of something. As they take the form of frequencies, e.g. the percentage of people with secondary education, they are simple probabilities. Finally, as rates of change, they are local first derivatives over time in some functions, e.g. the function of economic growth.
Each of those mathematical forms can be deemed representative for a heterogenous set of small phenomena, like small social entities sei. I assume that each set Nk = {se1, se2, …, sen} of n social entities manifests its current state in the form of a complex vector of variables: expected mean values, total sizes, simple probabilities of specific phenomena, and first derivatives over time in the underlying functions of change. Any set of socio-economic variables is an imperfect, epistemic representation of current local states in the individual social entities sei included in the set Nk = {se1, se2, …, sen}.
As I go through my notes and blog updates from the last 2 months, something emerges. The social entities I focus on, thus my sei‘s, are individual people endorsing individual social roles. set Nk = {se1, se2, …, sen} is essentially a set of people, i.e. a population. Each of those people has at least two coordinates: their place of residency (mostly defined as city vs countryside), and their social role. I messed around with a set like that in a neural network (see The perfectly dumb, smart social structure). The current state of the whole set Nk manifests itself as a vector Vse of socio-economic variables.
So far and by far, the most important variable I have identified is the density of population in cities, denominated over (i.e. divided by) the general density of population. I named this variable [DU/DG] and I assume it shows the relative social difference between cities and the countryside (see Demographic anomalies – the puzzle of urban density). The coefficient [DU/DG] enters into interesting correlations with such variables as: consumption of energy per capita, income per capita, surface of agricultural land, cereal yield in kg per hectare of arable land, number of patent applications per 1 million people, and finally the supply of money as % of the GDP. On the other hand, by studying the way that urban land is distinguished from the rural one and from wildlife, I know there is a correlation between density of urban population and the density of man-made structures, as well as the density of night-time lights.
Good. I have a set Nk = {se1, se2, …, sen} of n social entities, which changes all the time, and a vector Vse = {DU/DG; energy per capita; income per capita; surface of agricultural land; cereal yield; patent applications; supply of money} of variables pertinent regarding cities and their role. Between the two I insert my mild obsession, i.e. the set SR = {sr1, sr2, …, srm} of ‘m’ social roles.
Now, I go pictographic. I use pictures to make myself spit out the words I have in mind. I mean, I know I have words in mind, only I don’t know what exact words are these. Pictures help. In Figure 1 I am trying to convey the idea of proportion between the headcount of population and the range of social roles available to endorse. My basic assumption is that we, humans, are fully socialized when we endorse social roles that suit our basic personal traits, such as intelligence, extroversion vs introversion, neuroticism, conscientiousness, the way we look in seasoned black leather etc. The state of society can be explained as a balance between the incremental headcount of humans, on the one hand, and the incremental range of social roles to take. If the headcount of humans is n, and the number of social roles available is m, we are talking about ∆n/∆m.
When both sets, i.e. Nk and SR change at the same pace, i.e. ∆n/∆m (t0) = ∆n/∆m (t1), the society is in some sort of dynamic equilibrium, like a constant number of humans per one social role available. When the set SR of social roles burgeons faster than the pace of demographic growth, I mean when ∆n/∆m (t0) > ∆n/∆m (t1), logically there is less and less humans per one social role. This is social change by differentiation. New, idiosyncratic skillsets and behavioural patterns emerge. This is like an absorptive state, which can suck new humans in like easy.
On the other hand, when demographic growth in the set Nk races far ahead, and the set SR of social roles lags behind, i.e. ∆n/∆m (t0) < ∆n/∆m (t1), there is consistently more and more humans per one social role. That favours standardization and institutional definition of those roles, in the form of professions, public offices, ritualized social statuses etc. Society settles down into a nice order. Still, each such institutionalized social role grows barriers to entry around itself. You need to pass some exams, you need to be appointed or elected, you need to invest some capital… New humans coming to the world encounter those barriers, and some of them end up by saying: ‘F**k it! I stay outside of everything’. This is the tipping point, when social change is needed, so as to make social room for new humans.
Figure 1
Now, I transition into the role of cities in that social pattern. I am trying to picture the idea in Figure 2. If the state of social differentiation, we need some pattern for generating diversity. We need social interaction. Cities can be seen as a social contrivance which facilitates such interaction. Still, it comes to my mind sort of right now, we don’t really know (yet), to what extent digital interaction between humans can replace the real one in that specific respect, i.e. as a mechanism of creating new social roles. My gut feeling is that digital technologies can be at least imperfect substitutes of real social interaction. You Tube or Instagram may replace cities in their civilizational role of creating new social room for new homo sapiens. We might just be evolving from a civilization of city slickers living next to rednecks, into a civilisation of city slickers, rednecks and homo onlinus.
Figure 2
In the next step, I am wrapping my mind around the math side of the problem, which I try to picture in Figure 3. I guess that what I have in terms of empirical data to put in a neural network is mostly the vector Vse of social outcomes, which I can enrich with the headcount of population, and that would be the real-life material that a neural network could learn from. What that network could try and optimize could be the gradient ∆n/∆m or some variation thereof, as the exact number of social roles is technically unobservable with the current state of technology. When I think about the practical way of doing it, I can imagine a network pegged on optimizing some sort of hard-nailed output variable, such as the average number of hours worked per person per year (solid stuff, as it comes out of my so-far research). I drop the gradient ∆n/∆m among the input variables, and I try to discover what value thereof the network would yield after a few thousands laborious trials to produce artificial history.
Another angle of approach that comes to my mind is to take all the known empirical variables as input, and the gradient ∆n/∆m as the output. Then I make different clones of the network, with ∆n/∆m going various ways, like gently up, steep up, down a bit etc. I can check which of the clones displays the closest Euclidean distance to the source empirical dataset.
Figure 3
Now, the final step: I connect the concept of social role with that of conscious agent, as represented in the Interface Theory of Perception (Hoffman et al. 2015[1]; Fields et al. 2018[2]). Figure 4 represents my pictographic thinking about it. Social roles are temporary outcomes of learning and social interaction between Conscious Agents (CA). In other words, social roles form as humans exchange information about their conscious experience, which serves to translate objectively existing states of the world into material possible to process by our brain, so as to decide whether to run away from the tiger or maybe rather kill the tiger and take the antelope. We take action consequently, we contemplate its consequences, and we talk about it, and we show to each other how we can learn new stuff.
Figure 4
Discover Social Sciences is a scientific blog, which I, Krzysztof Wasniewski, individually write and manage. If you enjoy the content I create, you can choose to support my work, with a symbolic $1, or whatever other amount you please, via MY PAYPAL ACCOUNT. What you will contribute to will be almost exactly what you can read now. I have been blogging since 2017, and I think I have a pretty clearly rounded style.
In the bottom on the sidebar of the main page, you can access the archives of that blog, all the way back to August 2017. You can make yourself an idea how I work, what do I work on and how has my writing evolved. If you like social sciences served in this specific sauce, I will be grateful for your support to my research and writing.
‘Discover Social Sciences’ is a continuous endeavour and is mostly made of my personal energy and work. There are minor expenses, to cover the current costs of maintaining the website, or to collect data, yet I want to be honest: by supporting ‘Discover Social Sciences’, you will be mostly supporting my continuous stream of writing and online publishing. As you read through the stream of my updates on https://discoversocialsciences.com , you can see that I usually write 1 – 3 updates a week, and this is the pace of writing that you can expect from me.
Besides the continuous stream of writing which I provide to my readers, there are some more durable takeaways. One of them is an e-book which I published in 2017, ‘Capitalism And Political Power’. Normally, it is available with the publisher, the Scholar publishing house (https://scholar.com.pl/en/economics/1703-capitalism-and-political-power.html?search_query=Wasniewski&results=2 ). Via https://discoversocialsciences.com , you can download that e-book for free.
Another takeaway you can be interested in is ‘The Business Planning Calculator’, an Excel-based, simple tool for financial calculations needed when building a business plan.
Both the e-book and the calculator are available via links in the top right corner of the main page on https://discoversocialsciences.com .
You might be interested Virtual Summer Camps, as well. These are free, half-day summer camps will be a week-long, with enrichment-based classes in subjects like foreign languages, chess, theater, coding, Minecraft, how to be a detective, photography and more. These live, interactive classes will be taught by expert instructors vetted through Varsity Tutors’ platform. We already have 200 camps scheduled for the summer. https://www.varsitytutors.com/virtual-summer-camps
[1] Hoffman, D. D., Singh, M., & Prakash, C. (2015). The interface theory of perception. Psychonomic bulletin & review, 22(6), 1480-1506.
[2] Fields, C., Hoffman, D. D., Prakash, C., & Singh, M. (2018). Conscious agent networks: Formal analysis and application to cognition. Cognitive Systems Research, 47, 186-213. https://doi.org/10.1016/j.cogsys.2017.10.003
[1] Hoffman, D. D., Singh, M., & Prakash, C. (2015). The interface theory of perception. Psychonomic bulletin & review, 22(6), 1480-1506.
[2] Fields, C., Hoffman, D. D., Prakash, C., & Singh, M. (2018). Conscious agent networks: Formal analysis and application to cognition. Cognitive Systems Research, 47, 186-213. https://doi.org/10.1016/j.cogsys.2017.10.003
[3] Kauffman, S., & Levin, S. (1987). Towards a general theory of adaptive walks on rugged landscapes. Journal of theoretical Biology, 128(1), 11-45
[4] Hoffman, D. D., Singh, M., & Prakash, C. (2015). The interface theory of perception. Psychonomic bulletin & review, 22(6), 1480-1506.
[5] Fields, C., Hoffman, D. D., Prakash, C., & Singh, M. (2018). Conscious agent networks: Formal analysis and application to cognition. Cognitive Systems Research, 47, 186-213. https://doi.org/10.1016/j.cogsys.2017.10.003
