Week 4 (4/15,17): Data Science

[Gio-Choi]

Data science began its life as stastistics in astronomy, where scientists sought to combine observations across observatories to accelerate celestial science and superior navigation. The regularities these measurements demonstrated invoked the notion of natural law, with a small range of error. These methods inspired practitioners like Belgian astronomer and "social physicist", Adolphe Quetelet, Francis Galton and others to apply these methods to the study of society for the discovery of “social laws” and reinterpret errors of measurement as deviations from a standard (e.g., Quetelet's "Average Man"). What does social law mean for human free will and culpability? If burglaries, homicides and suicides are consistent, regular events—if they deviate from the norm but fulfill some natural law—then are their perpetrators culpable for their crimes? And if you can compute the causes and consequences of social patterns, can you simulate responses and alter them? We explore these issues and others associated with the generalization of mathematical, statistical and computational methods to human bodies and human minds with the rise of statistics, "data science" and the social and behavioral sciences  (from psychology, sociology and economics to education, criminology and business.)

Post your response as a Comment in reply to this message.

[suryachinnappa]

The term “statistics” has an interesting etymology which sheds light onto its origins, as this week’s readings allude to: it originated as the study of states (as in, political entities); namely, the tracking of information about them–in fact, its original English name was “political arithmetic”. In this, its origins and development are inseparable from the social sciences. This vastly differs from our modern conception of statistics; we often think of it more as allied to mathematics (or part of it) than anything else, and secondarily more affiliated with the so-called “hard sciences”, in opposition to the “soft” social sciences–indeed, even at UChicago, our Department of Statistics lies within the Physical Sciences Division, baking these perceived affinities into the bureaucracy of the university. However, in the readings from this week, this affinity is unsettled, particularly in the assumption of its longstandingness. As we see, the methods that comprise statistics first arose in the numerical tracking of information within states, arising particularly when this sort of information was growingly considered valuable due to the sociopolitical forces of the time. It was a great shock to these early statisticians when they realized that many of the things they were keeping track of seemed to follow periodic and constant patterns, when measured at the scale of an entire society. Certain things, like birth rates and child mortality, seemed to lend themselves particularly well to meaningful interpretation (for example, that the slight bias towards males in total births was counteracted by a slight bias towards males in child mortality, leading the population of the sexes to be roughly equal at matured age, fitting in with the divine design for maximal reproductive potential).

Other things, however, seemed deeply unintuitive and groundbreaking at an epistemological level. For example, the fact that various sorts of crimes, and even suicides, occurred at a steady, periodic rate in society over the course of years seemed deeply unsettling. Such actions are typically accounted for at the individual scale, explained (if at all) by the particular psychological and social state of the individual involved and the immediate conditions of the crime performed. With the observation of this seeming natural order, however, questions began to emerge as to how to adapt that worldview to the revolutionary and perhaps terrifying proposition that humans might be governed by natural law, just as the physical world was. Questions about individual agency naturally arose on account of this; after all, if there’s a yearly quota on the number of incidents of a certain type of crime, would removing one person from the circumstances by which they would commit this crime somehow magically predispose another to take their place?

A comedic sidenote lies in the relationship between statistics and data science. Statistics is often considered to be more focused on the mathematical aspect, with data science posed as its fresher, more interdisciplinary sibling, using empirical data collection and domain knowledge to enrich dry statistical theory–despite the fact that statistics itself was originally built upon these empirical–social–foundations!

Discovery: The Google Colab assignment was very insightful; while I was familiar with some of the statistical concepts covered before, the assignment seemed to place particular emphasis on the specific use cases of the different statistical formats, which I found very interesting. It reminded me further of the inexorable link between statistical theory and its social scientific origins. A textbook I once read made reference to statistics as a sort of mathematical engineering–with the historical context given by the readings and the interactive component in this week’s Explore, I feel like I understand that perspective more than ever: just how engineering is the application of the natural sciences to the benefit of social systems, so too is statistics the application of probability theory, combinatorics, and other areas of mathematics to better understand social phenomena.

[eeengland1]

The discussion of the nature of man in Quetelet’s treatise, and how seemingly random, even irrational actions of humans can be predicted in a confident fashion when looking at the aggregate of society— while a seemingly normal idea now— was certainly disturbing to Quetelet’s time. The first pages of the treatise (V-X, 5-12) discussed expected statistics, that is, relatively acceptable averages of aspects of human behavior such as birthrate and the anomaly between male and female birth rates— things that are and were commonly assumed to have a rational biological meaning behind them, thus unsurprising to predict.Indeed, the birth rate discrepancy turns out to be an evolutionary skew that normalizes later in the generation’s life, as male death rates are higher than females’. But A Treatise on Man’s more significant focus on the modeling of psychological behavior and seemingly impossible-to-predict aspects of human nature (i.e. crime and suicide), and it’s relative success in demonstrating how these irrational acts can be predicted brings questions on to what extent are our actions predetermined, or led by external factors. How much free will actually manifests itself in irregular, impossible-to-model activity in daily life?

The advent of statistics in society, bringing about these realizations, had allowed statisticians and political scientists of the time to create the “average man”, a model containing the aggregate and average data of human activity, and thus easy to predict. While predicting the outcome of a single individual was still as impossible as before, the average birth and death rates of society were easily modeled. The statisticians of the time liked adhering to the “facts”, raw data and the numbers their statistics produced, things generally indisputable. Even probability was seen as “speculation”; an absurd statement now, but less so at the time when the field wasn’t as developed. Of course, plenty at the time disputed statistics as an actual science, claiming that no advancement was being made in understanding the underlying causes for the data that was being produced, and that statistics was closer to a method than an actual science.

It seems however, that Quetelet was deeply interested in the origins of statistics, and how the field could be truly expanded into the social sciences. Quetelet gave significant credit to the astronomers who first developed this field, and actually disliked the “mass of statistical reformers” at the time. Quetelet, at least, wanted to try and find the underlying rules behind the statistics he computed, and pursued the idea of principles of human conservation, indeed, the laws that govern chaos.

Discovery: The collab notebook really ties the discussion of statistics and social sciences together! I also just really like probability mass and cumulative distribution functions and think they’re really useful in, well, most of statistics and probability, which is then useful in modeling a myriad of real-world topics and predicting outcomes. Also, it’s a shame that probability wasn’t a widely pursued field by “most” statisticians of the 1800s.

[mhmanek]

Adolphe Quetelet’s notion of “social law” was meant to extend Newtonian regularity to humanity: just as gravity specifies the probable path of a comet, the calculus of large numbers seems to specify the average height, age at marriage, or propensity to offend within a population. Yet treating these statistical regularities as “laws” does not force us into a stark determinism. They reveal boundary conditions—limits within which individuals make choices—rather than scripts that dictate those choices. Free will remains intelligible on the level of the person, while “social law” operates on the level of aggregates; the two descriptions coexist in the same way thermodynamics coexists with molecular chaos. Hence, recognizing social law refines but does not abolish the philosophical vocabulary of agency.

If burglaries, homicides, or suicides recur with clock‑like stability from year to year, that regularity seems at first to erode individual culpability: how can we blame a thief for fulfilling a statistical expectation? The better conclusion is subtler. Statistical stability does not make a given act inevitable; it only shows that the social environment supplies a certain volume of temptations, stresses, and opportunities that many—but not any particular—individuals will succumb to. Responsibility therefore bifurcates. Societies bear a collective responsibility for structuring risk, just as architects must account for average wind load, but the wrongdoer still exercises volition at the moment of decision. Law and ethics have long acknowledged such dual responsibility: the jurist asks both mens rea (individual intent) and causa socialis (social causes like poverty, policing, or gun availability) when judging crime.

Once we accept that social patterns are partially computable, it becomes tempting to imagine simulating counterfactual interventions and steering the aggregate. To a degree this is already routine: epidemiologists model vaccination campaigns, urban planners forecast traffic flows, and central banks tweak interest rates expecting probabilistic reactions. Yet simulation is not synonymous with control. Models translate noisy historical covariances into formal machinery, but they cannot capture every nonlinear feedback, local meaning, or emergent adaptation. Intervening on a social system is therefore closer to gardening than to machining: we can bias probabilities—make a neighborhood safer by adding street lighting, for instance—but we cannot guarantee specific outcomes for specific persons. Moreover, excessive reliance on simulation risks “moral distance,” encouraging policymakers to treat citizens as datapoints rather than deliberating agents.

In sum, social law reframes free will without negating it, shifts culpability toward a dual focus on individual intent and structural conditions, and opens a pragmatic but morally fraught space for probabilistic intervention.

Discovery
Working through the Colab notebook pushed me to see distributions not just as abstract curves but as living portraits of social life. Building PMFs, CDFs, and KDEs for the GSS data made the contrast vivid: the PMF spotlighted the modal age, the CDF revealed the “texture” of the whole age range, and the KDE smoothed away distracting noise without hiding the heavy‐tailed reality of income. I was surprised by how quickly a CDF lets you read off policy‑relevant facts—say, that barely one‑eighth of respondents are over 65, or that the top 5 percent of earners start at roughly $150 k. Fitting and visually testing normal and lognormal models drove home a Quetelet‑style lesson: models are provisional summaries, useful precisely because the data often refuse to fit cleanly. Filtering on sex and education with simple Boolean masks then exposed how structural categories twist entire distributions, not just their means. By the end, I felt I’d internalized a workflow—load messy social data, visualize from multiple angles, confront it with theory, iterate—that makes the numbers speak both mathematically and sociologically.

[evancholerton]

Right from the start, 19th-century statisticians began to challenge the idea that society was just the sum of free-willed individuals. Quetelet’s invention of l’homme moyen, the “average man,” made clear that when you compile enough births, deaths, or even crimes, a stable pattern emerges that seems to obey its own natural law. As Gigerenzer recounts, “Already in the early 1830s... the average man... really was the type of nature, just as the mean of astronomical observations... is the true position of the star” (Gigerenzer, page 54). That abstraction allowed reformers to think in terms of social trajectories, but it also risked flattening individuals beneath a blanket of determinism.

On the other side of that boundary was the ideal of objectivity embraced by statistical societies. The London Statistical Society adopted the motto Aliis exterendum, “to be threshed out by others,” and insisted that “statistics... was not concerned with causes or effects; indeed, its ‘first and most essential rule’ was ‘to exclude all opinions’” (Gigerenzer, page 38). Many statisticians resisted probabilistic speculation, preferring dry tabulation over inference. But in doing so, they risked evacuating meaning from the very social dynamics they were measuring.

The tension between reifying averages and obsessing over data purity echoes in today’s data science. The 19th century’s use of the law of large numbers helped justify the eerie stability of social phenomena like suicides or murders. This law, in practice, referred to the observation that many social events “occur in approximately constant numbers from year to year” (Gigerenzer, page 40). But stability is not inevitability. These regularities instead suggest social boundary conditions within which individuals still make choices.

This idea also influenced the development of modern probability theory. The frequency interpretation, which understands probability as a long-run regularity, was born not in games of chance but in the social worlds of insurance and statistics. Maxwell and Boltzmann, for example, used social regularities and Quetelet’s statistical distributions to justify statistical mechanics. As Gigerenzer notes, this was a striking example of how “social science” helped inspire the natural sciences.

Questions worth exploring: How do modern algorithms respect or obscure the same boundary between individual nuance and aggregate patterns? When machine learning systems learn from millions of data points, do they essentially reconstruct l’homme moyen, or can they preserve outliers and variation in meaningful ways?

Discovery

One detail that stuck with me was Quetelet’s use of conscript data to build his model of the average man. He treated thousands of chest measurements as if they were hundreds of identical statues, a kind of proto-datafication of the human body (Gigerenzer, page 41). This feels directly connected to how AI systems today rely on proxy data like credit scores or test scores as stand-ins for complex human traits. It’s a reminder that abstraction can be powerful, but it always raises the question of what gets lost in translation.

[Laurelli7]

There are definitely some kind of social laws that we can clearly see through big data approaches. People like Quetelet, Gigerenzer, and Porter changed social science dramatically by introducing statistical ways of thinking and applying big data to study society. Even before modern quantitative methods, it was clear that collective human behavior often differed significantly from individual actions. Big data is especially suited to capturing these collective patterns because it provides insights that traditional individual-focused perspectives can't fully reveal.
However, acknowledging collective patterns does not mean we should use these insights to determine individual guilt or innocence. The issue of culpability—whether someone is responsible for their crimes—should remain separate from these collective observations. The collective view and the individual view are essentially two different lenses for looking at human behavior. While they can complement each other, one should not directly influence decisions about individual responsibility. Collective social patterns do show correlations and regularities, but correlation doesn't necessarily equal causation. Just because the data shows consistent crime rates, it doesn't mean society "forces" specific individuals to commit crimes.
For these reasons, I strongly disagree with social deterministic perspectives like Gigerenzer's, where individual behaviors seem to lose their importance under collective laws. Accepting this determinism would mean that the collective pattern causes the individual actions—this isn't logically correct. The opposite logic, saying individual actions change collective patterns, isn't widely supported either. Both views oversimplify a complex relationship, where collective and individual perspectives coexist without directly causing each other.
If we start treating social laws as reasons to reduce individual culpability, we would face ethical problems and potential chaos. If individuals were no longer considered responsible for their actions because of predictable social patterns, the legal and moral systems would collapse—anyone could commit crimes without personal responsibility. Clearly, that would cause social catastrophe. This shows why individual culpability must always be preserved and separated from collective statistical insights.
Yet, this doesn't mean collective insights from big data aren't extremely useful. In fact, big data can guide us toward important decisions at the collective level. For example, policy-making, business strategies, and public health initiatives can greatly benefit from understanding societal trends and behaviors revealed by data. By recognizing these patterns, governments and organizations can proactively address root causes—such as poverty, education, or unemployment—to create better outcomes for everyone.
In short, while we must firmly maintain individual responsibility for actions like crimes, the insights provided by big data about collective human behavior are valuable and essential. They should inform decisions and interventions at the societal level, complementing—but never overriding—our fundamental concepts of personal accountability.

Discovery:
I used to interpret big data as a by product of Internet and statistic method, however now I realized that Big data is a natural next step when trying to think about problems through a collective perspective.

[malongo03]

From my perspective, social law and statistical prediction of human behavior are completely irrelevant to the existence of free will. This is because there both exist unpredictable systems that are nevertheless entirely deterministic and predictable scenarios that are entirely free.

The case of unpredictable deterministic systems is covered by chaos theory, where the past determines the future but the approximate past does not approximately determine the future. While we have made great strides in understanding chaotic dynamic systems, we will always be fundamentally limited in our predictions by the finite accuracy of our measurements. In addition, at least according to some interpretations of quantum mechanics, the behavior and evolution of systems of subatomic particles is also entirely deterministic even if it is proven that it is impossible to predict with total certainty the result of any arbitrary measurement.

The case of highly accurate prediction of complex human behavior is certainly spooky, but if we strip away the ambition and remind ourselves what is really going on, it becomes obvious that spooky does not mean deterministic nor does it invalidate conscious decision making. Imagine you're a social scientist. You watch my schedule and discover at 2pm on Thursday, I almost always attend this class. So you then predict that today at 2pm, I am most likely going to go to class today. You tell the professor, and therefore he marks a 0 for my attendance because this prediction means I “did not choose” to attend class, my presence is merely the result of probability. This is a ridiculous assertion! The prediction of my behavior has nothing to do with the existence of my conscious choice to attend class! Indeed, even if the prediction was told to me, this does not mean my legs suddenly begin to propel themselves forward and drag me to class against my will. I am still the one in full control of my decision.

These are simple examples, but I think they are sufficient to demonstrate how free will is detached from predictability. Indeed, I believe free will is detached from even complete foreknowledge, but that is a different conversation.

Discovery:
I was always wondering where the "heads tall" rules for drawing proportions that my father taught me came from, Quetelet gave me a very good explanation of how we derive this data empirically nowadays, versus how they did in antiquity relying on pure vibes.

[Alex7692]

The conjectural reality of "social law" implicates the predictability of crimes. It therefore seems possible to establish preemptive measures against crimes by identifying culpability ahead of time. Without diving too deep into the philosophical weeds, suppose we grant a Sartrean conception of "freedom" such that humans are "condemned to be free", to the extent that we nonetheless "choose" our own projects as dissatisfaction grows upon the state of affairs, the full-range capacity of free will remains intact. But as Sartre would phrase it, the only imposable limit upon "free will" is the "free will" of the Other. Insofar as intersubjectivity (our being-for-others) constitutes the limits of our freedom, it would seem to make room for the objective (or in risk of tautology, "nomological") facticity of social laws. To that end, freedom implies our full responsibility for the actions we take. But then the question comes to whether it is righteous to restrict that open possibility. Needless to emphasize that intervention itself would disrupt the existing social patterns in a way that any jurisdiction based on statistical norms may be self-defeating.

I'd therefore construe the issue here not as an ontological problem, but a hermeneutical (or epistemological) obstacle. Should we regard statistical data as explanatory as such, or merely descriptive without necessarily informing us about a true essence of regularity (and to that extent, we may as well be Humeans in refutation against causality as such)? Is "l'homme moyen" a natural kind or a constructed category for demographical purposes (to that Foucault would answer: it's the latter as a feature of the modern episteme)? To the extent that normativity / lawhood is itself constituted by paradigmatic shifts in our understanding of sociality, the perception of deviant behaviors is a contingent one. It may be useful to refer to them as "social law", but no "perfectly natural law" if that entails the hypostatization of human proclivity to do certain things in certain way (recall the Galton's synthesis of statistics with Darwinism; but also recall the use of computation in evolutionary biology that demonstrated the potential reducibility of macroevolution to random walk). As Gigenrezer suggests, there is an intimate link between the emergence of statistical thinking and 19th century liberalism. It is often more useful to hone in on the discrepancies and irregularities that is exhibited by individual agents that the wider social patterns supervene upon. Note that this does not entail the application of statistical results for the purpose of "rectification", but to gain a better understanding of why irrational behaviors exist in society in a rationally interpretable way.

Discovery

For my exploration I've read F. Y. Edgeworth's "The Element of Chance in Competitive Examinations", published in 1890 on Journal to the Royal Statistical Society. If I've understood it correctly, this "element of chance" is construed broadly in the article that consists in 1. The set of examiners who would grade the student's paper (assuming that it's referring to the differential standards / rigor of examiners); 2. The exact distribution marks at different examinations which influences the honor line (which we could perhaps understand by the more Americanized notion of "curving"); 3. the turnout of secondary (less important but still count as aggregate mark) subjects during an exam. To be honest, I am not sure what result of significance has come out of the analysis besides a substantiation of the factor of uncertainty in competitive examinations. (Or perhaps having derived a "safety/hedge distance", that this would be the important result?) On a more methodological note, it was interesting to see how the statistical instruments employed in the article are still fairly rudimentary: there's consideration over average, mean deviation (i'm not sure what would be its connection with SD?), square errors, and so forth, but none of any stronger tools were available to the author. It seems a lot more difficult to compute or even make clear sense of chance as opposed to identifying mere regularity.

[ucachyut]

To evaluate whether individuals possess free will in the context of statistical social laws, we must begin by defining free will itself. For this purpose, it is most useful to understand free will as an individual’s capacity for probabilistic action. Whether that action arises from genuine randomness or from hidden deterministic variables is ultimately irrelevant when analyzed within the broader context of society. This definition allows for the coexistence of free will and statistical regularity—what Adolphe Quetelet described as “social law.”

Social statistics, such as consistent rates of crimes like robbery or suicide, reveal patterns at the population level. However, these patterns do not imply that any particular individual’s actions are predetermined. Just as tossing a fair coin a thousand times yields roughly equal heads and tails without influencing the outcome of any single toss, so too can social data yield regularities without undermining individual autonomy. Each coin flip retains its probability of 1/2, regardless of past results. Likewise, each person retains their capacity for action, even within broader patterns of behavior. In this sense, free will is preserved as the freedom to act probabilistically, despite large-scale statistical consistency.

If one were to argue that crime is a manifestation of natural law, as Quetelet implied in his analogy between physical and moral regularity, it would logically follow that the punishment of crime is equally part of this statistical system. Over history, holding individuals accountable for deviant actions has itself been a consistent societal response. Therefore, punishment is not contradictory to the statistical framing of behavior, but integral to it. Accountability becomes not a violation of natural law, but part of the social mechanisms that produce and maintain it.

Furthermore, while social scientists can compute trends at a macro level, it remains practically impossible to predict an individual’s exact decision. As Gigerenzer and Porter emphasize, the probabilistic models of early statistics were never meant to yield certainty about individual behavior, only expectations at the aggregate level. This gap reflects the same uncertainty we find in flipping a single coin. Though the outcomes of large-scale behavior may seem deterministic from a distance—like an alien observing human society as its own object—the individual remains a site of unpredictability and freedom.

That said, the prospect remains open that individual behavior might one day be modeled through finer-grained probabilistic structures—perhaps by analyzing humans at the level of neural, cellular, or even molecular data. Whether this would reduce an individual to a large scale statistical model of cells or offer new tools for understanding human freedom remains a key question.

Discovery
One illuminating moment came in Quetelet’s assertion that "the laws which regulate the course of phenomena in the moral order, are as rigorous as those which govern the physical world" (Treatise on Man, p. vi). This early fusion of physics and social science reveals the origin of the idea that human behavior is computable, a core assumption behind much of contemporary artificial intelligence. It suggests that even in the 19th century, thinkers were already imagining society in ways that resonate with today’s predictive algorithms.

[benkim04]

Social law approaches human behavior in an interesting way in that it suggests that human behavior is dictated by a social norm that we tend to follow, with noise that causes us to deviate from the norm. This suggests that humans are actually just computers, in that we are a product of societal calculations and that our exercise of free will is actually just an exercise in rehearsing set standards. At the same time, it therefore suggests that any wrongdoing committed by individuals was bound to happen, or in some cases inevitable. That action, such as murder or littering, was part of the social calculus and suggests that the societal norm is built to include such behavior. That is, it is not the fault of the individual, but rather the fault of society for allowing a situation in which that behavior is validated. In this way, the theory of social law is both diminutive of human will while also being extremely generous towards deviation. I think it is improper to suggest that all of human behavior is calculable based upon a societal standard. Society is dynamic and its standards are changing constantly, whether decade to decade or year to year. Further, there is a sort of assumption that human rationality is dictated by reference to some societal benchmark. This is often not the case, as quite plainly many humans are selfish and would deviate from a societal benchmark should it be beneficial for them. When approaching the culpability of criminals, we can examine the contrasting case: are individuals who overachieve, generate scientific breakthroughs, are exceedingly generous, and generate art of high order responsible for their actions? Or are they again part of the larger social calculus? This case seems a little bit harder to determine. We often attribute factors like hard-work, tenacity, or sheer talent to such individuals. That is, features that make those individuals stand out from others. Furthermore, such individuals often come from backgrounds that enable such behavior. That is, they are able to positively deviate from the societal norm because they are given the possibility to do so. We can then return to the case of criminals. They certainly have committed feats that make individuals stand out from each other, but most probably because they come from backgrounds that not only enable such behavior, but necessitate such behaviour due to a lack of resources. That is to say, circumstance is most likely the biggest factor when determining human behavior, not an abstract societal norm. Still, it is helpful to think about punishment through the lens of social law, in that we must move the societal norm to erase situations that beget criminal behavior.

Discovery:
At the moment I am taking a class around neoliberalism and political theory, and social law has close ties with many of the ideas. Neoliberalism suggests that the market is the great societal calculator, and that there are natural laws that are “the rules to the game” that everyone must abide by should they wish to be players in the market. Further, it suggests that the market is run by not only winners, but losers. This ties in with social law in that the existence of losers, perhaps criminals, are vital to the system, and therefore may be intrinsic in the societal norm. Neoliberalism then also corroborates the existence of a social norm, that being the natural laws that people abide by. Finally, all individuals, whether they are winners or losers of the market, are still a product of the market and the calculator, suggesting that we ourselves cannot calculate outcomes or human behavior, but the market can.

[meeradasgupta247]

On Numbers and Notions
Meera Dasgupta

“Everything happens for a reason” is a phrase which brings comfort to many. After all, the ability to rationalize and compartmentalize the world into universals allows one to reconcile with the many unknowable things which they may come across in life. In particular, the readings for this week further this concept by exemplifying how, through the examination of group data clusters, or averages, individuals have come to have established an illustration of the normal person and, in turn, natural laws governing the human population. Though these averages are helpful to understand vast amounts of data, it is clear that they are not as objective as they are advertised to be - given that many aspects of data collection and analysis may be subject to human biases or error. More specifically, the universalization of the human condition not only dismisses these biases but fails to consider how outliers can be evidenced against the universal solely by means of existing. Hence, though statistics can illustrate big picture ideas, they cannot make generalizations about moral or natural law through numerical analysis.

In “The Statistical Probabilists,” by Gerd Gigerenzer, it begins by explaining how faith in statistics arose from the rule that it must “exclude all opinions” (p. 38). However, this reasoning seems to reside in the idea that numbers are unable to take on a subjective lens - that they are somehow beyond human construction. Still, how did numbers come to be if not for people? Perhaps the numbers themselves may be difficult to argue when situated within a properly acknowledged context but the analysis of numbers is subject to interpretation; after all, even the law of large numbers is interpreted to hold some greater meaning about natural law in the same text. The methodology of accumulating large amounts of data to better understand people is even flawed in the way that, according to the Treatise on Man and the Development of his faculties, by Adolph Quetelet, he claims that with “every thing being equal … we approach closer to truth” (p. 7). Yet, there is no way to ascertain that everything is equal when something such as social categorization is not numerical.

In fact, with all of this, the primary issue here still seems to be not the numbers themselves but rather the use of the objective as undeniable in order to make claims about what is right or natural. Though Quetelet in The Rise of Statistical Thinking, by Theodore Porter, seems to generally dislike those who utilize statistics to prove their beliefs, he is still deterministic in the way that he views individual actions to be ruled by uncontrollable forces which dictate the group. In fact, his observations lead him to believe in the “perfectibility of the human race” (p. 104) where “defects and monstrosities disappear more and more” (p. 104). This goes to show that by assigning outliers in his studies a moral qualifier, Quetelet too falls privy to his own conceptions of the masses and moral righteousness. By constructing fact, especially by relating statistics to the more established physical sciences of the time, Quetelet is able to conclusively create human fictions for the security of progress without change.

Discovery/Exploration

I decided to read “The Income and Property Tax” by William Farr in the Journal of the Statistical Society of London. This was interesting to me in the sense that an equal distribution of taxes in relation to the proportion of the property owned and also in relation to the time of collection was posed as the most fair route of collection. This was reasoned as also being the mark of intelligent societies which were more protected from anarchy or upheaval (p.17) - where taxation was a matter of justice. However, despite the calculations made, I wonder how effective these taxes were when applied, especially given that justice is a subjective consequence.

Citations

A. Quetelet (1835). Selections from Treatise on Man and the Development of his faculties. Preface, Introductory, Development of the physical qualities of man, Development of the Moral and Intellectual qualities of Man: pp. v-x, 5-12, 72-75,

Gigerenzer, Gerd, ed. The Empire of Chance: How Probability Changed Science and Everyday Life. Ideas in Context. Cambridge: Cambridge University Press, 1989, Sections 2-2.2 (“The Statistical Probabilists” and Chapter 2's conclusion): 37-45/68-69.

Porter, Theodore. The Rise of Statistical Thinking, 1820-1900. Princeton, N.J.: Princeton University Press, 1986, chap. 2 (40-70)

Farr, William. “The Income and Property Tax.” Journal of the Statistical Society of London 16, no. 1 (1853): 1–44. https://doi.org/10.2307/2338101.

[jpprieto]

There are many ways that we can give interpretations to the statistical explanatory substrata that emerges when we analyze population at a big scale. Quetelet and other early statisticians seemed to give importance to the means of the patterns, seeing the deviations from it as aberrations of the system. While as statistics developed, we saw that study of the variation patterns themselves prove more fruitful for analysis of phenomena both physical and social.

I find quite interesting that this sparked philosophical implications for free will. Perhaps it was precisely due to this framing of the tendency to the “average man”. However, I resonate more with the sociopolitical implementation of this. Even in a system that exhibits a distribution, it is not that we are each individually pushed to conform to such a curve. Instead, how I see it, the exact conditions of the system exhibit such phenomenology as to center around a response in aggregate, but there is a choice on where in the spectrum one falls. It is not that the curve generates the human response, but in my opinion rather than the choices generate the curve. It would be interesting to explain the psychology of this, when faced with an environmental condition, why does an average human brain tend to follow the central tendency?

To say that this takes culpability or free will seems erroneous. Within the environment, a certain person still chose to commit a crime or buy a pack of cookies. While I think that there still is personal responsibility, it is indeed important to see how systemic conditions can cause these events. Here I think it may be useful to use crime as a case-study. I believe that we indeed should try and simulate possible responses, both statistically from historical data and analytically through psychosocial investigation.

There will always be outliers, but if we talk about crime, we would want a system that would reduce the tendency of the population to extremes of behavior, to minimize the standard deviation of behavior. A very relevant matter is not only to think about criminality as a one time event, but also how it emerges across the lifespan of individuals. Right now, especially with a system that is so poorly designed for rehabilitation, we may have people that once they are labelled “criminals” they often continue to commit crime. But currently these people suffer diminished opportunities because of the label, affected experience and opportunities due to incarceration, and an effective criminal network through prison connections. Furthermore, prison is a school for crime. You learn about different crimes, about how others got caught; upon release, if their intelligence may before have been used for high-sophistication career paths, now with a network and knowledge the individual may see a low-friction possibility of criminal enterprise. Which when combined with the fact their effective life-time got reduced by imprisonment, is made more attractive due to its high-risk but high-return model. As such, I would argue, we are generating a distribution in which the mean is systematically biased to continue committing crime, and here ironically the outliers, the data points which require more energy to execute, is that of not committing further crimes.

Discovery: I enjoyed interacting with the code for statistical analysis. Personally I sometimes struggle with pure mathematics but enjoy deeply interacting with the frameworks through code or applied problems. This modern counterpoint to the historical reading of statistics brings to light how we went from the table analysis of Quetelet, where we compared means, to the powerful regression and pattern modeling we have nowadays. It may be easy to judge the interepretation of past early statisticians, who seemed to focus so heavily on the means. But without the genius of LaPlace or Fourier, it would take significant work to be able to plot the graphs or make the comparisons we do so easily nowadays. We see this echo in a very meta way, as there are central tendencies to ideology too based on historical context!

Link to notebook: https://colab.research.google.com/drive/1sBH4HScRFre7irn1Io3-UNBIKG-TZQ2f?usp=sharing

[peterzhou-cs]

When Quetelet discovered the constant patterns hiding behind the randomness and chaos of human individuals by performing large-scale statistics, he truly had opened a new page in both human intellegence as well as their social order. The idea of gathering complicated individuals together to balance out the randomness is ingenius, and is a genuinely innovative viewpoint that has profound applications to most disciplines. However, when the subject of study becomes human, the discerning power of statistics makes us uneasy: if human's behavior, even the most extreme ones such as suicide or murder, occurs a constant number of times in one society, the concept of free-will is shakened. Extreme viewpoints soon emerged, that human has no free will and the course of social affairs cannot be interrupted by man or even by God. (Bucker) Quetelet himself holds a milder opinion and attempts to leave a minimum bubble of free will for each individual.

The above seemingly metaphysical debate is crucial to human race. If we were to believe in Bulker, we would have ceased to participate in social affairs and become indifferent. In higher generality, once states are organized to collect enough data (as it was around the beginning of 19th century), it may use them to create common belief as numbers are manifestly objective and convincing, whereas common belief dominates our society. here are far more numbers in statistics than that appeared in other sciences before, which went beyond human's intuition and one must computes some values to comprehend the numbers. The comprehension and interpretation of these numbers turns out to be subjective and demands a whole subject of study. The first quantity being studied is the mean, or the 'average man', which is the outcome of balancing out the 'randomness'. It was then applied in Astronomy and mecidine. Much later was the concept of variation introduced, and scientists awared that the randomness is more than noises, but reflecting deep structure in the dataset. Then we introduced probability into the statistics and gained even more power: we can infer from the current data as if seeing the future from a crystal ball.

It is fascinating to recognize that not only technological advancements --- like the invention of stirrup or gunpowder --- or the rise and fall of great empires, pins down a markingstone or a turning point in human history. The emergence of some knowledge and perspectives alone can have a profound impact: we benefits immensely from its outcome in technology as well as social organization, without which it would be impossible to envision an organized society of billions. On the other hand, it exposed the deeper structure/order of humans to those who owned the data, and presenting a theoretical and practical threat that leads to a totalitarian society.

Exploration:
The colab exercises are pretty. Basic intepretations in Stats are so automated in the modern times.

[rshah1869]

The question of whether statistical regularity in social behavior undermines individual culpability and free will was central to nineteenth-century debates over the rise of statistics and social physics. Quetelet’s application of statistical averages as a framework for understanding human behavior helped him develop the concept of the “average man.” To Quetelet, the only way of understanding society was by observing “incontrovertible facts furnished by statistical data.” Across different areas such as crime, suicide, and marriage there was a striking and somewhat unsettling pattern of regularity. These acts, which seemed deeply personal and volitional, occurred year after year with near‑constant frequency. Quetelet interpreted these not as random acts committed by free individuals but as expressions of underlying social laws.

Theodore Porter notes that while Quetelet was an important thinker, his social physics was ultimately unworkable as a theory of society because of its determinism. Quetelet’s view assumed that individuals were so deeply shaped by their social environment that their actions could be predicted with the same certainty as planetary motion. However, he did not believe these patterns were immutable. Instead, he thought that by identifying the stable statistical tendencies within a population, reformers could rationally guide society toward improvement. His work thus opened the door to a new form of governance based on statistical insight, an early vision of social engineering.

Critics like Moritz Wilhelm Drobisch rejected Quetelet’s generalizations. They argued that statistical regularity in crime did not imply a universal human propensity to criminality but instead reflected the behavior of specific subpopulations, often marginalized or socially disadvantaged. Assigning a penchant for crime to the so‑called average man was both morally and empirically problematic. Émile Durkheim, one of sociology’s founding figures, took particular issue with Quetelet’s approach. Durkheim’s interest in suicide was in part an effort to disprove Quetelet’s statistical determinism. Rather than relying solely on the regularity of outcomes, Durkheim sought to understand the social conditions such as anomie, integration, and regulation that made individuals more or less likely to take their own lives. As a result, social theory evolved away from the search for deterministic laws and toward an analysis of structures, norms, and institutions.

Discovery
While coding the PDFs and CDFs of the given data and fitting them to different models I was struck by how intuitive and powerful this statistical approach becomes. It made me wonder what Quetelet would have thought of the advancement of statistical inference, particularly our ability now to fit models, evaluate residuals, and estimate uncertainty. Would he have embraced computational methods or worried about the detachment from direct observation? It also made me reflect on methodology. When measuring human behavior do ethnographic observations carry less weight than statistical inference? Can numbers alone tell the full story of society or do they require interpretation through lived experience and context?

[JoNeedsSleep]

(1) Free will can only be spoken at a certain scale. If the universe is governed by fundamental laws at the particle level, then in the physical sense we don't have free will. If you zoom out and look at historical trends where "large-scale regularities [is produced from] chaotic individual events" [1], we are merely one dot in some distribution. But we feel we have free will, and that's because perturbations in our will--motivation, beliefs, emotional states, memory--does have an observable impact on our lives.
(2) London Statistical Society took its motto as "Aliis exterendum - to be threshed out by others" [2]. Quetelet tried to find social laws of nature. Both assume an air of objectivity - and yet how you measure data, which set of variables you look at, baked in assumptions after assumptions, all of which can't be developed without the environment. I've been taking a causal inference class where the goal is explicitly to analyze past data under a causal framework (without a RCT usually) to guide policy. We have an a priori causal model which derives a non-unique statistical model, which then admits of a unique factorization. Importantly, we can falsify that philosophical model, but we can't in principle verify it. We take it by faith. The French social economist J.-B. Say criticized that "statistics was merely descriptive, and could offer little to the savant who sought to identify underlying laws." [3] I think statistics rose to its status today partly because it integrated the role of the philosopher, social economist, or political economist.

Discovery: it was cool to work on the google collab and see how different functions like pdf, pmf shine different lights to the same set of data. It's really all about interpretation.
[1] Gigerenzer, Gerd, ed. The Empire of Chance: How Probability Changed Science and Everyday Life. Ideas in Context. Cambridge: Cambridge University Press, 1989, Sections 2-2.2 (“The Statistical Probabilists” and Chapter 2's conclusion): pg 44
[2] Gigerenzer, Gerd, ed. The Empire of Chance: How Probability Changed Science and Everyday Life. Ideas in Context. Cambridge: Cambridge University Press, 1989, Sections 2-2.2 (“The Statistical Probabilists” and Chapter 2's conclusion): pg 38
[3] pg 46

[TonyYZ]

Response
Sociologists from the 19th century, like Quetelet, used means as an efficient tool to analyze features of social groups, from appearance to behaviors and even personalities. While means is a useful and concise scale for capturing the overall features of a population under investigation, their understanding of statistics puts itself in a paradox around factuality. While the factuality of statistics resists the urge to analyze “causes or effects,” which weakens statistics’ explanatory power, it is also difficult to fully preserve factuality when statistics is employed for analysis and counseling in practical matters (Gerd 38).
On the one hand, the factuality of statistics dismisses the deeper meaning of independent events. While statistics is applicable to stable features like physical characteristics, its analysis of classes of events could be clumsy, given the uniqueness of each individual event. Take writing creative works as an example. Quetelet compares author creativity with fecundity, to which he succeeds in applying the statistical tools, “The number of works can alone give an idea of the productive power of the author, as the number of children brought into the world gives us the fecundity of a female, without taking into account the value of the work produced” (73). His deliberate omission of the work’s quality probably stems from the difficulty of measuring this dimension in his age. Nevertheless, even if the contemporary concept of papers’ impact factor is considered, the work’s significance is embodied not just through citations but also through its relationship with the world. In this sense, each work has to be evaluated in its context, but means would efface this layer of meaning by flattening the author’s works into homogenous numbers. The preference for the general over the particular deprives the events of their multifaceted contributions to other events in the world.
On the other hand, statistics cannot fully reach factuality when it becomes practical. In Quetelet’s vision, factuality is supposed “to exclude all opinions” and only present the reality (Gerd 38). While he desired to “renounce analysis almost entirely,” the attempt to use statistics for state governance and “guidance of legislation” involves interpretation and analysis to some extent, or data would be pure numbers (Gerd 42). When Quetelet extracts data like crime rates and birth rates, the data is not ready in the world of people. Instead, it is produced by bureaucratic organizations through the classification of social groups, which implies assumptions about how the population should be divided. Therefore, data is a partial presentation of the population and could be manipulated for various purposes. For example, considering “crime [as] an attribute of the social type and not of individual deviants” could justify the government’s separation of the criminals from the population and open the space for stigmatization and misunderstanding (Porter 109). In other words, statistics’ participation in governmental affairs inevitably brings assumptions and biases in order to deduce policies from the data.

Discovery
On the field trip to the special collections on Thursday, I was drawn by the different mathematical notations used by the books. While some books seem to have been published later and were adapted to the modern mathematical notation style, older books like Kepler’s Astronomia Nova use Arabic numerals and an unfamiliar notation of vertical calculation. However, the 1308 Ars Brevis used purely the Latin alphabet for its symbols to fill in the wheels and tables. I wonder if the different notations that organize the symbols in various layouts influence these authors’ approach to computation, or is it the opposite direction of influence?

Referenced Images