Home /Modules/stanford-math/EQUATIONS-SOCIAL.md

EQUATIONS-SOCIAL.md

stanford-math/EQUATIONS-SOCIAL.md · 198 line(s) · markdown

Social-Structure Canon

34 equations, sourced from SMSocialCanon.sleela -- the single source of truth. This document is a human-readable mirror of that record; if the two ever disagree, the .sleela canon wins and this file is stale.

How social ties and social structure are actually measured in the social sciences -- network connection (density, clustering, centrality), the strength and overlap of ties (Granovetter), social capital (bonding vs bridging), structural holes, the human scale of stable relationships (Dunbar), the shape of real networks (small-world, scale-free, community structure), and the distribution of wealth (Gini, Lorenz, Pareto). These measures describe STRUCTURE -- ties and distributions -- and rank no person. Each row credits the scholar and year as the record gives them, with a one-line note on why it matters.

The one hard rule governs every row: values are faithful to the subject's real origin where publicly known; diffuse or unknown attributions are marked honestly (unknown, classical, attrib.), never invented to look legitimate. A standing boundary for this canon: these are measures of social structure -- network ties and statistical distributions -- not scores or rankings of individual people. Symbolic forms use plain-ASCII notation so they stay greppable and diff-friendly. The Why it matters column states each result's significance in one faithful line.

The 34 equations

#EquationFormFieldDiscoverer(s)YearWhy it matters
1Network densityD = 2 L / (N (N - 1))social network analysisclassical graph theoryunknownThe fraction of possible ties that actually exist; the simplest measure of how connected a group is.
2Local clustering coefficientC_i = 2 e_i / (k_i (k_i - 1))social network analysisWatts & Strogatz1998How tightly a person's contacts know each other; captures close-knit, 'everyone-knows-everyone' structure.
3Degree centralityC_D(v) = deg(v) / (N - 1)social network analysisLinton Freeman1979How many direct ties a node has; the most basic measure of how connected an individual position is.
4Betweenness centralityC_B(v) = sum_{s != v != t} sigma_st(v) / sigma_stsocial network analysisLinton Freeman1977How often a node sits on the shortest path between others; identifies brokers and bridges.
5Tie strength (strength of weak ties)strength ~ time + intimacy + reciprocity; weak ties bridge groupseconomic sociologyMark Granovetter1973Granovetter's insight that weak ties (acquaintances) reach new information strong ties cannot -- key to how jobs and ideas spread.
6Dyadic embeddedness (tie overlap)overlap(a,b) = |N(a) cap N(b)| / |N(a) cup N(b)|social network analysisGranovetter (overlap principle)1973The more two people's networks overlap, the stronger their tie; the structural basis of the weak-ties argument.
7Homophily indexH = (same-group ties - expected) / (total - expected)sociologyMcPherson; Smith-Lovin; Cook2001Measures the tendency for similar people to connect ('birds of a feather'); shapes the structure of every social network.
8Social capital (bridging vs bonding)returns ~ f(bonding within-group, bridging across-group)sociologyPutnam; Coleman; Bourdieu2000The value in social ties: bonding ties hold a group together, bridging ties link groups -- both drive collective action.
9Structural holes (constraint)C_i = sum_j (p_ij + sum_q p_iq p_qj)^2economic sociologyRonald Burt1992Burt's measure of the advantage of bridging gaps between otherwise-disconnected groups; a source of opportunity and influence.
10Dunbar's numberN_stable ~ 150 (cognitive limit on stable ties)evolutionary anthropologyRobin Dunbar1992A cognitive ceiling of roughly 150 stable relationships; why communities and units tend to a human scale.
11Preferential attachment (scale-free degree)P(k) ~ k^(-gamma)network scienceBarabasi & Albert1999'The rich get richer' in ties -- explains why a few hubs dominate real social and web networks.
12Small-world path lengthL ~ ln(N) / ln(k)network scienceWatts & Strogatz1998Why any two people are linked by only a few steps ('six degrees'); short paths despite local clustering.
13Gini coefficient (wealth inequality)G = (sum_i sum_j |x_i - x_j|) / (2 N^2 mu)economics of distributionCorrado Gini1912The standard single-number measure of how unequally wealth or income is distributed (0 = equal, 1 = all held by one).
14Lorenz curveL(p) = (cumulative wealth of poorest p fraction)economics of distributionMax O. Lorenz1905Plots cumulative share of wealth against share of population; the visual from which the Gini coefficient is computed.
15Pareto (wealth) distributionP(X > x) ~ (x_m / x)^alphaeconomics of distributionVilfredo Pareto1896The heavy-tailed law behind the '80-20 rule' of wealth; the original model of concentration at the top.
16Reciprocity (directed ties)r = (mutual dyads) / (total ties)social network analysisstandard SNA measureunknownHow often ties are mutual rather than one-way; a basic indicator of balance and cohesion in a network.
17Modularity (community structure)Q = sum_c (e_cc - a_c^2)network scienceNewman & Girvan2004Measures how cleanly a network splits into communities; the basis of community-detection methods.
18Assortativity coefficientr = Pearson correlation of degrees across edgesnetwork scienceM. E. J. Newman2002Whether well-connected nodes link to other well-connected nodes; distinguishes social networks from technological ones.
19Eigenvector centralityx = (1/lambda) A xnetwork sciencePhillip Bonacich1972Importance proportional to the importance of neighbors; a foundational network measure.
20PageRankPR = (1-d)/N + d sum PR(j)/outdeg(j)network sciencePage; Brin1998Ranks nodes by a random surfer's visit frequency; the algorithm behind web search.
21Katz centralityx = (I - alpha A)^(-1) . 1network scienceLeo Katz1953Credits all paths with distance-decay weighting; a classic influence measure.
22Closeness centralityC(v) = (n - 1) / sum_u d(v,u)network scienceGert Sabidussi1966Importance by proximity to all others; measures efficient reach in a network.
23Bonacich power centralityc = alpha (I - beta A)^(-1) A . 1network sciencePhillip Bonacich1987Power depends on connections to the less-connected; captures bargaining advantage.
24Freeman degree centralizationC = sum (c_max - c_i) / max possiblenetwork scienceLinton Freeman1979How unequal a network's degree distribution is; a measure of centralization.
25Shannon diversity indexH = -sum p_i ln p_isocial measurementClaude Shannon1948Information-entropy measure of variety and evenness; widely used across the social and life sciences.
26Simpson diversity indexD = 1 - sum p_i^2social measurementEdward H. Simpson1949The probability two random members differ; a standard diversity measure.
27Theil index (inequality)T = (1/N) sum (x_i/x-bar) ln(x_i/x-bar)social measurementHenri Theil1967An entropy-based, decomposable inequality measure; separates within- and between-group inequality.
28Atkinson inequality indexA = 1 - (1/x-bar)(mean of x^(1-e))^(1/(1-e))social measurementAnthony Atkinson1970Inequality with an explicit inequality-aversion parameter; grounded in social welfare.
29Zipf's law (city sizes)size ~ 1 / rankurban economicsGeorge Kingsley Zipf1949City size falls inversely with rank; a striking regularity of human settlement.
30Gravity model of migrationF_ij = G P_i P_j / d_ij^2spatial social scienceErnst Georg Ravenstein1885Flows grow with sizes and fall with distance; the base model of migration and trade.
31Benford's law (leading digit)P(d) = log10(1 + 1/d)social statisticsBenford; Newcomb1938Leading digits follow a logarithmic distribution; used to detect fabricated data.
32Sex ratio at birth (Fisher principle)parental investment drives a 1:1 ratiosociobiologyRonald A. Fisher1930Equal parental investment drives a balanced sex ratio; a classic of evolutionary reasoning.
33Metcalfe's law (network value)value ~ n^2network economicsRobert Metcalfe1980A network's value scales with the square of its users; a heuristic of network economics.
34Granovetter threshold modelcollective action occurs above an activation thresholdcollective behaviorMark Granovetter1978Collective action ignites once enough others act; explains riots, fads, and adoption.

Validate

These equations are compiled, not just documented. Build the source compiler, then check and run the sibling demo (which prints every entry and the running total across all canons):

SLEELA_LIB="$PWD/stanford-math" impl/build/sleela check stanford-math/SMArchive.sleela
SLEELA_LIB="$PWD/stanford-math" impl/build/sleela run   stanford-math/SMArchive.sleela

Sources & attribution

Names, forms, discoverers, and years follow the standard social-science / network-science record; the Why it matters notes summarize each result's established significance. Key anchors: Granovetter (1973), Freeman (1977-79), Burt (1992), McPherson-Smith-Lovin-Cook (2001), Watts-Strogatz (1998), Barabasi-Albert (1999), Newman-Girvan (2004), Gini (1912), Lorenz (1905), Pareto (1896). See SMProvenance.sleela and the repo README.md. This archive records and attributes published results; it does not originate, own, or certify them.

Definitions & relevance (second-layer additions)

The entries below are the second-layer additions to this canon, organized by type. Each gives a short definition (what the measure is and what its symbols mean) and its relevance (why it is foundational to the study of social structure). As everywhere in this archive, these describe structure and distribution — not persons; they rank positions and quantify inequality, never individuals.

Network centrality

Eigenvector centrality — x = (1/lambda) A x. A node's importance is set proportional to the summed importance of the nodes it connects to, so influence flows recursively through the network rather than being counted one hop at a time. It is foundational because it captures the intuition that being connected to well-connected others matters more than raw tie count; it is the mathematical ancestor of PageRank and a standard tool for finding structurally influential positions.

PageRank — PR = (1-d)/N + d sum PR(j)/outdeg(j). The stationary distribution of a random surfer who follows links with probability d and jumps to a random node otherwise; a node's score is the long-run fraction of time the surfer spends there. It is relevant far beyond the web: as a damped, teleporting eigenvector centrality it ranks importance in any directed network, from citation graphs to transport systems.

Katz centrality — x = (I - alpha A)^(-1) . 1. Credits a node for every path that reaches it, discounting longer paths by a factor alpha per step, so distant connections still contribute but with diminishing weight. It matters because it bridges local degree and global eigenvector centrality, and it stays well-defined on directed, acyclic structures where plain eigenvector centrality can fail.

Closeness centrality — C(v) = (n - 1) / sum_u d(v,u). The inverse of a node's average shortest-path distance to all others, so a high score means the node can reach the rest of the network in few steps. It is the natural measure of reach and efficiency — how quickly information or influence originating at a position can spread to everyone else.

Bonacich power centrality — c = alpha (I - beta A)^(-1) A . 1. Generalizes eigenvector centrality with a parameter beta whose sign decides whether being tied to well-connected others is an advantage (positive) or, in bargaining, a disadvantage (negative). Its relevance is that it formalizes the counterintuitive idea that power can come from connecting to the weakly connected — central to exchange and negotiation theory.

Freeman degree centralization — C = sum (c_max - c_i) / (max possible). A graph-level measure of how unequally centrality is distributed, comparing the most central node to all the rest against the most star-like possible network. It matters because it moves the analysis from individual positions to the whole structure, quantifying whether a group is organized around a hub or spread evenly.

Diversity & inequality measurement

Shannon diversity index — H = -sum p_i ln p_i. Borrows information entropy to score a population's variety: it is maximized when categories are many and evenly represented, and zero when one category dominates. It is a foundational, field-crossing measure of heterogeneity, used from ecology to the diversity of occupations or opinions in a social group.

Simpson diversity index — D = 1 - sum p_i^2. The probability that two members drawn at random fall in different categories, so higher means more mixed. Its relevance is as an intuitive, bounded companion to Shannon entropy that weights common categories more heavily — a standard concentration/diversity gauge.

Theil index (inequality) — T = (1/N) sum (x_i/x-bar) ln(x_i/x-bar). An entropy-based measure of inequality whose signature property is decomposability: total inequality splits exactly into a within-group part and a between-group part. That makes it foundational for attributing inequality to structure — how much of it lies between regions, sectors, or classes versus within them.

Atkinson inequality index — A = 1 - (1/x-bar)(mean of x^(1-e))^(1/(1-e)). Measures inequality with an explicit inequality-aversion parameter e, so the analyst states how much weight to place on the bottom of the distribution. It matters because it grounds an inequality number in social-welfare theory rather than in a purely descriptive statistic.

Distributional & spatial laws

Zipf's law (city sizes) — size ~ 1 / rank. The population of the k-th largest city is roughly the largest city's size divided by k, giving a strikingly regular rank–size curve. It is relevant as one of the most robust empirical regularities in the social sciences, constraining theories of how settlements and firms grow.

Gravity model of migration — F_ij = G P_i P_j / d_ij^2. Flows between two places rise with the product of their sizes and fall with the distance between them, by analogy with Newtonian gravity. It is the workhorse model of migration and trade, still the baseline against which spatial-interaction theories are tested.

Benford's law (leading digit) — P(d) = log10(1 + 1/d). In many naturally-occurring datasets the leading digit is 1 far more often than 9, following a logarithmic law. Its relevance is both theoretical (scale invariance) and practical: deviations from it are a standard red flag in fraud and data-integrity auditing.

Sociobiology & network economics

Sex ratio at birth (Fisher principle) — equal parental investment drives the population toward a 1:1 ratio. Fisher's argument is that whenever one sex is rarer, parents producing it gain a reproductive advantage until balance returns. It matters as an early, elegant equilibrium argument that seeded evolutionary game theory and the economics of investment under frequency dependence.

Metcalfe's law (network value) — value ~ n^2. The usefulness of a communication network grows roughly with the square of its users, because the number of possible pairwise connections does. It is relevant as a heuristic for network effects and critical mass, widely invoked (and debated) in the economics of platforms and standards.

Collective behavior

Granovetter threshold model — collective action ignites once the fraction of others already acting crosses an individual's activation threshold. Because thresholds vary across people, a population's distribution of thresholds — not any single person — determines whether a riot, fad, or adoption cascade takes off. It is foundational to understanding tipping points and why outwardly similar crowds behave very differently.