PaperlayerView exact revision on arXiv

Source-linked AI summary

Social physics

Marko Jusup, Petter Holme, Kiyoshi Kanazawa, Misako Takayasu, Ivan Romic, Zhen Wang, Suncana Gecek, Tomislav Lipic, Boris Podobnik, Lin Wang, Wei Luo, Tin Klanjscek, Jingfang Fan, Stefano Boccaletti, Matjaz Perc

arXiv:2110.01866v2physics.soc-phcs.SInlin.AOq-bio.PE

TL;DR

Social physics addresses how physics-based methods can organize research on societal phenomena, where unconstrained empirical hypothesis spaces can produce disconnected knowledge. This review surveys applications across cooperation, crime, migration, epidemics, and environmental and climate challenges, while highlighting both successful analytical frameworks and remaining limitations.

  • Problem

    Without firm theoretical expectations, behavioural research can test an almost infinite set of hypotheses and acquire disconnected knowledge despite rigorous experiments.

  • Method

    The paper reviews social-physics research across societal systems, including higher-order cooperation, criminal recidivism, migration, epidemic transport, environmental management, and climate dynamics.

  • Results

    Analytical epidemic-arrival-time estimates accurately capture outcomes for almost all populations in the worldwide air-transportation network.

  • Takeaways & Limitations

    Social-physics models provide a basis for studying societal phenomena while leaving opportunities to improve realism, including heterogeneous urban layouts in crime models.

  • Takeaways & Limitations

    The review notes that conventional Hadley-cell analysis does not fully resolve latitude-longitude structure because its equation accounts only for the latitudinal direction.

Abstract

from arXiv · show

Recent decades have seen a rise in the use of physics methods to study different societal phenomena. This development has been due to physicists venturing outside of their traditional domains of interest, but also due to scientists from other disciplines taking from physics the methods that have proven so successful throughout the 19th and the 20th century. Here we dub this field 'social physics' and pay our respect to intellectual mavericks who nurtured it to maturity. We do so by reviewing the current state of the art. Starting with a set of topics that are at the heart of modern human societies, we review research dedicated to urban development and traffic, the functioning of financial markets, cooperation as the basis for our evolutionary success, the structure of social networks, and the integration of intelligent machines into these networks. We then shift our attention to a set of topics that explore potential threats to society. These include criminal behaviour, large-scale migrations, epidemics, environmental challenges, and climate change. We end the coverage of each topic with promising directions for future research. Based on this, we conclude that the future for social physics is bright. Physicists studying societal phenomena are no longer a curiosity, but rather a force to be reckoned with. Notwithstanding, it remains of the utmost importance that we continue to foster constructive dialogue and mutual respect at the interfaces of different scientific disciplines.

1. Prologue: The physical roots of multidisciplinarity

Social physics is defined broadly as physics-informed research addressing societal problems, rooted in the historical exchange between social science and statistical physics. This review organizes the field around forces enabling modern life and threats that perturb it.

  • Social physics comprises active research topics addressing societal problems to which physicists have contributed substantially.
  • Probability and statistics moved from social science into statistical physics, while physicists now apply statistical-physics methods to quantify social phenomena.
  • The review treats physics as a root of multidisciplinarity and emphasizes the field’s broad scope.
  • Its topic selection asks what enables modern living and what perturbs or threatens it, including cities, traffic, markets, crime, migration, contagion, environment, and climate.

2. Urban dynamics

Urban-dynamics research uses physical and statistical models to define cities, explain city-size and spatial-growth patterns, and analyze movement, navigation, segregation, and infrastructure. These models reproduce selected urban regularities but can differ markedly from real cities.

  • Defining cities: Cities can be defined administratively, by population-density thresholds, or by recursive commuting relationships between smaller and larger regions.The density method uses boundary thresholds, while the commuting method begins with sufficiently populous seed regions.
  • Size of cities: City-size models generate Zipf-like power laws through population redistribution and rich-gets-richer growth mechanisms.The multiplication-diffusion model produces emergent power laws across a broad parameter range, while larger cities attracting more people grow faster.
  • Spatial growth: The Zanette-Manrubia-Solé model reproduces many features of real city growth but embeds the oldest regions completely within newer built environments.
  • Networks within and of cities: Urban-network studies examine spatial organization, segregation, and infrastructure, including nonlinear segregation produced by local tolerance rules.In Schelling’s model, final segregation increases nonlinearly with both the tolerance threshold b and occupancy f.
  • Trajectory analysis: Trajectory studies find people’s next locations predictable at around 90%, while route and navigation measures quantify detours and failures to find short paths.

3. Traffic flows

Traffic-flow research treats vehicular traffic as a complex system whose simple models explain qualitative phenomena across multiple scales. Observed patterns include phase-like flow regimes, phantom jams, hysteresis, and congestion waves that merge into wide jams.

  • Motivation: Traffic is studied as a self-organized, decentralized complex system with emergent behavior connecting short and long spatial and temporal scales.
  • Traffic-flow models: Traffic models span continuous density descriptions, discrete car-following theories, and cellular automata.
  • Observed phenomena: The tent-shaped flow-occupancy relation suggests free-flow and congested dynamic phases, with synchronised flow and stop-and-go motion at higher densities.
  • Observed phenomena: Phantom traffic jams can arise without an external trigger and move opposite to the direction of traffic.
  • Traffic-flow models: The Nagel-Schreckenberg model reproduces flow-density curves and phantom traffic jams, while coupled-map lattice models reproduce several empirical traffic characteristics.
  • Future outlook: New autonomous-vehicle data and cheaper collection through drones or tower-mounted cameras may support discovery of additional statistical laws of traffic.

4. Econophysics

Econophysics applies statistical-physics concepts and models to financial markets, explaining empirical price patterns through interacting-agent dynamics and trend following. Its models also connect market regimes, trading behavior, and intervention analysis.

  • Foundations: Econophysics models agents as interacting particles whose preferences depend on their social environment rather than fixed rational-choice assumptions.Empirical work focuses especially on domains with large datasets, including financial markets.
  • Historical development: 1991 power-law price changes and 1995 symmetric power-law tails after re-scaling helped establish statistical-physics analysis of financial time series.These studies contributed to the field’s early expansion and visibility.
  • Agent-based modelling: Deterministic dealer interactions with N ≥3 can generate chaotic, noisy price dynamics, while positive trend-following coefficients produce scale-invariant fluctuations.The model identifies nonlinear interactions and trend following as distinct mechanisms for realistic price-series characteristics.
  • Agent-based modelling: The dealer model’s Langevin dynamics yield price-change power laws whose exponent depends on d, matching reported empirical patterns.Here d measures the extent of trend following; contrarians can have negative coefficients.
  • Applications: Risk-averse dealer adjustments mimic bid-ask spread widening, loss cutting, and profit booking, enabling assimilation of financial time-series data for intervention planning and market-response prediction.These extensions increase the model’s realism relative to simpler dealer behavior.
  • Applications: A small number of triangular-arbitrage dealers can boost cross-currency correlations among the US dollar, Japanese yen, and Euro consistently with empirical observations.The effect is modeled through rapid circular exchange transactions.
  • Macroscopic models: The PUCK model reproduces market regimes ranging from nearly random walks to exponential, double-exponential, or finite-time-singularity behavior in abnormal conditions.Its quadratic-potential threshold is b(t) = −2: below it, price fluctuations grow or decline exponentially.
  • Macroscopic models: In the PUCK model, trend following acts as inertia, while b(t) < −2 produces negative viscosity associated with accelerating bubbles, crashes, and hyperinflation.The model also reproduces power-law price changes, abnormal short-time diffusion, and volatility clustering.

4.4. Order-book modelling: Financial Brownian motion

Order-book modelling treats financial price movements through an analogy with colloidal Brownian motion. The resulting layered view distinguishes orders that drive price movements from those that resist them, while motivating tests of continuous-price assumptions.

  • Order-book representation: Order-book models represent bids and asks on a price axis, with transactions occurring when incoming prices meet or cross the best opposing order.This provides a microscopic description of market interactions.
  • Financial Brownian motion: The colloidal-particle analogy places an imaginary particle between the highest bid and lowest ask, treating accumulated orders as surrounding molecules.The particle’s motion provides an intuitive representation of market-price movements.
  • Layered order book: Buy and sell orders divide into inner and outer layers with opposite behavior: increases in one layer correspond to decreases in the other.This layered categorization differs from treating all orders on one side equally.
  • Layered order book: Inner-layer orders show high positive cross-correlation with price velocity and act as a driving force, whereas outer-layer orders show negative cross-correlation and act as drag resistance.Figure 15 identifies γc = 18 as the point where the cross-correlation’s nature changes.
  • Model assumptions: The continuous-price assumption in models such as the Langevin equation requires justification through an estimate of the financial Knudsen number.The estimate relates one-direction price-movement distance to the colloidal particle’s inner-layer diameter.

4.5. A kinetic approach to financial market microstructure

Microscopic trading data supports a kinetic description of high-frequency traders as strategic liquidity providers. Empirical scaling relations connect individual trend following to collective order-book structure and distinguish HFTs from low-frequency traders.

  • Microscopic evidence: Microscopic trading logs record anonymized submissions, cancellations, executions, trader identifiers, and bank codes from the Electronic Broking Services foreign-exchange market.These data enable direct analysis of individual trading decisions.
  • HFT behavior: HFTs are defined here as traders submitting more than 2,500 limit orders weekly and typically maintaining two-sided quotes as liquidity providers.Their quotes seek bid-ask-spread profits and may also benefit from liquidity rebates.
  • HFT behavior: The buy-sell spread is the difference between an HFT’s best bid and ask, while its mid-price averages those two quotes.The spread can represent either a round-trip profit estimate or risk from adverse selection.
  • Trend-following behavior: Statistical correlations between past market-price changes and individual HFT mid-price changes provide evidence of trend-following behavior.The analysis uses tick time, incremented whenever a transaction occurs, and compares one-tick historical and future changes.
  • Trend-following behavior: After re-scaling, a common master curve appears for at least the top 20 HFTs, while conditional variance is independent of the historical price change.The mean adjustment follows the trend, but its magnitude is described as an intrinsic HFT property.
  • HFT–LFT comparison: HFTs typically keep fewer than 10 live orders with one volume unit per order, whereas LFT order volumes follow a power-law distribution.The statistics show less variation in HFT strategies than in LFT strategies, supporting the distinction between the groups.
  • Emergent order-book structure: Trend following induces collective order motion that accounts for the layered structure of the order book.This links individual strategic behavior to a macroscopic market-structure pattern.

4.6. Solving the microscopic model via kinetic theory

Kinetic theory reduces the high-dimensional stochastic dynamics of HFT order placement to tractable macroscopic equations. Under mean-field assumptions, the framework analytically derives order-book and timing statistics that agree closely with simulations.

  • Financial Liouville equation: The financial Liouville equation is an exact time-evolution equation for the N-body probability density and requires no approximation at its derivation.It is equivalent to the original stochastic model.
  • Financial BBGKY hierarchy: The BBGKY hierarchy reduces the exact many-body dynamics to lower-dimensional one- and two-body distributions, but the resulting equations require closure.The one-body distribution describes relative mid-price density conditional on spread.
  • Financial Boltzmann equation: Applying the molecular-chaos mean-field approximation yields a closed financial Boltzmann equation whose large-N leading-order steady solution is a tent function.The next-to-leading-order solution is also accessible for detailed mean-field analyses.
  • Financial Boltzmann equation: The average order-book profile is analytically derivable for any spread distribution ρ(L), indicating that the microscopic HFT model is analytically tractable.The derivation approximates the spread summation by an integral over all spreads.
  • Financial Langevin equation: Additional coarse-graining produces a financial Langevin equation whose tick intervals have exponential statistics and whose noise and interval statistics are analytically obtainable under mean-field assumptions.The macroscopic dynamics depend on dimensionless parameters including ˜c and scaled price changes.
  • Numerical confirmation: Analytical predictions for the order-book profile and exponential price-change distribution are confirmed numerically, with especially precise agreement for the order-book profile.The analysis attributes this to rapid decay of two-body correlations under the model’s collision rule.
  • Numerical confirmation: The kinetic formulation depends on molecular chaos, whose validity is supported here because large-N interactions make repeated collisions between the same HFT pair unlikely.This addresses why mean-field behavior can work despite the one-dimensional price space.

4.7. Consistency between theory and data

The kinetic model is tested against microscopic, mesoscopic, and macroscopic financial data, with theory matching observed order-book and price-change patterns across time scales.

  • Individual HFT buy-sell spreads are well approximated by a γ distribution, which implies the theoretical average order-book profile.
  • The model’s normalised average order-book profile agrees closely with the theoretical curve without additional parameter fitting.
  • At one-tick scales, price-change distributions follow an exponential law whose decay length κ depends on the selected time period.
  • Rescaling price changes by κ removes time-period dependence, collapsing the two-hourly distributions onto one exponential master curve.
  • At longer time scales, weekly price changes develop fat tails fitted by a power law with exponent α = 3.6 ± 0.13.
  • The decay-length distribution has exponent m = 3.5 ± 0.13, and α ≈ m provides an additional consistency check.

4.8. Future outlook: towards market ecology

The paper links microscopic HFT trend-following behaviour to the broader goal of understanding market ecology. Identifying strategy interactions could support realistic market simulators and regulatory interventions.

  • The microscopic trend-following model describes HFT behaviour at the time scale of one tick and is solved using kinetic theory.
  • Regression analysis estimates HFT trend-following coefficients and incorporates delayed, coarse-grained price signals.
  • The moving-average weights follow an exponential scaling law for 85% of examined HFTs.
  • HFTs divide into short-, intermediate-, and long-time-scale trend followers operating over about 4, 20, and 40 ticks, respectively.These correspond to approximately 30 seconds, 3 minutes, and 6 minutes.
  • The remaining 15% of HFTs use other trading strategies.
  • Understanding interactions among strategies could make regulatory market simulators plausible and help regulators plan interventions for liquidity and stability.

5. Cooperation

The review examines how evolutionary game theory, network structure, temporal dynamics, and empirical findings illuminate the evolution of cooperation. It emphasizes general rules, mechanisms, and the need to reconnect theoretical models with experiments.

  • 5.5. Cooperation in networks with higher-order interactions: The review identifies a persistent gap between theoretical cooperation models and empirical evidence, motivating stronger connections between theory, experiments, and behavioural disciplines.It also discusses cooperation effects in coupled layers, where small thresholds can produce suboptimal synergy and dampening can promote cooperation unevenly across layers.
  • 5.1. Social dilemmas: Evolutionary game theory models strategy frequencies and payoffs, identifying stable states and evolutionarily stable strategies across social dilemmas.The replicator equation tracks strategy frequencies from per-capita and average payoffs; prisoner’s dilemma selects defection, whereas snowdrift supports coexistence.
  • 5.1. Social dilemmas: Dilemma-strength parameters can make dyadic games with different payoff matrices equivalent in terms of evolutionary outcomes.The review further identifies positive assortment as the common feature linking kin selection, group selection, reciprocity, and network reciprocity.
  • 5.3. Extended network models: Adding third strategies can produce cyclic dominance, and evolutionary dynamics vary with the topology of the underlying network.Cyclic dominance describes intransitive relations in which A dominates B, B dominates C, and C dominates A.
  • 5.2. Cooperation in networks with pairwise interactions: Network structure promotes cooperation in the prisoner’s dilemma but can reduce cooperator frequency in the snowdrift dilemma.Under weak selection, cooperation is favoured in pairwise networks when b/c > k; heterogeneous networks provide especially favourable conditions.
  • 5.4. Cooperation in temporal networks: Temporal-network cooperativeness depends on both interaction aggregation and the relative speed of evolutionary and network dynamics.Increasing g can improve cooperation beyond static-network levels despite burstiness, while intermediate aggregation periods resist defection most strongly and unfavourable temporal parameters can nullify these effects.

6. Networks and communities

Networks are central to social physics, and community detection uses network structure to identify basic constituents of complex systems. The section emphasizes stochastic block models as generative, statistically grounded tools, while noting their extensions, limitations, and future competition from scalable machine-learning methods.

  • Networks span social physics applications including online, physical, animal, financial, transport, power-grid, climate, medical, nutrition, and sports systems.
  • Community detection simplifies complex networks by identifying constituent groups while retaining information about how the whole system is connected.
  • Statistical inference of communities: Degree-corrected SBMs accommodate real-world degree heterogeneity by assigning each node a parameter controlling its expected degree independently of community affiliation.
  • Statistical inference of communities: Canonical and microcanonical SBMs agree asymptotically for sufficiently large degrees and intercommunity link counts, but can differ substantially in small or sparse networks.
  • Statistical inference of communities: Nested SBMs address the resolution limit, while non-parametric formulations determine the number of communities internally rather than treating it as an external parameter.
  • Future community-detection methods are likely to favor practical approaches that scale to billions of nodes and support multilayer, dynamic, and incomplete networks.

7. Human-machine networks

This section reviews how AI-driven human-machine networks can support social-good applications while raising concerns about opacity, fairness, privacy, and accountability. It covers AI fundamentals, learning methods, cooperation between agents, and directions for human-centric trustworthy AI.

  • Human-machine networks: Data science seeks to derive knowledge from increasingly large, diverse, and frequent data, while augmenting human intelligence and decision-making for better lives.
  • Human-machine networks: AI-driven human-machine networks are reviewed across human and machine behaviour, methodological AI fundamentals, and research directions for future breakthroughs.
  • Literature walkthrough: Machine behaviour studies how AI agents affect society, culture, economies, and politics across individual machines, machine networks, and human-machine networks.
  • Fundamentals of artificial intelligence: Artificial narrow intelligence performs well on specific, well-defined tasks but cannot augment humans beyond the limited domain in which it learned to operate.
  • Fundamentals of artificial intelligence: Reinforcement learning trains agents to maximise cumulative lifetime rewards through feedback in complex or uncertain environments where optimal policies are unknown beforehand.
  • Learning to learn: Meta-learning exploits prior learning experiences to generalise to novel tasks, including supervised few-shot learning and optimisation-based approaches such as model-agnostic meta-learning.
  • Learning to learn: Meta-learning methods have achieved human and superhuman performance on simple one-shot classification tasks, while their broader role in artificial general intelligence remains prospective.
  • AI agents for promoting cooperation: In social dilemmas, learning with opponent-learning awareness can lead commonly defecting reinforcement learners to cooperate and develop tit-for-tat behaviour.

8. Criminology

The review presents crime as a socially embedded, dynamically spreading phenomenon and surveys statistical-physics models for understanding and mitigating it. Findings emphasize that interventions can have nonlinear effects, while social context, networks, and reward–punishment systems shape outcomes.

  • Motivation: Crime recurs and proliferates across societies, with deterrence remaining difficult even in strongly monitored and policed states.The challenge is especially severe where social responsibility is weak or greed overrides moral constraints.
  • Approaches: Statistical-physics approaches model crime through hotspots, self-exciting point processes, agent-based models, evolutionary games, and criminal networks.These approaches target both the emergence and diffusion of crime and the likely effects of policing interventions.
  • Crime hotspots: Nonlinear feedback between criminal positions and attractiveness fields produces complex aggregation patterns resembling real residential-burglary hotspots.The hotspot model exhibits four regimes of attractiveness dynamics associated with different burglary realities.
  • Crime hotspots: Only subcritical hotspots can be permanently eradicated by suitable suppression; supercritical hotspots are displaced rather than removed.The result follows from setting the crime rate to zero at selected locations for a specified period.
  • Crime prediction: Point-process methods outperform fixed-kernel approaches by better balancing exogenous and endogenous contributions and inferring patterns directly from data.They also perform better for robberies and car theft, where near-repeat effects are less prominent.
  • Social dilemmas: Informants can drive an initially crime-dominated population toward a crime-free state, although utopia may be elusive in highly adversarial societies.A deterministic version shows that any initially positive informant presence, I_0 > 0, leads to utopia regardless of δ, θ, and ϵ.

9. Migrations

The review frames migration as both beneficial and destabilizing: underlying conflicts and climate change may drive future migrant waves, while tolerance and integration shape whether societies reach cooperation, newcomer dominance, or antagonism. It also links integration outcomes to cultural similarity and shows that right-wing populism can undergo tipping-point transitions.

  • Migration pressures: European integration and free movement require costly efforts to harmonise relations, as the EU migrant crisis demonstrated.The crisis also brought momentary border controls and populist policies, although a devastating phase transition was averted for now.
  • Migration pressures: Migration can alleviate labour-force deficits and ageing-related age-structure imbalances, but unresolved conflicts and climate change are expected to produce future migrant waves.The review asks whether the current world order is prepared for these pressures.
  • Tolerance and populism: Tolerance toward immigrants is conditional, and sudden inflows correlated with increased support for an anti-immigrant right-wing populist party in Germany.The review therefore calls for quantitative analysis of immigration, integration, and local tolerance together.
  • Tolerance and populism: Immigration policies can produce mutualism, newcomer dominance, or antagonism, depending on the resulting fitness ratio and whether cooperation remains intact.Mutualism preserves a sustainable local majority, newcomer dominance produces a newcomer majority, and antagonism is an absorbing state caused by complete breakdown of cooperation.
  • Tolerance and populism: Sufficient tolerance and reasonable integration rates support mutually beneficial interaction, whereas slow integration with low tolerance can lead to turmoil and violence.The review concludes that successful policy requires measuring and monitoring local tolerance and newcomer integrability.
  • Integration and culture: Cultural similarity is associated with integration outcomes: culturally proximate immigrant groups feature in successful cases, while cultural distance and non-mixed households are linked to weaker economic integration.Axelrod’s model further suggests that local social influence can polarise culture rather than integrate immigrants without concerted central-government effort.

10. Contagion phenomena

The review covers digital epidemiology and analytical models for understanding and forecasting contagion across populations. It highlights both the promise of diverse data sources and the importance of accounting for network structure and mechanistic epidemic dynamics.

  • Digital epidemiology: Google Trends queries correlate with reported influenza cases and opened the way to near-real-time epidemic detection.
  • Digital epidemiology: The original Google-based algorithm produced inaccurate estimates because it ignored influenza seasonality and changing search behavior.ARGO addressed these limitations by incorporating time-series structure, and later work added spatial and temporal synchronicities.
  • Digital epidemiology: Twitter supports disease surveillance across scales; dengue predictions in Brazil reached a correlation coefficient of 0.98 nationally.
  • Digital epidemiology: Digital traces also reveal public-health signals, including vaccine sentiment, adverse-event candidates, social connections, and mobility-related disease and economic effects.Studies linked online vaccine sentiment with regional vaccination rates, identified 72 definite adverse-event cases after expert review, and found tradeoffs between disease transmission and worker mobility.
  • Digital epidemiology: Combining multiple sources, including Google Trends, Twitter, Flu Near You, and CDC data, improves influenza predictions up to four weeks ahead.
  • Analytical models: Epidemic thresholds decrease as metapopulation networks become more heterogeneous in node degree.The threshold marks where an epidemic wanes or intensifies, and recurrent mobility patterns have also been incorporated into this line of work.
  • Analytical models: A closed-form probability distribution for epidemic arrival times addresses a previously unresolved analytical gap.

11. Environment

The environment section examines how social physics and mechanistic models clarify environmental risks, ecological tipping points, and management trade-offs. It highlights evidence that scientific understanding can support regulation while stressing limitations in current chemical-risk assessment.

  • Environmental degradation is difficult to prevent because mitigation imposes direct and opportunity costs, encouraging economies to externalise environmental harms.
  • Strong scientific evidence and quantified costs helped motivate rapid international action on ozone-depleting substances, followed by recovery of the ozone layer.
  • Chemicals are usually tested and regulated independently even though environmental mixtures can exacerbate toxicity.
  • Mechanistic DEB models link environmental energy and material availability to growth, reproduction, mass and energy balances, and hazard rates.
  • In the predator–prey model, increasing predator mortality produces invasion and persistence thresholds, with intermediate mortality allowing adult prey to peak before predator extinction.

12. Global climate change

The climate section connects climate change with societal outcomes and develops network-based tools for modelling, forecasting, and detecting critical transitions. It reports broad evidence of climate impacts alongside unresolved gaps in human dynamics and mechanistic interpretation.

  • Climate variables are associated with mortality, agricultural income, labour productivity, electricity use, economic output, and interpersonal aggression across multiple settings.
  • More than half of the observed increase in global average surface temperature from 1951 to 2010 was caused by anthropogenic greenhouse gases and other forcings.
  • Global climate change: Climate models treat human dynamics as external, motivating integrated assessment models that link socioeconomic systems with the biosphere and atmosphere.
  • Climate networks: Climate networks represent locations as nodes and connect them according to similarity between climate records, commonly using correlation-based link construction.
  • Climate networks: The climate-network approach reliably predicted the onset of the 2014–2016 strong El Niño event in 2013, with accuracy higher than cited state-of-the-art climate models.
  • Climate networks: Network divergence identified propagation pathways and predicted more than 60 % of extreme-rainfall events in the Central Andes, exceeding 90 % during El Niño conditions.

13. Epilogue: Keeping the dialogue open

The epilogue argues that social physics depends on interdisciplinary exchange rather than disciplinary dominance. Quantitative modelling must be paired with expert input, appropriate approximations, and sustained dialogue.

  • The review omits topics including the physics of art, agriculture, gastronomy, ethnology, and civil unrest to maintain a contiguous account.
  • The authors acknowledge that their broad definition of social physics may be contested, especially at boundaries with engineering, artificial intelligence, and climate modelling.
  • Imperious behaviour by physicists can threaten mutual respect and jeopardise the success of multidisciplinary collaboration.
  • Formulating relevant hypotheses and model assumptions requires expert input because intuition and common sense cannot replace domain knowledge.
  • Physicists’ simplifying approximations can conflict with disciplines that prioritise exceptions, making dialogue important when applying general models to complex systems.
  • Open dialogue, patience, and care can produce insightful and impactful research needed for humanity’s continued prosperity.
Loading 2110.01866v2…