Modeling Societal Dynamics with Historical Data
Abstract:
Individual human societies have increased in scale and complexity from tens to billions of people over the last six millennia, accompanied by repetitive collapse and fragmentation. We discuss efforts by ourselves and colleagues in the Seshat Databank project to understand these complex dynamics using a variety of techniques, highlighting two recent quantitative models exploring the drivers of increasing social scale and, conversely, crisis fragmentation in different historical contexts. We discuss ongoing work to adapt these models to contemporary societies, noting the benefits and challenges of translating historical insights to help navigate the complex challenges faced in the modern world.
Bios:
James Bennett is associate faculty at the Complexity Science Hub in Vienna. Since 2015 he has been a part of Seshat: Global History Databank. His research investigates the dynamics of human history, in particular the rise, spread, and fall of societies, from the Neolithic to the modern. Prior to this he was Vice President of Recommendation Systems at Netflix and responsible for the Netflix Prize.
Daniel Hoyer is a computational historian and complexity scientist. He has been part of Seshat: Global History Databank since 2014 and is currently an affiliated researcher with the Complexity Science Hub, Vienna and the SocialAI lab at the University of Toronto. His research seeks to understand societal responses to shifting ecological, social, and economic contexts that determine well-being outcomes in the past, as well as how this may shed light on critical social pressures today.
Summary:
Focus: how do societies function and how do groups of humans work and evolve?
Many patterns and underlying dynamics in today’s society are similar to how they have worked in the past
Coordination
Resource management
Interaction with environment
Observation: most past societies and civilizations don’t exist today
Driven into collapse by common challenges
Many lessons to be learned from this
Seshat databank (https://seshat-db.com/)
100s of years of data on societies
Since 2011
Societal Dynamics (SoDY): translates this into relevant insights
Turning history into data
Cliodynamics:
Scientific approach to study history
Theory testing, statistical analysis, model building
Big quantitative data + deeper quantitative descriptions
Approach:
Focus on region in the world
Track the sequence of polities that existed at that location
Analysis: develop codebook, consult with experts
Quantify “social complexity”
51 variables drawn from various theories about social complexities
E.g. monetary system, legal code, population
Tracked over time within and across polities
400+ societies from late neolithic to early modern
Analyze, test theories, develop causal models
Example: do changes in military technology cause polities to increase in area?
Track data for different societies
Compare to simulations that encode the dynamics predicted by theories
Observation: increases in social complexity driven primarily by transition to sedentarism, agricultural productivity, war
Model: Nomads-Agrarians
Spread of societies in Eurasia 1500 BC - 1500 AD
Historical Observations:
~750 BC the land controlled by polities started rising
More states over 3000 years concentrate in Egypt/Sumeria/China/Turkey region, with many less dense regions (Europe, Russian steppe, sub-Sahara)
Larger polities, even very large empires, don’t last longer
Agent-based model
Types: Agrarian states and Nomadic confederations
Generated synthetic history with geography, technology, wars, competition
Investigate impact and timing of a few exogenous contingencies
Regional agricultural productivity
Asymmetrical threats from military innovations
Regional efficient military transport improvements
Observations from model:
Agrarian states:
5% of productivity is reserved for warriors; actual warrior population varies between 1-3% during growth, then starts to saturate the surplus, thus limiting the state’s growth
Dynamic: occupy more land -> produce more warriors -> more productive -> occupy more land
Once warriors can no longer expand the state (military logistics constraint) they saturate the new surplus and start to fight each other; then start a civil war, producing a split
Asymmetric military shock:
Nomads provides the shock; much smaller population compared to agrarian states
Develop new cavalry tactics (horses, compound bows) that defeat sedentary armies of agrarian states
Agrarian states adopt same technologies: auto-catalytic power race between the two (mirror empires along the steppe-agrarian border)
Simulation:
~50 Central Asian nomadic tribes in 1500 BC
Nomads never annex agrarian state land and vice versa
Nomads can confederate with each other, confederations collapse stochastically
Co-evolution where both nomads and agrarians get stronger and larger, with border groups taking over their deeper neighbors
Agrarian states learn from nomads and each other
Model predicts overall 3,000 years of history roughly correctly, capturing the development of steppe empires, China, Europe, etc. but with some error in predicting population; distribution of state sizes is mostly right
Some parameters can be estimated directly from history, others relatively: military logistical efficiency; need to be calibrated to ensure agreement with overall historical record
Estimated from data: agricultural productivity, strike depth of nomads
Estimated relatively: military logistical efficiency
Model: Demographic cycling in Europe during mid-Holocene
Populations move from Turkey into Europe -> France -> Atlantic islands over 4-5k years in multiple pulses
Observed dynamic of initial population growth -> decline -> regrowth
After a population moves in there is a drop in population ~400 years later, followed by slow regrowth over following millenia
Hypotheses for what drives the boom and bust pattern:
Booms and busts driven by changes in climate
Don’t find long-term shifts that explain these collapses
Driven by inter-village conflict
Populations move out to new lands
Remaining populations don’t have anywhere to expand to, turn into aggressors who attack the villages that moved out earlier
After war period, return to stability in a more empty land
Model of this dynamic is more consistent with archeological evidence
Future
Dynamics of crisis
Seshat Databank’s CrisisDataBase Project
What drives crises?
What can be done about it?
Quantify social pressures, societal responses (e.g. revolution), and post-crisis trajectories (recovery, continual instability, collapse)
What factors determine the ultimate outcomes of crises?
Can we rewrite the history of the near future
Multi-path forecasting
Use current state and historical developments from ~10-50 years prior to make near-term forecasts
Using dynamic / complex systems framework; 'cycles' are emergent under certain conditions, but not essential (in some conditions spikes of discord don't cycle - one of the ways the perspective differs from Howe & Strauss work cycles of discord modeling framework
Analyze possible policy interventions through scenario exploration simulations