Provenance
The creative direction is based on the project's Sahilverse commission: human capital, informal retail, education, agency and unintended institutional consequences. No private anecdotes, identifiable customers or family financial records have been encoded. Published Sahilverse manuscripts were not independently reviewed for this rebuild, so the experience is not presented as a verified adaptation of any specific text.
Why this is a model, not a forecast
All coefficients, starting populations and budget units are invented. Six mechanisms demonstrate plausible directions but not measured magnitudes or causal relationships in Delhi, India or elsewhere. The rules are deliberately contestable. The deterministic pseudo-random sequence permits reproducibility, not external validity. A different choice of parameters can reverse a policy ranking.
Model outline
120 agents: income capacity, wage, skill, health proxy, interpersonal trust proxy, caring obligations, work state, educational opportunities, debt, reserves. Four districts: rent, jobs, housing units, maintenance, travel burden, retail cash and employer cash. City: partial public budget, bank liquidity proxy, prices and employment. In each period: district market response → household employment and spending → scenario interventions → fiscal updates → aggregate measurements.
Scenarios: credit access influences enterprise income, loan servicing and shortfalls; training trades lost time for skill and later wages; rent caps relieve rent but lower maintenance and potentially supply; transit relief influences travel burdens and public spending; schooling slowly influences future earnings; schedule flexibility eases some care burdens while introducing a staffing friction. These are simplified storytelling mechanisms, not validated representations of policy programmes.
Explicit caveats
Not a closed-economy monetary simulation: wage inflows and consumption outflows need not conserve cash. Employer liquidity, banks and public budgets are partial proxies; they are not fully reconciled double-entry balance sheets. There is no migration, elections, inflation expectations, corruption, land-market equilibrium, endogenous births, or differentiated firms with bargaining strategies. Household roles are archetypes, not demographic data. Trust, stress and health are fictional game variables, not diagnostic scales. The inequality measure uses shifted nonnegative wealth for a simple proxy and must not be compared with official Gini coefficients.
Reading trail (conceptual background, not validation)
Thomas Schelling — Micromotives and Macrobehavior (micro choices, macro patterns); Epstein & Axtell — Growing Artificial Societies (agent-based models); Selbst et al. — Fairness and Abstraction in Sociotechnical Systems (limits of abstraction); Journal of Artificial Societies and Social Simulation (methods and debates). These readings inform the conceptual domain; no empirical coefficients were taken from them.
How to investigate
Compare identical seeds with distinct policy strengths. Select a single citizen and examine their period trace. Change the seed to look for sensitivity. Rewind and replay. Export the full JSON ledger, including both worlds and the individual event histories. Treat any pattern as a claim within this artificial world only.