Economic dynamics in the presence of general automation

A relatively simple state-space model of a 2-sector economy as general-purpose task automation technology is introduced.

AI-enabled cognitive task automation will provide new business efficiency opportunities, but also place downward pressure on wages even as GDP continues to increase. A notable new failure dynamic may force people to exit the economy when market-driven wages drop their income below subsistance. This drop in economic participation need not yield the corresponding drop in GDP expected in economies to date, where historically much of the economy retained non-automatable roles.

Effectively there's a greater risk that the horse breaks away from the carriage and that the wellbeing of large portions of the population become market externalities: The economy ain't buyin' what they're selling and, accordingly, does not respond to their demand signals either.

A six state differential algebraic equation (DAE) model is presented here to investigate the dynamics of the introduction of general-purpose automation technology (AI in its various forms) into an economy like our own (USA circa 2025). The intent is to use such a model in economic policy planning to ensure that technology is adopted at rates and this regulatory support to arrive at desirable economic stability points rather than unstable regions or undesirable stability points.

The phase trajectories plotted below for different sets of initial conditions show that a wage subsidy can steer the system up to a high-wage and high-GDP stability point, if it starts at a sufficiently high initial wage level. Once initial wages drop below a threshold, subsidies only suport a path to a less desirable stability point with much lower GDP and a tens-of-percent of the population excluded from economic participation by downward wage pressures from the market.

See the pdf in this package for a the set of differential algebraic equations. Scripts to build the system of equations and simulate various initial conditions and candidate federal regulatory policies are included.