Computational model of selection by consequences: patterns of preference change on concurrent schedules.

Saule and colleagues built a computer 'organism.' It picks one of two levers.

Each lever pays off on its own timer, like two vending machines with different restock speeds.

When the payoff rates change, the program 'mutates' and keeps the moves that earn more points.

The team asked: will this digital creature shift its choices the same way real animals do?

Yes. The virtual player moved its time and responses to match the new payoff ratio.

The size and speed of the shift looked almost identical to pigeon and rat data.

A simple rule—keep what works, drop what doesn’t—was enough to copy the matching law.

McDowell (2004) ran the first test like this on single schedules; Kulubekova et al. (2013) widened it to two-lever choice.

Rojahn et al. (2012) later added reward size as well as rate; together the three papers show the same Darwin-like engine handles richer tasks.

Avellaneda et al. (2025) went back to live rats and found the sensitivity number itself drifts when overall payoff rises.

Their update does not kill the 2013 model; it just lets the ‘keenness’ dial move, giving you a smoother curve for your Excel sheet.

If a three-line code rule can mimic your client’s choice patterns, you can test schedule tweaks on a laptop before touching the classroom.

Try running a quick simulation when you plan to fade richer to leaner reinforcement. If the virtual kid stalls, your real kid probably will too—so adjust the steps before session one.