Games might teach people how to manage agents better than work does
Could games help people learn to lead AI agents?
Johannes Cortez argues that games could give people a lower-stakes way to practice instructing agents and seeing the consequences. In his view, learning to lead agents could be a side effect of wanting to win, but the game still has to be fun first.
The Mochi loop is simple: give an instruction, watch what the agent understood, then deal with the result.
That is also the real problem people have with agents at work.
Was the bad outcome caused by the model, the available tools, the instruction, or the human who assumed the agent could read their mind?
A game gives you that feedback without risking payroll, customers, or production.
Sitrep still has to be fun first. Nobody wants corporate training wearing a tactical-shooter skin.
But if players get better at leading agents because they wanted to win, that is a pretty good side effect.
Johannes Cortez: GitHub · dev.to
More from Johannes Cortez · The next generation of games · Sitrep field notes