Recently, I wondered. What if you have an AI that tracks events as nodes in a graph, and then associates these events with each other by proximity, maybe with a directed graph with propabilities attached to them. And each time events occur either simultaneusly or in close succession, you add this to the graph. Over time, you could aggregate some kind of pavlovian conditioning, where the listener expects certain things when perceiving events. You could have causality and correlation chances for events, and when faced with decisions, make educated guesses based on past experiences.
Just a thought. Not really related to stardew valley.
Just a thought. Not really related to stardew valley.