

It’s more likely that Connecticut comes alphabetically after Colorado in the list of state names and the number of data sets it used for training that were lists of states were probably abover the average, so the model has a higher statistical weight for putting connecticut after colorado if someone asks about a list of states
You can do timed mutes of keywords, though the times aren’t as customizable as some people might like.