Concerning Fair, I appreciated the optimisation task as well as the open-endedness which allowed for several ways to work through the story to discover different facets of the overall microcosm.
One thing, though. The randomised sales conversations seemed a bit off technically. Very often, I would receive what seemed like two or three parser responses in a row, although I had only picked one option. Often, these reactions would even contradict each other. Of course, I don’t have access to my transcripts right now, but it went something like:
"Your potential buyer is happy with this answer. Good job!
I prefer optimization possibilities presented after basic gameplay mechanics and world familiarity, but I can understand this is not always easy to design for.
It wasn’t intentional, but it was one of those things I that was sticky to work out, so I left it in as it occasionally gave you a double response as if the person reacts two ways… Essentially what you are seeing is responses to the choice you made (sometimes which is random) and then the overall reaction when you near the end of the conversation. The way it works is there is a variable which starts at zero. A positive response increments it, a negative response decrements it unless it is zero.
To sell a book, the variable must be at least 2 by the end of the conversation. If you get two good responses and then one bad one you won’t sell a book. If you get one bad and two good, you will. So it’s possible that the last question you answered is a good response, but the game is alerting you that the conversation isn’t going well. Or your last response is bad, but the number stays above two so you get a good reaction and sell a book.
I tried to improve this by making some of the strong reactions more neutral so it wouldn’t seem so weird if you got a bad and a good one together, but if you have an early version this may not be implemented.