'From Game to Behaviour Model' is my ongoing scientific work, stalled a bit because of AFCALC. I can explain core premise in two ways.
First and short way: that article is about writing an automatic solver for the game Princess Maker 2 by GAINAX. :)

 But that's not scientific enough. So, second and long way.

Suppose we have a completely deterministic world, in which any action taken results in well-defined consequences.
And suppose than in this world exists an active agent, defined by set of numeric characteristics which we will call an agent's "state".
The world itself consists of static predefined set of actions, each action modifying the agent's state.
Obviously, we can state the desired state and then state a problem of finding a sequence of actions which will lead an agent from it's initial state to desired state.

So, this work is about solving a problem, in which an agent finds such a sequence by itself, following some mechanism, which we will call "behaviour". Other details are in the paper.
Mathematical tools by which this problem is solved in my work are genetic algorithm probing a modified Mamdani fuzzy controller.

I wrote initial draft of the paper in the 3rd year of studying in the university and assembled a GUI program for the calculations. Program is written in Visual Prolog 7.1, so it works only in Windows.

Program code and the text of the paper both in the Public Domain according to CC0 license.

--
STATUS @ 2011.09.24 21:34
Project made public via GitHub.

TODO:
1. Rewrite the draft of the paper in English.
2. Add the prologue to the paper explaining the roots of the work lying deep into Princess Maker.
3. Decouple the program logic from the interface. Maybe it's already done.
4. Rewrite the core logic so you can build a library and separate application linked to it.

Estimated work: 40 hours.
Planned finish: November 2011.