It looks good to me, Laurent!
Just to detail it some more, I think the precise buttons at the bottom we need at this point are:
Current Govt Profile (negotiated values)
Ruler Preferences (only screen where the player can change values)
Military Political Block (values desired by MPB)
Religious Political Block (same)
Capitalists PB (same)
People's PB (same)
Bureaucracy PB (same)
For playing purposes, there should be just two groups of variables: pol powers and policies. Not three, as in the model (and in the picture). In other words, the player doesn't need to know and shouldn't even care, that Private Property and Social Policies are not handled in the same way as all other policies (Civil Rights, Slavery, etc) inside the model. In fact, I think it'll be more confusing if we split them in the interface as we're doing now. The player just needs to understand there're policies and there's a structure of power that determines who defines the values for those policies. I would, therefore, put all policies in the right window (move PP and SP to the right window).
In that right window, though, we may, if we believe it'll help the player, group policies by themes using colors. FE, Civil Rights, Slavery, Ethnic Discr and Religious Discr could be blue (the "liberties policies"), Private Property, Social Policies and Tax Rate could be red (the "economic policies") and Foreign Affairs could be green (the "International policies").
Another thing: I might be taking your picture too literally, and if so, ignore me, but just in case: social role values, as shown in the left window, should not be shown and should not be subject to change. What should be shown are pol powers (ruler, people, capitalists, military, religious and bureaucracy), hopefully through a pie chart.
I don't know how difficult it is, but it should be nice if in the ruler's window, where the player has to enter his desired polpower distribution, the interface would allow you to move the rays of the pie (the lines dividing the pie segments), thus changing the size of the pie segments (thus changing the distribution). That's at least the friendliest interface I can imagine for a group of variables that are forced to sum 100%.
What do you think?