“Knavery and flattery are blood relations”

I’ve been doing experiments in my LOST mini-storyworld to determine the correct settings for the belief equation that I came up with. Here’s the belief equation:

What I’m trying to determine is what weight values should be substituted for the 0.0 after pF_H, fM, and C_S to influence actor reactions.

I first started playing around with the parameters in Scriptalyzer, setting the weight sliders and recording the results.

It was helpful but I realized that I needed something more visual to help me see at a glance the effects of the weight changes.┬áSo I started creating these little 5×5 matrixes on graph paper, setting various parameters, and then running through a whole series of flatteries.

I realized it was going to take me quite a while to run through all the calculations particularly since I had to create the little matrixes by hand and then seeing what each actor’s reaction to Jack’s flattery was.

My next thought was to create a spreadsheet where I could input values for various Traits, their properties, and their associated aTraits and pTraits and see the final results. I actually created a spreadsheet and programmed a macro to color the individual cells based on a numeric value.

But then realized that since I don’t know the actual Storytron algorithms anything I coded up would not reflect what’s actually going on behind the scenes in the engine. (Though a visual spreadsheet-like tool in SWAT, a cross between Erasmatron’s AlgoLizard and SWAT’s Scriptalyzer, that allowed you to do calculations and see results visually would a great help.)

So I returned to my LOST mini-storyworld, change values and settings manually and run through various iterations.

The first series of experiments had the following parameters:

  • False_Honest was a visible trait
  • KnowsMe[all other actors, Jack] was true
  • Jack’s False_Honest was set to three approximate values using the trait slider – Totally False, Neither False nor Honest, Totally Honest
  • The weights for pF_H, fM, and C_S were set to three values – 0.0, 0.4, and 0.8. Only one weight was changed at a time, the other two were kept at 0.0.

Here are the results.

After some forum discussions about the rules around how pTraits are initialized and set, I did a second set of experiments with the following parameters:

  • False_Honest was an invisible trait
  • All non-Jack actor aF_H was set to match their F_H
  • KnowsMe[all other actors, Jack] was false
  • Jack’s False_Honest was set to three approximate values using the trait slider – Totally False, Neither False nor Honest, Totally Honest
  • In the Relationship editor, the checkbox next to pF_H for all other actors in relation to Jack was unchecked so pF_H would be calculated automatically.
  • The weights for pF_H, fM, and C_S were set to three values – 0.0, 0.4, and 0.8. Only one weight was changed at a time, the other two were kept at 0.0.

Here are the results.

To see the results at a glance, here’s two “heat maps” from both sets of experiments. The one on the left is where False_Honest is visible; the one on the right is where False_Honest is invisible.

Each group of seven squares displays is the results of changing pF_H, fM, and C_S weights when Jack’s False_Honest varies from Totally Honest (top) to Totally False (bottom). Within each of the seven squares in a group you can see the weight changes

  • pF_H – Changes from 0.0, 0.4, 0.8 going across the top of the square from left to right.
  • fM – Changes from 0.0, 0.4, 0.8 going diagonally through the square from upper left to lower right.
  • C_S – Changes from 0.0, 0.4, 0.8 going down the left side of the square.

So, should False_Honest be visible or invisible and what should the equation weights be?

I’m leaning towards having False_Honest be an invisible trait because I like the distribution of actor reactions irregardless of equation weights. Also, Facundo made an excellent observation on the forums that reinforces whey False_Honest should be invisible:

“A visible trait is ideal to model characteristics that are easily spotted at sight (height, perhaps social status, sex, etc). When dealing with traits about behavior a visual examination of a character would not give much clues about it. Under which circumstances honesty could be considered a visible aspect?”

So False_Honest is going to be invisible and I’ll let the other actor’s aTraits help set their pTraits towards Jack (even though I don’t fully understand the algorithms behind the scenes).

Remaining items regarding flattery:

  • Weights for belief equation parameters.
  • Substitute aFalse_Honest for Credulous_Suspicious (Chris’ suggestion)
  • Actor EmotionalReaction scripts in response to Jack’s flattery.

Maybe I can have this all wrapped up by the end of the Labor Day weekend.

This entry was posted in Side Projects and tagged . Bookmark the permalink.