Regular readers of this design diary — if there are any — might have noticed that I have made no new diary entries for nearly a month. I have been working on a thousand distractions, none of which I shall explain here. I’m back.

I’m feeling impatient, so it’s time to slash through these problems with deals. I am reverting to the original Siboot design in which each actor starts each morning with accurate knowledge of one of the auragon counts of each of the other actors. The actual knowledge would be assigned randomly.

Lie detection and over-determination of auragon counts

Here’s a tricky little problem: with four actors, each one knowing a total of three auraong-counts, each auragon-count is known by exactly one person. This means that the holder of that information can lie about it and nobody could expose the lie. I could let each actor know four auragon-counts, but that would given each actor knowledge of two-auragon counts, which is dangerously powerful.

What about increasing the actor count? If there were five actors, each one knowing one auragon-count for each of the other actors, that would mean that a total of 20 auragon-counts are known, with only fifteen auragon-counts in play. Thus, five auragon-counts would be known by two different actors. If you tell a lie, there’s a one in three chance that your lie could be discovered — if the right person asks the other right person about the right auragon count.

This opens up a cornucopia of worms: lots of wonderful interaction possibilities, along with lots of design headaches.

I could take it even further with six actors, producing a total of 30 known auragon counts out of 18 actual auragon counts, yielding a two-thirds potential of being caught. With seven actors, we get 42 known auragon counts with 21 actual auragon counts, yielding a 100% potential for being caught. That doesn’t mean that you will be caught; it means only that you could conceivably be caught.

But seven characters is way too many. Even six is a lot. I think that five should be sufficient.

Results

Here’s the deal system that I now have:

A { requests | recommends | demands } deal with B

B { what do you want? | rejects } A

A tells B “I want X”

B { reject | tells A “I want Y” }

A { rejects | accepts }

B { delivers | lies }

A { delivers | lies | cheats }

For telling auragon-counts, the format is:

A reveals to B that C has auragon-count of D in amount E; source is F

All lies attribute the information to another source; this provides an additional layer of coverage should the lie be exposed. This creates the delicious situation in which B asks F whether F told A C’s auragon-count. In that case, F could tell the truth or lie, with the intent being to besmirch A or defend himself against being caught in a lie. See what I mean by ‘a cornucopia of worms’?