There was also the Argument of Increasing Decency, which basically held that cruelty was linked to stupidity and that the link between intelligence, imagination, empathy and good-behaviour-as-it-was-generally-understood – i.e. not being cruel to others – was as profound as these matters ever got. This strongly implied that beings capable of setting up a virtuality so convincing, so devious, so detailed that it was capable of fooling entities as smart as – say – Culture Minds must be so shatteringly, intoxicatingly clever they pretty much had to be decent, agreeable and highly moral types themselves. (So; much like Culture Minds, then, except more so.)

But that too might be part of the set-up, and the clear positive correlation between beings of greater intellectual capacity taking over from lesser ones – while still respecting their rights, of course – and the gradual diminution of violence and suffering over civilisationally significant periods of time might also be the result of a trick.

A bit, after some adjustments for scale, like the trick of seeding another society with the ideas for a holy book that appeared to tell the truth on several levels but which was basically just part of an experiment, the Contents May Differ thought, as it reviewed the results of the latest sim runs.

The sims it was setting up and letting run were all trying to answer the relatively simple question, How much difference will it make if the Gzilt find out the Book of Truth is a fake?

And the answer appeared to be: Who the fuck knows?

Once you started to think that the only way to model a population accurately would be to read the individual mind-states of every single person within the real thing – something even more immoral than it was impractical – it was probably time to try another approach entirely.

As a good, decent, caring and responsible Culture Mind, the Contents May Differ would never run a sim of the Gzilt people at the individual level to find out anyway, even if it could have, and – apart from anything else – had decided some time ago that even resorting to such desperate measures wouldn’t solve anything in any case. Because there were two Problems: the Simming Problem and the Chaos Problem.

The Chaos Problem meant that in certain situations you could run as many simulations as you liked, and each would produce a meaningful result, but taken as a whole there would be no discernible pattern to them, and so no lesson to be drawn or obvious course laid out to pursue; it would all depend so exquisitely on exactly how you had chosen to tweak the initial conditions at the start of each run that, taken together, they would add up to nothing more useful than the realisation that This Is A Tricky One.