In the case of the people talking, it seems that the surrounding nodes would need to broadcast their shatter conditions so that the node’s currently being simulated know what factors to calculate… Perhaps even, the only factors that need to be simulated are those that relate to shatter points of surrounding nodes… Like sarcasm, its not even thought about (thus not simulated) amongst those who don’t care. Has this been thought about already?
-Excerpt from a comment by Luke
You raise an interesting point, and one I think bears examination.The idea of simulating only the bare minimum required by circumstances is a sound one. It avoids many of the problems of creating masses of extraneous information, and transmitting this information between characters. On the other hand, one of the core ideas in Fledgeling is that of imperfect perception, either by way of being incomplete, incorrect, or deluded. And, in real life, one of the characteristics of shatter points is that you rarely know where they are until you reach them. In other words, they are very difficult to perceive.
But this raises a problem in conjunction. If the archon (The GM shard instance tasked with arbitrating local conflicts. If the GM is the general, directing the entire campaign, the archon is the commandant, or whatever.) only requests information that it needs, then the characters can back-sim likely conditions and indirectly percieve more than they should about their surroundings, future, and past. This happens to inexperienced GMs all the time when they ask all the characters to “roll perception” or “roll a will save”. The act of asking for that specific information clues the players in to what is happening.
One solution is the brute force approach, the equivalent of checking every skill every round for every character. While this might work in a computer context where brute force is at our disposal, we have much better alternatives available. (You’ve probably already figured this out actually, but I’ll spell it out anyway for clarity) Namely, the GM keeps track of the players stats and does the rolls in secret.
The method we plan to use in Fledgeling is something like this. Each character has a detailed model of itself. This includes all manner of complexity that is not normally communicated to the external simulation. The character communicates only a simple model of itself to the archon. The archon runs all simulations on this simplified model, and communicates circumstances back to the character.
Some of these circumstances are the actions of the character itself. If there is time for the character to react, the character may close the loop. However, many actions such as triggering an avalanche, or offending someone in a conversation (as well as most reflexes) are not really subject to character decisions. In such cases, the archon tells the character “you did such and such, and this is the result” and then the character can retro-sim (as appropriate or desired) the internal causes for these actions based on the much more complex internal structure.
Let us use the avalanche valley in a concrete example. A character alone walks through a snow-bound valley. The archon knows this character’s general history, and that they don’t know much about avalanches. The archon decides that they make a loud noise at some point, based on the simplified model, and an avalanche starts. The archon informs the player of the volume of noise they were making, and that they see a huge wall of snow descending the side of the mountain. The character can then decide that they were awed into song by the beauty of the landscape, or were calling out to see if anyone else was in the area, or whatever other reason. They may also make decisions based on this information. Perhaps the character attempts to run, or contemplates their life until they are overtaken by the debris.
You see in the example how the potential for an avalanche was much more difficult to perceive than the avalanche itself, and how there was no warning of (and thus no opportunity to react to) the conditions required to alter the landscape. The archon kept the internal state a secret from the character until the results were perceptible.
However, imagine that the avalanche is not quite so easily triggered. Perhaps there is a slight shifting of the snow, and the hillside creeps a bit before letting go. The archon could inform the character of the rumble coming from the hills, the tiny bits of snow skipping down the packed ice. At this point the character can choose to alter their state, perhaps “I’m being quiet” or “I’m running away in terror” or some-such. Or, if they don’t know better, perhaps they begin climbing the mountain to investigate. Whatever the case, the character now has some way to choose their actions, instead of the archon choosing for them.
Note also the practical applications of this model on real life. If people are easily and quickly offended, then their offense will rarely be based on actual decisions made by other people, and will be founded mostly on random chance circumstances. The ideal is to have ductile robust reactions to others while also telegraphing our internal state, which gives them plenty of time to react and adjust to the feedback. The communication will then be based on the conveyance of character qualities instead of superficial chance, and any offense will result from perception, and not from presumption.
Put another way, we should be slow to anger and abounding in love, enduring all things, and speaking kindly to one another. Well worn advice, but worth repeating.
What I’m saying is that, from a practical standpoint, the archon may be able to know what information to use in its simplified model based solely on the nature of the local shatter points. The idea of the archon is new to me, but it makes a lot of sense. Though I suppose that the archon must also simulate in relevance to what was requested by its superior which in turn are due to the shatter points in the superior’s local space.
-Luke, in a letter to the author
Ahh, sorry for the misunderstanding. Yes, you’re absolutely right! The archon will be able to pare down the simulated dimensions based on whatever is relevant to altering the world state. You are also quite perceptive in pointing out that this responsibility trickles down from higher archons.
Left un-checked, this could result in ALL archons simulating EVERY aspect, as each attribute and dimension of the world is usually relevant in at least one way. However, This tendency is countered by the approximate nature of the simulation involved. Only the largest nodes are likely to be involved in simulation of attributes that are relevant at large scales. Gravity, for instance, will likely not be simulated for individual people floating around a space station. They DO exert gravitational force on each-other in reality, but it is so slight that it can be ignored. The same goes for adjacent pebbles sitting on a hillside. Gravity is only relevant in bulk, not for individuals. They will feel the force of gravity of the planet, but probably not that of the mountain (or that of the sun). The same goes for many other attributes, such as the effect of plant life on atmospheric composition, or the cost of office supplies affecting company policy.
Of course, one COULD require this level of simulation. If, for instance, one was making a gravitational anomaly map of the planet, individual mountains (and composition of the mantle) would be relevant, and need to be determined with higher accuracy. This can be done on the fly, however, and generally need not be continuously tracked.
As long as it maintains the appearance of continuity, much of the detail can be abstracted away, or simply never dealt with at all. And, keeping in mind the importance of perception, and the relatively high shatter points of most non-inteligent systems, even more can be abstracted and ignored.
Proverb solved! In fledgeling: If a tree falls the forest, and nothing nearby cares about that level of sonic energy, then it does not, in fact, make a sound.
However, this can lead to some interesting consequences. If a character is listening intently, they will lower the “shatter point” of their local node with regard to sonic energy. Since sound travels quite a distance, this will put them in touch with a large area, and require anything that makes sound within a wide radius to inform them of such. The same goes with light energy (sight), scents, and so forth. In short, animals (highly reactive agents) and intelligent beings (humans and the like) tend to link the universe together and connect distant locations to the degree that they are aware of their surroundings.
When you have many such agents close together, they tend to form feedback loops, depending on their temperment. Or, put another way, mobs are unstable. This effect is balanced against the speed of light, since distant events can not effect our actions (and vice-versa) at any higher slope than C. This provides a comforting (though rather high) limit on the rate of social interaction, and a very practical way to limit mass hysteria (lower the rate of data transfer, or move people farther apart).
Fascinating thoughts? I hope so!