Alicornutopia

Witches
So there is a new thing for Azurite: witches! They do magic! Woo~

Now, what they actually do here is "probability manipulation." If you tell someone "think of an elephant," they will think of an elephant. However, there is a nonzero probability that they will think of an elephant regardless of this. A witch can nudge this probability that way a bit farther, and cause things to happen that way. The simplest way witchcraft expresses itself looks like luck, the dice rolling a bit more favourably than they otherwise would. A witch has a more instinctive notion of what their actions will cause, as well as access to a "bird's eye view" of the web of causality. Witches who are good at following threads of causality are almost precognitive with them.

Doing the manipulations is simultaneously intuitive and difficult: when a witch does magic, their and other people's brains try to "elide" over it, finding increasingly ridiculous justifications and explanations about how that was not magical at all, no, sir. The more improbable the witch's desired outcome is, the harder it is to focus on it and remember what they're doing while they're doing it and cause it to happen, and the more readily everyone's brains will find a way to rationalise it into something that fits their physicalist intuitions. A witch could make someone trip, or think about something, or make a coin come out heads ten times in a row, or make an object float, and people would find ways to rationalise it all away - and the level of difficulty here is so that making an object float for five seconds is all but impossible on concentration alone. But the witch will, intuitively, pull strings to make things work better for them, nudging plans and people along their way without even noticing it. Witches tend to be likeable, lucky, popular, and often get what they want, even if they don't notice they're doing it all.

A witch can also bind causally-neutral actions to certain effects. For example, a witch could (in principle, if not in practice) determine that when they swish and flick this stick of wood and say "Wingardium Leviosa," the object they're pointing at will float according to their will. Binding like that is always harder than actually performing the magic itself, more easily forgettable/"elidable." However, once it is done, whenever anyone at all performs that action the bound magic effect will be done, and it will not be forgotten or elided over. A witch can also use that and ramp up, building rituals made up of different parts for larger effects, binding the small parts to smaller effects so they don't forget what they're doing partway through. Therefore, witchcraft is best at creating effects that are made up of a variety of smaller ones.

The difficulty is linked to the relative probability of the action and the effect. A witch could bind, say, the action of putting a pen (any pen) inside a sink then picking it up and biting it to the effect of someone stumbling and tripping. Given that this action is very unlikely to happen in general, and tripping happens fairly often comparatively speaking, it's not hard to bind. It becomes even easier to bind if the witch instead binds the action of putting a specific pen inside a specific sink and then biting it. That's also the scale at which witches are most comfortable acting: nudging dice, making people trip, causing stray thoughts, air drafts, this kind of thing.

Another use for binding is repetition: Making a whole crowd trip by magic, even one person at a time, is very difficult and if a witch tries naively they'll eventually forget what they're doing. However, if they bind a flick of their wrist to someone stumbling, they can do it repeatedly to each individual person on the crowd and make everyone trip without forgetting anything.

The witch who originally bound an action is the only one who can unbind it, and it is as difficult to unbind as it was to bind.

So if somebody were to bind the action of pointing any stick of wood at a person and saying “avada kedavra” to having that person die, and were to then idiotically test this on themselves, it would be impossible to unbind?

Pretty much, yep.

Mind you, doing that would be quite impossible given the probabilities involved. This magical system is such that killing someone magically is best done by cleverness (making them stumble off a cliff) or large rituals made of several parts (this bit makes these neurons misfire, this bit makes them choke on their saliva, ...).

Oh, does recursive binding not work? And I was thinking specifically that killing somebody is an instantaneous action which might be easier to focus on because you can safely lose focus once it happens.

I'm not sure what you mean by recursive binding.

Killing somebody is not the kind of thing this system treats as a 'primitive action" if they're not literally in the ICU with two feet and a hand in the grave already.

(Not that there are really "primitive actions" in that sense, it's just that the probability of that happening randomly is by default too low.)

Recursive binding presumably meaning binding a tool to produce bindings on other tools?

I don't think it'd work if you wanted versatile extra bindings, and I think it would be prohibitive to have it happen anyway since the random chance of any object becoming bound without the intervention of a witch would be… negligible? So if you wanted to be able to alter that probability, it'd have a high difficulty to get it to "likely", and so it'd have an even higher difficulty to produce a binding on something else that produced that effect.

Or is that not what you meant?

I meant recursive binding invocation, not recursive binding production. I bind saying the words to actions which themselves are bindings for other things. Making a binding which was more probable to invoke than for the action it was bound to happening naturally would be difficult, but it would let you break complicated results into smaller pieces.

If I understand you correctly, you mean setting something up like a chain, with "action X" bound to "result Y" where that result is also "action Y" which has been previously been bound to "result Z" so performing "action X" ends up causing "result Z"? If that's the case, chaining like that is not naively possible, a bound action needs to be performed by a person. It is possible to bind "action X" to a result that will cause some other person to perform "action Y" but if "action Y" itself is pretty unlikely to be done a priori (which it normally will be, for bound actions) then that doesn't really seem to help.