Chapter 5 Page 10

Keith Schiffner

oh dear.

kiva

does the tracking spell look for a persons aura? if so then maybe it can not recognize Arianna because her color changed? and by that change became another person, and is not the one he remembers anymore.

kiva

this is just a guess

Hakira

Ok, one of these people need to suggest sleep, and sleeping pills/magic…

Dart Devious

Cue dramatic music!

David Argall

Zera seems to making a logic error here. If tracking spells do not work on the dead, that does not mean a failed spell means the target is dead. It might have failed for any other reason. Here Zera tells us such spell can fail even under near ideal conditions [not often, but “almost completely” means it can fail] .
We already have the idea it can fail because the target has changed so much that the target is “dead” and the spell fails. A more general idea is that a very large number of people would not want to be tracked, in particular those that our girl seems to be working with. That means there has been heavy and talented work on these exceptions and there may be quite a few ways she could be dead to the spell and live on [or not live on. Undead are another possibility.]
However, Zera should try tracking while excluding all knowledge he got from contact with the child. The knowledge of the child has several possible ways to work against finding the adult mage.

Happyroach

I think his implication is he was close enough to her that it should be nearly infallible.

Also it may be that the spell isn’t giving a “Cannot give accurate location of person” response, but a “Not Reading Aura Error: Abort Retry Fail?” message.
- David Argall
  
  Possible, but the way he talks, the spell does not work that way. If it did, he would likely say something like… “The spell would say X is within 1 inch,or foot, or mile or thousand miles of here, but says ‘no such person’ if he is dead.” Instead his wording implies the spell gives the same error message any time it fails to locate the target.
Daniel Demski

“Almost completely infallible” is close to “completely infallible”. If it were completely infallible on the living, failure would guarantee the target is dead. Since it is almost completely infallible, failure almost guarantees the target is dead. So, the reasoning is not justified by logic, BUT it’s quite sound according to probability theory… the test provides good Bayesian evidence.
- David Argall
  
  2 problems: “Almost infallible” is apples to the oranges “almost sure to work”, and [related] the math don’t work that way.
  A guard on the cliff is almost infallible in spotting approaching soldiers. No matter how they sneak and hide he can always spot them, … unless it is night…He sees them under 99 out of 100 possible conditions…just that one little exception… And it may be a very little exception if we can be sure they will not come by night, or a great big one if they know the guard can’t spot them at night. But that means our almost infallible guard has a chance of seeing approaching soldiers of 100% or 0%.
  
  Getting a little closer to our case… We have a unit of 100 men, of whom 95% answer roll call at any given time, a close to infallible way to know who is present. Now a dragon passes by and has one unlucky fellow for lunch. We need to find out who died and call roll. 95 answer and 5 don’t. Very normal, but we know 4 of those 5 are alive and only one was chow [or maybe all 5 are alive and it was some visitor who learned too much about dragon diet.] Our almost infallible “spell” doesn’t even tell us someone is more likely than not to be dead. And if we use the same figures for our case, the “dead” girl has a very high chance of being alive.
  
  We can vary our figures, but each such test will have both false positives and/or false negatives and the false can be huge in percentage.
  - Daniel Demski
    
    Actually, the math does work that way! It’s called a “Bayesian update”.
    
    I’m not completely sure I see what you’re getting at with your example, so correct me if I missed it; but it looks like what you’re saying is that we could take the test as evidence, and therefore become more certain that Arianna is dead; but because of what amounts to a false positive (if we look at it as a test for deadness), we will become more wrong than we were.
    
    The concept behind Bayesian evidence is that this is an acceptable price to pay. Most of the time, when the spell fails, it’s because the person is dead. So when we learn that the spell fails, we “update” in the direction of the person being dead. We should never become completely certain of it, because of the existence of false positives; and the more probable false positives are, the less evidence the test provides.
    
    In the example with the dragon, suppose first that we are 100% certain one of our own men died. We and 100 other people saw what the dragon did quite clearly. Then the test of doing the roll call and hearing 95 men answer gives us a lot of information: it narrows down the possible deaths to just 5 men.
    
    Suppose, instead, that we are a mere 50% certain that one of our men was taken. Well, hearing 95 men respond still assures us that none of those men died. But suppose one of the habitual non-responders is named Adam Kneetrodder. Before performing roll call, we have a 50% probability the dragon didn’t eat one of our men, which implies Adam Kneetrodder is alive. In the remaining 50%, Adam is one of 100 men who are equally likely to be eaten; and 1% of 50% is half a percent – so what I’m trying to say is that “before the update”, we had an overall 0.5% probability of Adam Kneetrodder being the person the dragon ate.
    
    Once the roll call is taken and we hear 95 responses, there is still that 50% chance that it wasn’t one of our men; but in the other 50%, the situation is now different. Adam is one of five men who could have been eaten- and one fifth is 20%. 50% of 20% is 10%. So “after the update”, we believe there is a 10% chance Adam Kneetrodder has been eaten. This is rational given what we know about the roll call and our uncertainty about what the dragon did.
    - David Argall
      
      Which is not what Zera was saying. His language was that the tracking spell was decisive, not a mere 10% chance “Adam” was “eaten”. [We can be generous and say he meant 95% instead of 100%, but he means far above the 10% we are calculating here.] He should not be at all that sure on account of the spell. [Of course, even very bright people jump to very stupid conclusions, especially when they are obsessed as he seems to be, but he is still highly likely to be wrong.]
      On the next page, Zera says he has cast the spell many different ways, but the assumption of greater accuracy [.05x.05x.05…] is suspect. If all we do is call the roll again, say in reverse order, we can easily get the same 5 no answers [2 are sleeping, 1 dead, 2 engaged in criminal behavior and don’t want anybody knowing where they are] and the chance remains at 20%. Since the spell seems easy to cast, repeat casting looks to be no problem, but of no use since it merely gets the same answer in most cases.
    - Daniel Demski
      
      To me, “almost completely infallible” probably means something like 99.9% reliable. I’m going to go ahead and assume he’s correct about this — so, the error rate is something like 1 in 1000, including whatever bizarre cases might be involved specifically with Arianna.
      
      To start with, Zera had some sort of convincing evidence that Arianna was dead years ago. Let’s say he was already 98% sure Arianna was dead. If the tracking spell is actually independent evidence, it moves him to 99.998% certainty. But you’re right to point out that evidence doesn’t always combine like this. So suppose the 98% certainty leaves a 1% error for something like “she was rescued by someone” or “I was mistaken about what I saw”… basically, mundane sources of error. And the other 1% is there to represent “something very strange is going on” or “I’m deeply mistaken about the way the world works”.
      
      Well then, the mundane sources of error will tend to be ruled out by the spell at its usual rate (error in 1 out of 1000 cases). But the “very strange” sources of error are different. Many of them wouldn’t affect the tracking spell, but many of them would… so I’ll just simplify things and pretend all of them affect the tracking spell. Well, checking the tracking spell should still move him from 98% certainty to 98.999 percent certainty; practically 99%.
      
      Additional tests, like he provides on the next page, keep providing a little bit of evidence by ruling out some of the “very strange” cases. Very optimistically, he might work his way up to 99.5 percent certainty.
      
      So it’s true that he should be reserving some small amount of uncertainty to account for really weird cases. But the tracking spell still provides a good amount of evidence; and given the utter rarity of whatever happened to Arianna, he’s not making a mistake by taking that evidence seriously.
      
      Except there *is* other evidence available. Firstly, someone has appeared using the name Arianna Marcel. The chances of this happening while Arianna is dead are low; Zera mentions that the name is rare, and that basically he can’t think of a reasonable scenario for a bounty hunter to pick it up as an alias. So this should provide some evidence that Arianna is alive.
      
      Let’s suppose that 90% of the cases where he’d hear her name are ones where she’s alive. He might start at his 98% certainty, then her name comes up in connection with this case. This moves him to 82% certainty that Arianna is dead. Then the 99.9% accuracy of the test could move him back up. If he fails to account for “very strange” cases, that would get him to 99.982%. I think that’s essentially what we’re seeing in the narrative. On the next page he acts like he’s sure Arianna is dead. 99.982% is close enough that I’d call it certainty too.
      
      But if he accounts for the possibility of “very strange” cases, the tracking spell should merely bring him from 82% up to something like 91% (since half of the cases where she’s alive are strange ones). And in fact, Arianna’s name has popped up in the context of “613”, who escaped custody in a “very strange” way! I could make up some more numbers, but it seems pretty clear the tracking spell shouldn’t move him up to 90%. It still counts as evidence, but it’s much weaker.
      
      Still, overall I’m arguing that he should rationally remain something like 80% certain that Arianna is in fact dead. The biggest factor is the probability of a bounty hunter deciding to work under that name. If I had sad this was fifty-fifty instead of one in ten, I would have ended up staying above 96% certainty.

Sabreur

Iiiiiiiinteresting.

I know I should be more empathetic about his obvious sense of loss, but I can’t help but be fascinated by his lecture instead. Is our Arianna the same one he knows? Is she undead or something? Did the tracking spell fail for another reason? Does it have anything to do with why she’s not affected by Vector’s eyes?

Obviously I don’t expect to get actual answers to any of these questions anytime soon. But it’s looking like a fun journey to find out!

Lexi

Do all spells sound like the tardis?

SparkyDragon

Oooh

interesting

Boy, when you do catch up to Arianna the first thing you need to do is borrow that sleeping spell of hers.

Suros

Ah, so that’s what the cone is. Wonder what she tried to track before.

Purphoros

So… wrong. The more information he adds, the more likely the spell will fail. People change! Especially over the course of 15 years. So, the more details you add, the higher the probability that some of the traits have changed, and the spell fizzles.

That is not even accounting for information that wasn’t true to begin with, like when you are too nice to tell your old grandmother that you hate her cabbage soup, or when your little sibling thinks your favorite color is yellow when you are really just picking it to let them have the blue figures.

Almost infallable? Not even close.

sargothwise

The obvious expansion is that she’s a lich who came back to life to have vengeance upon the living