Hacker News new | past | comments | ask | show | jobs | submit login

Ok, I'll bite. What happened, then, that's any simpler than a 3rd person? Because I have no idea what happened.



Bite what? I'm not proposing any explanations. I'm pointing out that it doesn't make sense to assume a third person, for whom we have no evidence, was in the car.


My understanding is that you use Occam's razor on a list of possible answers. A third person isn't less likely unless there is some more likely choice. I don't know what that would be.

Edit: You're sort of glossing over that "nobody in the driver's seat" is unusual by itself. All explanations I can think of for how this happened are unlikely.


You don't need to have a "list" of explanations. Any explanation would either assume there were more than two people in the car (why stop at three?) or it would assume there were no more than two people in the car. Any explanation that assumed there were more than two people in the car would "multiply entities" more than any explanation that assumed there were only two people in the car, so the Razor would always cut out the former in favour of the latter.


>> Edit: You're sort of glossing over that "nobody in the driver's seat" is unusual by itself. All explanations I can think of for how this happened are unlikely.

That means you can't think of an explanation for it, not that there isn't one.

Again as I said in other comments, if you find yourself thinking "I can't think of a better explanation", that's a sign you need to sit down and think of a better explanation.


You can't describe how one explanation doesn't fit with Occam's definition of simpler, without saying what the other explanation is.

That's the whole point of Occam's Razor: "simpler" (or any other adjective we might use there) requires a comparison.


The Razor, as originally stated by William of Occam, is: "entities should not be multiplied without necessity" (see wikipedia) [1].

Clearly, assuming a third person was in the car is "multiplying entities without necessity". Note also that the original formulation of the Razor makes no mention of hypotheses, competing or otherwise, or of any requirement for comparisons.

Even going by the more modern interpetation of (stated informally) "prefer the simplest hypothesis", a comparison can be made "on the fly": any hypothesis that needs a third person in the car is less simple than a hypothesis that doesn't need a third person in the car. So we can prune away the entire branch of the hypothesis search tree that begins with "suppose there was a third person in the car" without wasting any time considering those hypotheses even if we don't yet have any competing hypotheses.

And, btw, we sure do: there were two people in the car and they caused the crash. That's a competing hypothesis. I don't know why people keep saying "there's no competing hypothesis". That you don't like that hypothesis makes no difference.

You'll notice also that the more modern interpretation ends up agreeing with the original: any hypothesis that needs a third person in the car is "multiplying entities without necessity" (the entities being the people in the car), whereas any hypothesis that doesn't need a third person is leaving the number of entities (at least people in the car) well enough alone.

_____________

[1] No, please, really do see wikipedia. I think it will help clarify much confusion about what the Razor is. For instance, there are five or six different formulations of the Razor, each with its own, multiple, interpretations. But I can't see any one that "requires a comparison". That sounds to me more like a Calvinball rule. If you want to say I'm "using it wrong", then start by clarifying which formulation and what interpetation of it you are using, as I have done throughout this thread.


> (see wikipedia)

> Note also that the original formulation of the Razor makes no mention of hypotheses, competing or otherwise, or of any requirement for comparisons.

Well, the Wikipedia article states, in the very first line:

> ...or more simply, the simplest explanation is usually the right one.

and further down it defines what a "razor" even is:

> The term razor refers to distinguishing between two hypotheses either by "shaving away" unnecessary assumptions or cutting apart two similar conclusions.

So both of these do seem to directly relate to comparisons. It's not simply a misunderstanding of the term to say that Occam's Razor involves comparisons of competing hypotheses.

> any hypothesis that needs a third person in the car is less simple than a hypothesis that doesn't need a third person in the car.

The trouble is you need to define what you mean by "entities." Obviously this does not need to literally refer to bodies. It means facets and complications of the hypothesis.

So we need to decide which hypothesis requires more entities: one that supposes the possibility of a third person, or one that supposes the possibility of people tricking an autopilot to do what no one believes it ought to be able to do.

The point is, we need something beyond what we have been told. If it were simply "drunk man crashes into tree," we need no further explanations, no further entities. That story explains itself. But if the initial story isn't by itself sufficient -- "autopilot drove itself at high speed in a narrow cul-de-sac with no one in the driver's seat, against all programing and prior experience of Tesla cars" -- then we must hypothesize additional entities. Wonky programming. Hacked system. 160 lb weight on the driver's seat and fake hands on the steering wheel. Third person.

I don't know that the third person hypothesis is, in fact, simpler. But I don't think you can say that Occam's Razor suggests that that should not be a rational hypothesis. I don't know how many additional entities would be required to get a Tesla to do what was suggested in the article.


Yes, wikipedia gives different interpretations of the Razor that, as you say, "directly relate to comparisons". None of them requires a comparison. I mean, you won't find that rule anywhere: "to use the Razor, you must make a comparison, otherwise you can't use the Razor". That is Calvinball.

>> The trouble is you need to define what you mean by "entities."

I think this is splitting hairs. A person in the car is clearly an "entity". Assuming a third person in the car is "multiplying entities".

>> Wonky programming. Hacked system. 160 lb weight on the driver's seat and fake hands on the steering wheel. Third person.

... or a 160 lb bag of stones. Or three giant rabbits. Or a pair of labradors. Or... etc.

We really don't need to assume any of this. The two people in the car and sufficed to have caused an accident. Assuming more people, or anything else, is multiplying entities without necessity.


> That is Calvinball.

Stop with the condescension. The wikipedia article literally defines "razor" as "distinguishing between two hypotheses." If you disagree with that, edit the article, but don't tell me to read the article ("No, please, really do") and then dispute it.

> The two people in the car and sufficed to have caused an accident.

You keep saying that. I don't know it's true. Do you know it's true?

Like others have said said, this would go against every prior experience with Teslas, and our understanding of how the system works. It would be the first Tesla, ever, to have gotten into an accident on a narrow residential road with no one in the driver's seat.

You need something else in the explanation, even just "the system actually works in some other way." That is also multiplying entities. That was my point, which you missed.

You can't simply state, ipso facto, that the description given "sufficed to have caused an accident," when numerous people have disagreed with that.


There is a misunderstanding I'd like to clear up. By saying "Calvinball" I am not being condescending. I am calling you out for trying to invalidate my entire line of reasoning by introducting a new rule to Occam's Razor, that it must be used in a comparison (between two hypotheses). This is a rule that is not there in the original formulation of "entities should not be multiplied beyond necessity". It is also not there in subsequent interpretations, including the interpretation you quote from wikipedia. It's a rule that you came up with and it is a rule that you came up with for the sole purpose of helping you win the internet argument. So it's a Calvinball rule.

Besides which, as I said before, there do exist many hypotheses that we could be making and the Razor is useful in prunning many of them away before we have to waste our time formulating them and exploring them in detail. For example, all the hypotheses that start with "Suppose there was a third person in the car" need not be considered because "there was a third person in the car" is "multiplying entities without necessity", and so hypotheses that make such an assumption are less likely to be true than hypotheses that do not.

The misunderstanding I'm pointing out and that I advised you to read wikipedia to clear up, is that you seem to understand the Razor as requiring at least two competing hypotheses to be fully formed to the point that they could be written down, perhaps even in a formal language. There is no such requirement. That's your Calvinball rule, made up on the spot and for the sole purpose of winning the internet argument. Or, perhaps, only for the purpose of forcing me to waste my time by considering all the unnecessary hypotheses that start with "suppose there was a third person in the car".


This is so frustrating, because you're not seeing my point that, without a simpler hypotheses, it doesn't make any sense to dismiss hypotheses on the basis of the Razor alone.

And, again, from what everyone understands of how cars and Teslas work, the initial description doesn't have enough elements to be a complete story.

This is honestly what this whole conversation sounds like to me: Suppose we find a baby stuck high up in a tree.

Me: "How do you suppose it got there?

You: "No idea. All I know is that there's a baby in a tree. The baby got there."

"Perhaps someone put him there?"

"Occam's Razor states you shouldn't multiply entities unnecessarily. Another person would be another entity."

"Ok... I mean, maybe there was a flood and-"

"And maybe a flying pig put him there! What don't you get about Occam's Razor???"

"So how do you think he got there?"

"I have no idea."

"Ok... So, I think that someone putting him there is the most likely -"

"How do you not understand that Occam's Razor doesn't require comparisons!"

"WTF?"

At this point, I don't give a damn about what Mr Occam said. This nitpicking over the original formulation is completely irrelevant. All I've been trying to think from the beginning is what the most likely explanation is.

A third person may well NOT be the most likely, or simplest explanation, but since you haven't provided one this conversation has been pretty useless.

Occam's Razor says you shouldn't multiply entities unnecessarily. That last word, "unnecessarily," I realize in retrospect, is the one I've been focussing on this entire time, while you've been talking about "entities."

If the initial story isn't sufficient -- and the fact that this would be the first Tesla ever to drive at high speed in a narrow cul-de-sac with no one in the driver's seat suggests it isn't -- them we MUST multiply entities. Bugs, trickery, GPS malfunction, whatever. We can then argue about which entities are reasonable and which aren't, if we like, but that's separate from the fact that we need to multiply them.

(And, indeed, it is this word "unnecessarily" in the original formulation that does require simplER hypotheses, which is why that wikipedia article that you told me to read all the way through is chock-full of reference to "simpler" or "competing hypotheses." Without another explanation, you have no idea if the additional entities are necessary or not. All explanations require at least one "entity," except perhaps the Big Bang. The particulars of the case determine how many entities are necessary and sufficient.)


I am seeing your point. You say that there needs to be a comparison between at least two hypotheses and then the Razor will help us choose one if it's simpler than the other.

There are many competing hypotheses to choose from in this case. They include all the hypotheses that assume there was a third person in the car; and all the hypotheses that do not assume there was a third person in the car. Any hypothesis that does not assume there was a third person in the car is "simplER" than any hypothesis that assumes there was a third person in the car. Assuming there was a third person in the car is to multiply entities beyond necessity, i.e. as you say "unnecessarily".

You can check my comments in this thread to verify that I've said this many times. The comments that wonder about what "entity" means are yours, not mine.


> Any hypothesis that does not assume there was a third person in the car is "simplER" than any hypothesis that assumes there was a third person in the car.

Nonsense.

A hypothesis that there was a third person in the car is simpler than a hypothesis that there were only two people in the car and an advanced animatronic ice sculpture in the driver's seat which melted after the crash. (I remember my minute-mystery solutions.)

A hypothesis that there was a third person in the car is simpler than a hypothesis that there was a rare combination of a bug in the software, a GPS malfunction, a solar flare, and a paint spill that looked like a painted lane marker.

There are an infinite number of hypotheses that don't contain a third person that are more complex than the hypothesis that there was a third person.

> You say that there needs to be a comparison between at least two hypotheses and then the Razor will help us choose one if it's simpler than the other.

If you think that was my point, you didn't read my comment above at all. That was not my point. It was that, IF the known facts of a story aren't sufficient cause to result in its conclusion, then there is required to be at least one more cause than has been explained.


>> There are an infinite number of hypotheses that don't contain a third person that are more complex than the hypothesis that there was a third person.

Yes! You're right, and I'm very excited now because you seem to understand how it works.

Note that all those hypotheses that you bring up also "multiply entities beyond necessity" - so they are "more complex" than any hypotheses that don't, and we don't need to examine them, we can just prune them out without even having to state them.

So to correct what I've been saying above that was indeed too general, "any hypothesis that does not assume there was something else in the car is simpler than any hypothesis that assumes there was something else in the car" ("something else" meaning "something besides what was actually found in the car"). Please correct me again if you think I'm still wrong.

I think we're getting to something we can agree on now, yes?


Yes, but you haven't accepted the central premise of all my posts yet, which is that entities should only not be multiplied "without necessity."

If the initial known facts of a story aren't sufficient to describe an event (a baby in a tree, a car doing something it never has before) then the actual cause must involve additional entities, whether a person or a solar flare.


>> Yes, but you haven't accepted the central premise of all my posts yet, which is that entities should only not be multiplied "without necessity."

This is not the central premise of all your posts. For most of our interaction the central premise of your posts was that I'm misunderstanding the intended use of Occam's Razor. In recent posts you concluded that the reason I misunderstand it is that I don't take into account the "necessity" part of the Razor's original formulation.

Yet, I have constantly said that we don't need to assume that there is a third person in the car because the two people we already know were in the car suffice to explain the crash and that therefore assuming a third person in the car is multiplying entities beyond necessity.

In fact, I keep repeating the "beyond necessity" part like a broken record, so how come you're now insisting that I'm missing that particular point?

I think what you are saying is that you "need" to assume a third person in the car, otherwise you can't explain what happened. Well, that is a "need" in the same way that "I don't have an iPhone, therefore I need to get one" is a "need". It's not so much a "need" as a "want". You can't think of anything better and you want to explain what happenned, so you make up some entity that must be responsible for what happened. So you have an explanation you're happy with and your "need" for an explanation you're happy with is satisfied.

However, that's still "multiplying entities beyond necessity" because you may "need" an explanation, but you don't need a third person in the car to explain the crash.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: