<<< Date Index >>>     <<< Thread Index >>>

Re: OT: learning curve



On Tue, Nov 04, 2003 at 09:59:56AM -0500, David T-G wrote:

> Dave, et al --
> 
> ...and then David Yitzchak Cohen said...
> % 
> % On Mon, Nov 03, 2003 at 11:30:05AM -0500, David T-G wrote:
> % 
> % > % - Firstly, and to your point, given that everyone "knows" that a steep
> % > %   learning curve is a hard one, people won't know what you're talking
> % > %   about if you refer to a shallow learning curve
> % > 
> % > So?  "Everyone" "knows" that Outhouse is a mail program, but not many
> % > here (and definitely not I :-) are content to go that route or even let
> % > the assumption stand.
> % 
> % Did anybody ever show you a reason to believe Outlook is a bona fide
> % email program?  I know, for instance, that DOS and Windows fail miserably
> 
> People send mail with it all of the time.  No, that isn't good enough for
> me, but neither is using the shape of the graphed curve to apply a third
> criterion to the definition of the term at hand.

The shape of the graphed curve (if you're hoping to get on top of it,
that is) matters a fair amount - anybody who climbs rocks knows that ;-)

> % Our definition for a learning curve seems to make at least as much
> % sense as "the original one" (see my related post for more on that), is
> % more compatible with the reality of "steep curve == difficult to climb"
> 
> When was the last time you climbed a curve?  I only climb hills (and
> rocks, as noted below).  I would argue that the most common physical
> defintion for curve is 'a relatively smooth bend in a road or other
> course', while a hill is obviously 'a well-defined natural elevation
> smaller than a mountain' or 'an incline, especially of a road; a slope'.

That's a good point :-(

What's more, if we plot my own version in the Mathematician way (putting
the function along the y axis) rather than the Economist way (putting
the function along the X axis), we end up with effort along the x axis
and knowledge along the y axis.  In fact, my real-world example (what
you know vs. how much you can do) suffers from exactly the same problem.
Sadly, I think your position makes much more sense than mine, even if
nobody ever steps forward with some evidence supporting the first-use
claim of the URL.

> Of course, Rene's attempt to do so got us into this mess in the first
> place ;-)

<shot type="cheap">
Yeah, well, he's good at getting us into messes ;-P
</shot>

> But the graph is not meant to literally represent some hill (or cliff)!

I'm still a tad uncomfortable with the term "steep" being abused to mean
"easy to climb," but I'm inclined to believe the "original" definition
is still better, anyway (for reasons I already mentioned above).

> Just because we call it a bell curve doesn't mean that we expect to be
> able to strike the paper (or chart or computer screen or ...) and get a
> ringing tone!

...but it does mean we expect it to look like a bell - if somebody showed
you a graph that looked like a peeckock (spelling?) and called it a bell
curve, I'm sure you'd stare at him and wonder what planet he fell from ;-)

This point is rather moot for our purposes, though, as I already agree
that hill != curve, and it doesn't seem like anybody else is arguing my
side anymore ... oh, well. . .

> Here, think of it as a rocket's trajectory instead of a curve.  You want
> to be on a fast rocket which goes up quickly instead of in a little plane
> which has to struggle to gain altitude.  Now think of which one is 'better'
> in terms of the goal of getting 'up there' to a certain knowledge level.

That analogy sucks, IMHO, since you don't think of a "curve" when
goin into outer space, either.  If anything, an airplane is a far more
pleasant ride.  As I said, though, this is all so much irrelevance until
I can think of an argument for my old view that actually holds together
under scrutiny.

> % Incidentally, there's another common real-world use for the learning
> % curve idea: what you can do vs. how much you had to learn.  Notice that
> % the curve resulting from this derivation agrees with the curve immediately
> % above, not with "the original."
> 
> Sure.  I can think of common real-world uses for lots of things that
> aren't correct.

This particular real-world use is THE only major application of the rule
(and it's guaranteed to be 100% correct by all academic definitions -
since it's THE use for which the whole idea of a learning curve was
invented), so any definiton of learning curve which contradicts this
real-world use CAN'T be correct (since it'd be implying that the guy
who coined the term screwed up, essentially inventing the wrong term
even for his own use).

As it happens, merely flipping the graph along the xy axis makes it
agree with your definition.  Flipping the graph is justified (required,
I'd say) by the fact that we're not economists, so there's no reason
for us to work backwords.  I'm not quite sure why I didn't think of this
problem earlier. . .

> Were I not so picky I might even say "I could care less"
> about them (except I *am* picky and so I *could* care a lot less but not
> in the way that most people don't when they say that).
> 
> 
> % 
> % For times when the original definition makes sense and the common one
> % doesn't (e.g. Web == what most people call web site; Web site == what
> % most people call data center; etc.), by all means use the original (and
> % give footnotes to define words, if necessary - just like you already do
> % when you use technical terms that your audience may not understand).
> % For times when the original definition (if this even was the original
> % definition) makes less sense than the common one, I'd lean towards the
> % common one in the absense of any evidence to the contrary.
> 
> To each his own.  I tend to lean toward Doing It Right whenever possible
> and helping others discover what's right along the way.

I prefer to Do It Right (TM) too, but you're abusing the trademark
if you consider "right" without proof of original use.  So far,
nobody has shown even substantial evidence - much less outright proof.
I can't possibly call something "right" without proof of original use.
The best I can do is call it "apparently better" than the common use,
and that's exactly what I do now.

> Why else would I
> have
> 
>   http://justpickone.org/hoaxinfo.html
> 
> and take the time to follow up on urban legend emails?

Incidentally, I couldn't find individual followups to individual emails
(excepting your brother's outrage against them in general) on your Web.
Everything's just pointers to the rest of the 'net.

 - Dave

-- 
Uncle Cosmo, why do they call this a word processor?
It's simple, Skyler.  You've seen what food processors do to food, right?

Please visit this link:
http://rotter.net/israel

Attachment: pgp6mhDSQzFWF.pgp
Description: PGP signature