From: Anders Sandberg (asa@nada.kth.se)
Date: Wed Dec 19 2001 - 17:40:17 MST
On Wed, Dec 19, 2001 at 12:33:05PM -0800, Ben Damm wrote:
>
> However, I think the test is truly ridiculous! I understand that the test
> is taken as a whole, that each question can be answered incorrectly (that
> is, against any condition you have) because of some other factor. My
> question is, what kind of profiles can score highly on this test yet not
> be autistic or have Asperger's? What kind of profile can score low on
> this test and be autistic? "Eighty percent of those diagnosed with autism
> or a related disorder scored 32 or higher." Sounds to me like 20% fell
> through the cracks! Why!
Remember that this is not far from all the other "How high is your
EQ?"/"Measure your paranoia potential"/"What is your personality
archetype?" tests you find in magazines. This one seems to be based on a
real test that has been validated (hence those numbers), but most likely
it is just a random subset of questions and has far less statistical
validity than the original.
Creating a workablie psychological test is hard - there are entire
courses or research groups about this at some psychology departments -
and takes a tremedous amount of experimentation, statistics and
analysis. Just look at the effect of cultural factors on the autism
test; I'm certain far more swedes than americans would be classified as
autistic by this test, although there is likely no higher incidence of
autism in Sweden - it is just that the way we would understand and
answer the questions is different from the way an american would.
Normally such tests are translated at great expense to fix such cultural
factors, at least if they are intended to have some validity.
> I'm not claiming that the test is bogus (after all, I scored high, maybe I
> really am abnormal along with the other 95% of the population), but I
> would like to hear some informed feedback on how people think the test
> stands up against a rigorous logical dissection. What kind of delusion
> does someone need to score high or low, and be placed in the wrong
> category by this test? (Heck, maybe even asking this question puts me in
> the "autistic" category, but if that's true, humanity is in serious need
> of more autistic people.)
You can create quite accurate tests with surprisingly bad questions. It
is all an issue of creating a huge test with questions that manage to
cover the relevant area, getting a lot of people to do it, then examine
statistically what answers correlate with which group of people. If it
turns out that all autists answer "purple" as their favorite color, even
if nobody has the faintest idea why, then that question can be used in a
slimmed down test that actually manages to identify many autists. Logic
doesn't come into it. It is all a question of statistics: there will be
some type one and two errors, a good test manages to keep them
reasonably small. But since autism is a spectrum, it doesn't make much
sense to put any rigid limit on what score implies autism and what score
implies neurotypicality.
> Someday psychologists will discover that there is no normal. Then it's a
> not-so-simple matter of letting the rest of the world know.
Psychologists have known that for a long time. But the news spread
slowly because most people doesn't want to hear them...
-- ----------------------------------------------------------------------- Anders Sandberg Towards Ascension! asa@nada.kth.se http://www.nada.kth.se/~asa/ GCS/M/S/O d++ -p+ c++++ !l u+ e++ m++ s+/+ n--- h+/* f+ g+ w++ t+ r+ !y
This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 08:12:40 MST