Serious Problem With New Student Assignment Plan
School Board votes on the plan on Wednesday, June 17, 2009 - contact them
ASAP!
(contact info is at the bottom of this page)
My objection to the proposed Student Assignment Plan centers around bullet # 1
of the Processing Applications for Attendance Schools section (and the
corresponding identical line in the Option Schools section), which states:
"Process all first choices first, then process all second choices, etc."
This sounds simple and fair. It is neither. It is a change from the
old choice algorithm with enormous and terrible implications. It makes it
all-but-impossible for most families to get a reasonable shot at a popular
school. The implications are subtle, but absolutely disastrous. I strongly
recommend (as a parent and as a professional software engineer) that we simply
substitute the corresponding bit of the old algorithm, as in: "For each
application, consider the first choice; if no seat is available, then consider
the second choice as if it had been listed first."
I need to emphasize just how incredibly disastrous the proposed
choice algorithm would be. It essentially makes the whole
choice process nearly worthless for most families. I disagree with the
assertion that more parents will be able to understand it, and that the new
system punishes only naive players. I think the implications of the old
system are not difficult to explain; the school district just didn't do a good
job of it. They explained how it worked, but didn't sufficiently emphasize
the fundamental conclusion that listing a long shot doesn't hurt you. The
most important thing to understand was that you should just go ahead and list
the schools in order of decreasing preference - put the one you most want at
the top of the list, and then your next choice, etc. The primary point of
confusion was that in the distant past (1999? 2000?) the system was
different, and listing a long shot could hurt you, and many people still
thought (and said) that. All you really had to know in recent years was that
listing a long shot school (that you wanted) could NOT hurt you. If anything,
I think it's harder to explain the implications of the proposed new system
(don't list a school first or even second unless you really think you have a
good shot at it), never mind explaining how to maximize your child's chances
of getting into a desired school. This is crazy.
The biggest problem with the proposed algorithm is that you effectively have
at most one chance to get into a popular school, and that ONLY if you list it
first. And doing so significantly hurts your chances for your second and
subsequent choices, unless they are very unpopular. So first off, you either
have to be very fortunate in where you live (if you happen to want your
attendance area school), or you have to do a lot of research and guesswork to
figure out what school to list first. But even if you do that, you
effectively only have ONE shot. No matter how much research you do or how
accurately you can analyze and guess what enrollment patterns will be, there
is no effective way to list an acceptable backup choice if it is also popular.
For example, for my son, we chose to list Thornton Creek first, followed by
Salmon Bay, followed by various other schools. We knew our chances for
Thornton Creek were not very good (we lived outside the cluster), and Salmon
Bay was not likely either. But we could take a shot at them, knowing that
they were the best choices for our son. With the proposed new choice
algorithm, we probably could not have listed either (or would have been
extremely unwise to do so). Why? Because with long odds against either of
our first two choices, by the time the system got around to processing its way
through everyone's THIRD choices, only unpopular and/or distant schools are
likely to be left.
I wonder about the reason for this proposed change. Here's what I can think of:
-
It's easier to implement in the software. Okay, but...not really. The old
system wasn't very hard to implement - a trivial difference for a professional
software engineer. I am one (résumé at http://www.cybermato.com/chris/resume/),
and I'm the one who did an independent analysis of the old software a couple
of years ago (see http://www.cybermato.com/projects/school-assignments/
for more info). And in any case, even a moderate difference in implementation
complexity would be worthwhile, compared to the harm that would be done by the
proposed change.
- They are deliberately making the choice process so risky, painful, and
useless that few will actually use it. Possible, but...sounds paranoid to me.
I don't think the school board or the school district staff are operating that
way. I think they have some very difficult problems to solve, and I wouldn't
want to be in their position. Let's move on.
- The new proposed system sounds very simple, straightforward, and fair, if
you don't look closely enough at the implications, and most people involved
have not analyzed it carefully enough to spot that. I'm putting my money on
this one. If we point out the huge problems with this proposal, and the fact
that the old algorithm was designed the way it was for a good reason (because
it was fair and worked well), I think they will be willing to do the right
thing, and make the minor change in specification with little (if any) cost
and huge benefits for the parents (and the school district staff who won't be
plagued by parents whose kids were cheated by a poorly designed algorithm).
So, is there any disadvantage (from the perspective of the parents) of
applying the old choice algorithm (wherein if you didn't get your first
choice, then you were considered for your second choice just as if it had been
your first) to the new SAP? No!! In fact, it's better for everyone. It is
in fact EASIER to explain than the proposed new system, if you focus on the
fundamental issue (it's safe to list any school you want, and you should list
them in your true order of preference - there is no harm in listing a long
shot). So it's better for the parents and most importantly the kids. It's
better for the school district as well, because they don't have to try to
explain the new system to a population that is finally getting to understand
the old one; they don't have to try to explain the counterintuitive and
ultimately horrific and torturous implications of the proposed new algorithm;
they don't have to deal with confused and disappointed parents who try to use
the system in good faith and are failed by the new algorithm. (True, there
will always be disappointed parents since some schools will always have more
demand than available seats, but if it were me, I'd rather be dealing with
that knowing that it was a fundamentally fair system and not everyone can win
than with the proposed new system where it's just unfair to begin with.)
Perhaps someone will argue that making this change will impact the
implementation schedule for the software. I doubt that. But even if so, I
will happily offer my services to the school district at absolutely no charge
whatsoever. I would be happy to submit a sample implementation, with source
code, tests, and the results of large-scale simulations to validate the
results. (If anyone would like to chip in to defray my costs in doing that,
please get in touch with me via the contact form at http://www.cybermato.com/contact/.)
It's not that hard to do. I know - I've analyzed the actual software that
runs the old system!
See Also
The following are also worth a look:
Who Are You, Anyway?
See more information about Chris
MacGregor and other Cybermato
Consulting projects.