## Wednesday, June 11, 2014

### Thank-you Albert Andry

All of us remember some teacher or professor who had a disproportionate affect on us.  Today, I am going to write about one of those professors, Albert Andry.

I had the good fortune to run across him in an airport.  I was flying in, he was flying out.  We had overlapping lay-overs.  He was in a talkative mood.  He wanted to talk about tests and testing.

His goal in testing was to get good resolution.  He said it is stupid to only use 30 points of a 100 point scale.  He never wanted to see 100% and he never wanted to see 0%.  Rather, he wanted to see most of the population smeared between 15% and 85% with a few outliers.

He was successful.  The highest score anybody ever got on one of his tests was a 98%.  He had two people tied for the low.  One student scored a cumulative of 40 points out of 300 (three tests).  He had another student who scored 5 points on a 100 point test.  Yes, they tried to answer every question....at least they filled up the blank paper with scribbling.  An important aside is that he was teaching 400 level (senior level) classes in Engineering....classes like Vibrations and Linear Control Theory.  It was not possible to just walk off the street into these classes. You had to pass many prerequisites to enroll.

Albert Andry eventually settled on a format of four questions.  Two of the questions were always straightforward, just like the homework questions.  By his reasoning, if you honestly did the homework you ought to get 48 points on just those two questions.  He figured the computations were long enough and complicated enough that most people would make a sign or a decimal point error, hence the two lost points.

One of the questions would be from the fringes of the body of knowledge.  Esoterica or perhaps something that a diligent student could derive within the time constraints of the test.

The last question was a "trick question".  That is, it could not be answered unless you actually considered the physical problem that was being presented.  Two that I remember were a block on a dry surface with a preloaded spring, define X for all time.  The other problem I remember was a gigantic computational problem, something like, perform a bottom sweep-out of a [U]T [K] [U] where [K] was a 12 X 12 matrix.  In the case of the gigantic problem the zero terms were arranged such that one only needed to multiply one term in the [U]Y by one term in the [K] by one term in the [U] matrices.  The bastard.

He tried to tip the students off about the "trick question" by making the plug-and-chug approach daunting if not impossible.  He wanted to ingrain the habit of having engineers actually attempting to comprehend the physical problem before whipping out the calculator.

A few of his comments stuck with me.

"Luck counts but only fools count on luck."

"Your employer won't pay you the big bucks to tell them 'It cannot be done.'  They pay you the big bucks to tell them three ways they can achieve their goal."

He also commented that the test results were distinctly bimodal.  Approximately 65% of the students clustered around 35%.  And remember, they should have had an easy 48% if they had been honest about doing their homework. The other 25% were clustered around 80%.  The remainder were randomly smeared in the gap.  The same students maintained their place in each mode.  There was not much flip-flopping.

His conclusion was that between two-thirds and three-quarters of the credentialed professionals are "pretenders".  Approximately one-quarter can actually function in their profession as effective problem solvers.

Further, he concluded that it was not an issue of technical knowledge.  The first two problems required rote/regurgitation of the assigned material.  Rather, it was a matter of attitude and fire-in-the-belly.  He did not expect graduation to fix that deficiency.

Thank-you Albert Andry, where ever you are.

1. I've often wondered, how much of actual engineering work is mathematical calculation/computation, how much is intuition, how much is trial-and-error and how much is 'best guess'.

I ask that as a long-time engineering wannabe. I graduated from DeVry in 1972 -- Electronics Engineering Technology. That credential qualifies one for work as a 'technician'; i.e. a high-fallutin' tinker.

2. Best-guess and trail-and-error are bride-and-groom. Trials are wasted effort if A.) The guesses are ill considered and B.) if the "design space" is not thoroughly explored. What is the point of buying 5 containers of ice cream if all five are some variant of vanilla?

Intuition has great value when it is based on valid experience. Good intuition (or back-of-napkin calculations) are good for about two significant digits of performance. Calculation/computation will hone the edge and get the third significant digit.

Intuition is of huge value early in a project. You might enjoy this post http://eatonrapidsjoe.blogspot.com/2013/10/some-limits-to-central-planning.html It is also valuable when people start lying...the BS indicator starts gonging.

I know that it is possible to assign percentages. It is more a matter of different skills leading the parade and the others supporting....with the lead skill changing as the project matures.....or goes in the ditch.

3. Sadly, he probably wouldn't be allowed to teach in today's environment with that 'attitude'... Thankfully he probably DID weed out some wannabes though!