?

Log in

No account? Create an account

Previous Entry | Next Entry

why engineering is so hard

I don't often agree with sakky over in College Confidential's College Discussion forums, but his response in the Why is engineering so hard? thread really resonated with me. It captures a lot of my past frustration while in school, how I would study hard but not perform well, without being given any good guidance on how to do better. I could study hard, and learn material well that I had access to, but I could not "prepare perfectly" – I could not study in such a way that I would always do well on whatever exam I was given. And for people who did do well all the time, I didn't know what enabled them to do that (and in some cases, neither did they). At any rate, it was discouraging and frustrating, for example when having to defend the fact that despite not doing well, I was really studying hard.

Later in life, I started to question whether these high achieving students would do well regardless of what exam was given. (Assuming reasonable questions, of course – this rules out things like "derive Ohm's Law from Maxwell's Equations in 30 seconds".) For example, if I got to give them an exam based on things I thought were important, how well would they do? And if they didn't do well, would it matter? Suppose I was a world-class researcher/professor – would it then matter? Some people interpret the grade that a world-class researcher/professor gives as a blanket endorsement (or not) of the student, but that endorsement may not be warranted. (Recall a time I've mentioned when a physics professor flunked 60% of the MIT freshman physics students in the fall of 1978. The professor would not change her mind – the grades stood. But it is not clear that all of these students really deserved to fail, as the grading was much harsher than what it had been in recent offerings of the class.)

This also relates to my job searching during my (not so) recent periods of unemployment, where it was even less possible for me to "perfectly prepare" for the interviews. There was just far too much material to cover. It was very discouraging and frustrating having to explain to people why I was unemployed despite having a master's degree in CS, etc. "The interviews are too difficult; there is no way for me to reasonably prepare for them" was not a sufficient explanation for some people (especially who aren't in the computer field).

Comments

( 16 comments — Leave a comment )
fauxklore
Nov. 28th, 2008 10:10 am (UTC)
One thing that is different about engineering (and math / science, in general) versus other fields is the progression of information. If you take six English classes at the same time, only your eyesight will suffer. But if you didn't grasp some of the fundamentals of some of the engineering prerequisites, you will always be playing catchup. You can't do fluid mechanics and heat transfer without vector calculus, for example.

The trickier aspect to explain has to do with problem solving. This is really the important skill in engineering in the real world and is a very difficult thing to test in examinations. Maybe if you had group projects and a TA sat in with each group to watch their process, you might be able to get a realistic grasp of how the students were doing, but that doesn't seem realistic.

I'll also note that the academic emphasis on individual accomplishment is somewhat at odds with the real world. For example, I had something I was working on the other day which I had an idea about an approach to. I was able to approach a colleague (who has a background on one aspect of what I was working on) and say, "can I explain how I'm thinking about this and get your input?" And I made some adjustments to how I was going about the problem as a result. That doesn't happen much in school.
gregbo
Dec. 2nd, 2008 03:18 am (UTC)
Hmmm ...

Some English classes have prerequisites (or at least say they do). Also, not all HS English programs are equal. There seem to be professors out there who complain that their students can't write coherent English sentences, let alone provide literary criticism, etc.
gregbo
Dec. 3rd, 2008 05:31 pm (UTC)
Actually, I thought of a better example from music. There are music classes that depend on the others – you can't take composition without knowing harmony, orchestration, and so forth. Plus there is something to be said for playing at least some of the instruments you're writing for. A composition may be harmonically "correct" but physically impossible to play (by humans). (But unlike engineering, perhaps musicians don't have to rigorously prove this.)
fauxklore
Dec. 4th, 2008 12:01 am (UTC)
Actually, I think the music analogy supports my point about the prerequisites what makes a field time consuming. Musicians spend hours and hours practicing. Playing an instrument 6 hours a day should surely count for as much as doing problem sets 4 hours a day.
gregbo
Dec. 4th, 2008 04:32 am (UTC)
But is it that prerequisites exist, or is it some particular aspect of them? In the English classes, you'd be in just as much trouble as in an engineering class if there were prerequisites that required a lot of time to learn and/or review.

I thought of another comparison to music study that might shed some more light. For a recital or audition, you're required to prepare some pieces to perform. You know in advance what the pieces are, so you know what to practice. Except for very obscure pieces, you also have some "reference" works to listen to to get an idea of how they are performed. (In the case of obscure pieces, the composer, date of composition, usage of compositional tools, etc. provide some insight into what the piece should sound like.) So the only real unknown is how the judges will feel about your performance, as opposed to what topics will be covered on an engineering exam, what techniques are required, etc.

I remember on the CS112 (intro to modeling and analysis aka queuing theory) final during my first quarter of grad school, there was a question requiring integration by parts. If there were any sample problems, problem sets, etc. that required integration by parts, there weren't many relative to the other techniques that were required. So someone might very well not remember how to do integration by parts on the final, although if they had an idea it might be required, they might have reviewed it. But if they reviewed integration by parts, what other things weren't they reviewing that they might have needed to? There were too many possibilities -- all the things from integral and differential calculus, some linear algebra, lots of combinatorics and probability.

This might be an argument for making engineering exams comprehensive in the way that recitals or auditions are. For example, you could be assigned an electronic circuit to design and build. On the day of the exam, you have to build it for the judges who make sure it is working correctly.
fauxklore
Dec. 4th, 2008 10:07 am (UTC)
Again, I think it's the progressive nature of the prerequisites. That tends not to be true for things iike English classes except those dealing with, say, formal linguistic aspects.

As for music study, there are certainly auditions in which one does not know beforehand what one will have to play. A guy from my home town applied to Julliard and was dazed when they put a piece of sheet music in front of him and asked him to play it. Our piano teacher had always played a piece through when it was first assigned and Emil's ear was good enough that he could know what he had to do from that, so he hadn't actually learned to read music!

Again, I would argue that integration by parts is part of a repertorie of techniques. If said music student were weak on, say, arpeggios, he could make the same omplaint if one of the pieces he was asked to play had a lot of arpeggios.

If you had really thoroughly learned all of the possibilities for the exam so they were second nature, then it wouldn't have been an issue if any particular technique (i.e. integration by parts) were required.

By the way, my experience at MIT was that almost all of my engineering exams were either open book or allowed a page of notes, so one could easily have reminders of quite a lot of techniques. I've always thought the page of notes technique is an excellent one for engineering exams mostly because making the page of notes is an effective way to study. And many of my exams in grad school were take home exams. Things like designing something and proving to the instructor that it worked were also common as a major portion of grades.
gregbo
Dec. 4th, 2008 09:49 pm (UTC)
As for music study, there are certainly auditions in which one does not know beforehand what one will have to play.

True. But these auditions are more like their engineering exam equivalents. The student has to figure out what to do on the spot. But in music (at least in piano), this creates an interesting phenomenon. You have expert sightreaders who struggle to memorize all but the most elementary pieces; you have those who play by ear excellently but struggle to read all but the most elementary works. In other words, you have specialization based upon the impossibility of being able to do everything that can possibly be done on the instrument. And there are even pieces that can't be sightread because it takes more than a few minutes to work out fingerings for them. If a judge decided to require such pieces, they'd get the same result as the engineering professors who give exams where the mean is something like 25% – there's no practical way for every student to do well, although some might do well because they happened to remember some things that were on the test, but would do poorly on another exam with different questions of the same difficulty.

Again, I would argue that integration by parts is part of a repertorie of techniques. If said music student were weak on, say, arpeggios, he could make the same omplaint if one of the pieces he was asked to play had a lot of arpeggios.

Sure, if he was asked to sightread a piece, and (for some reason) did not practice arpeggios enough. But in terms of how fundamental a building block some technique is for a discipline, can we argue that integration by parts is just as crucial as arpeggios? I don't think so, because integration by parts is a somewhat specialized technique that is not always applicable.

If you had really thoroughly learned all of the possibilities for the exam so they were second nature, then it wouldn't have been an issue if any particular technique (i.e. integration by parts) were required.

Well, now we're getting somewhere. What level of effort is necessary to learn all of the prerequisites for a given engineering class so that they're second nature? And what isn't being learned that the student might need to know (perhaps outside of engineering)? On the final in question, I actually remembered how to do integration by parts, so I could answer that question. But there was another question that was worth a lot more points that I didn't answer correctly. I came up with a quotient that required differentiating, which turned into a horrible mess. I should have used the chain rule, which would have simplified the math. So not only did I get the wrong answer, I lost valuable time that I needed for the rest of the final. If I had thought I would need to spend more time practicing those parts of calculus that were needed on the final, I would have done so, but then I would not have had time to practice some other things that I needed to because I was having trouble with them.

Granted, it had been a few years since I'd used calculus regularly when I took the class, so I was at somewhat of a disadvantage relative to when I was an undergrad. But I had been working hard on example problems, etc. – it was only the tests themselves that were giving me problems, not so much understanding concepts. But because of the weighting of the tests, I came off looking like I was capable of doing less than I actually was.

Now I would speculate that some of these Korean students I'm hearing about are being prepared from a very young age to do just that – do engineering like it was second nature, at the expense of other things they might arguably spend time on.

(to be continued ...)
gregbo
Dec. 6th, 2008 06:17 am (UTC)
(... continued from last post)

There are students in other countries that are trained from a very early age to be excellent engineers, scientists, etc. They get (at least) the kind of tech-oriented education that US students used to get back in the 1960s, when the UC system was the envy of educational systems not just in the US, but around the world.

But their education doesn't (can't) cover everything. Some of them have poorer English skills than people who come up through the US K-12 educational system, even if they aren't academic tech superstars. But it doesn't seem to matter much, because they're admitted to US universities, hired by US employers in the US, etc. despite these limitations. Some of them don't even seem to be concerned that their education might be deficient in some ways, as long as they can outdo US students. For example, in a thread some time back on Aurelie Thiele's blog about the MS in Analytical Finance, an individual remarked that "universities loose good international students because of the restrictive TOEFL score required." Well, perhaps their TOEFL scores would improve if they spent more time on English. But at what cost, especially to their math/science ability, comes that extra time on English?

Another area where this is relevant is how TAs are selected. (Interestingly, in the CC thread that prompted my post, the subject of the quality of TAing has come up.) I once had a TA for 6.035, the compilers class, who might have been a world-class computer language researcher, but his English skills were really poor. There were widespread complaints about his inability to communicate clearly in English. I personally had a lot of trouble understanding him, which added to my overall frustration with the class. I would argue that it was more important to us, the students in the class, to have someone who was competent enough to TA the class and also had enough English skills to communicate needed information clearly.
fauxklore
Dec. 6th, 2008 07:18 am (UTC)
First, I would argue that both integration by parts and the chain rule are used frequently enough that they are part of the legitimate engineering mathematics repertoire. Any mathematical trick that can be used for more than two problems gets called a "method." What I would expect of a student is, say, to realize that the answer to a problem involving cylindrical fluid flow will be a Bessel function.

Second, my experience is that the education in some of those other countries is not really great preparation for real engineering practice, which requires creative problem solving. Which is something that, frankly, I'm not sure can be taught.

In fact, communications skills are vital to engineering success if one is to be more than a drone sitting in a back room for their entire career. I've probably said before this before, but I still consider the secret of my success to be my willingness to pick up the phone and call people I don't know. I've gotten where I am less because I can understand the technical issues but because I can do the "geek to English" translation to get other people to understand why they matter.

Finally, I agree that the inadequate English of TAs (and, sadly, some professors) is a serious problem for many students. But second language acquisition actually falls into that category of "progressive" learning that I've already been talking about. If you don't have a certain basic vocabulary, all of the grammatical rules in the world do you little good. My guess is that the extra time they'd spend on English would have more to do with grammar and less to do with the pronunciation that usually is the barrier to understanding.

For what it's worth, TOEFL is probably not a great test of English proficiency. Personal interviews would be the best approach, but would probably be too complicated to arrange.
gregbo
Dec. 7th, 2008 09:27 pm (UTC)
First, I would argue that both integration by parts and the chain rule are used frequently enough that they are part of the legitimate engineering mathematics repertoire.

I'm not trying to argue that they're not part of the curriculum, but what is the right emphasis to place on each, especially when time is limited (such as when you are preparing for a final, or taking one)? Admittedly, integration by parts is less universally applicable than the chain rule.

I actually had to look at the queuing theory textbook to refresh my memory about the issue at hand (inverting transfer functions), since I don't do this regularly. The technique that is most emphasized in the textbook (through development of the theory and examples) involves taking complex residues. But you don't always know right off whether you will need the chain rule, the quotient rule, or any calculus at all. It depends on how the equations come out. Sometimes you can just invert the resulting terms by inspection, because they're in a canonical form. But the professor's book emphasizes the complex residue method, and most of the problems I recall involved differentiating a quotient, so it wasn't unreasonable for me to approach the problem that way. It was just unfortunate that I ran into trouble somewhere and came up with an overcomplicated expression, and that I didn't realize early on that I could have used the chain rule, which would have simplified the solution.

As I recall, most of the people in the class were surprised to see a question involving integration by parts. The going sentiment seemed to be that the professor wanted to see if anyone could solve such a problem, which was not totally inconsistent with his teaching and testing methods, but not likely.

Anyway, keep in mind that the backdrop for all of this is why engineering is hard, and some people become discouraged and pursue other careers. US companies say that they don't have enough engineers so they have to go outside the US to get them, but they don't take into account why this is so. US education, typically speaking, does not prepare all students for a demanding engineering career from a very young age. We teach people other things, like history, Spanish, English grammar, and so forth, in the hopes that they will be well-rounded – able to pursue a number of career paths. Perhaps it is just not possible to create the most elite engineering force on the planet without making engineering the primary focus of a student's life.

In fact, communications skills are vital to engineering success if one is to be more than a drone sitting in a back room for their entire career. I've probably said before this before, but I still consider the secret of my success to be my willingness to pick up the phone and call people I don't know.

Agreed. It's unfortunate that there are (software) engineering companies who consider it more valuable to code madly away in one's cubicle than to have a discussion about how a project might be better organized in order to improve its chances of success. I was glad to hear that some companies like Google are now being required by the change in economic conditions to financially justify projects, rather than just throw ideas against the wall and see if they stick (when some discussion would reveal that they wouldn't).
jessiehl
Dec. 1st, 2008 02:43 pm (UTC)
I don't often agree with sakky over in College Confidential's College Discussion forums...

Bwahahaha I've largely stopped reading his posts, because even when I agree with them they are too long-winded and pompous (though I might be the pot calling the kettle black, or something). I had forgotten that you ever read CC.

Recall a time I've mentioned when a physics professor flunked 60% of the MIT freshman physics students in the fall of 1978.

Huh, my dad has told me about this as well (he was an MIT undergrad at the time). Until you gave the date here, I don't think I realized that you must be talking about the same incident.

Do you think that what you are talking about really makes engineering different from other things? I mean, I'm not sure why it would be inherently easier to prepare perfectly for, say, a political science test (and I definitely don't think it's easier to prepare for a science or math test, but when you say engineering I am generally reading it as math/science/engineering anyway).

The interview process is a different matter. A lot of tech companies seem to focus on whether you know trivia about languages, stuff like that, which does not in my opinion have much to do with whether you will be good at the position. And I do get the impression that this is less of an issue in other fields.
gregbo
Dec. 2nd, 2008 04:27 am (UTC)
After a few years away from it, I have been reading and posting to CC off and on for the last three months or so. I thought you might've recognized my posts from the subject matter or writing style.

Do you think that what you are talking about really makes engineering different from other things? I mean, I'm not sure why it would be inherently easier to prepare perfectly for, say, a political science test (and I definitely don't think it's easier to prepare for a science or math test, but when you say engineering I am generally reading it as math/science/engineering anyway).

I don't know for sure. It may have to do with how grading is done in other majors, or how employers, grad schools, etc. interpret grades as part of the whole package. But as for my own experience in CS classes, I often felt that I couldn't prepare perfectly. Perhaps if I give an example it will make more sense.

In my first quarter of grad school I took a class called CS181, which is like the 6.045 automata/formal language class I've written about earlier. The class is usually taught by Sheila Greibach (of Greibach normal form), but this time it was taught by Stott Parker. I had my old notes and tests for 6.045 to use, some sample CS181 questions from past years (including things asked on Major Field Exams, which are part of what PhD candidates have to pass at UCLA to earn their doctorates), plus some things I'd picked up off the net and from friends. So I had a lot of material to help me study.

I was doing really well on the problem sets – getting nearly full or full credit on all of them. I also was doing reasonably well with the sample questions I had; granted, I couldn't answer every question and sometimes had to ask for help, but I was getting enough of the material correct to give me the impression that I would get an above average grade on the midterm and final. I even answered questions in lecture (usually a rarity, given my shyness), so I had a fair amount of self-confidence ... until the midterm and final came around.

I won't go into a lot of detail about my actual grades, but they were a little below average on both the midterm and final. I was disappointed not just because I'd worked really hard, but because I had no reason to believe I wouldn't do better. There was nothing (except the fact that I couldn't answer all of the sample questions) that would have lead me to believe that I wouldn't do better, but there isn't anyone who can answer every sample theory question (except perhaps people who've just gone through CS theory quals, or wrote a CS theory book). If I had been able to prepare perfectly, I would have had some indication that there were some types of problems that I needed to work on, and would have worked on them (instead of things that never came up).

Now, perhaps in some cosmic sense, I am better off for having worked more problems. But considering that the midterm and final were worth 40% of the final grade each, a lot of my effort was wasted.
jessiehl
Dec. 2nd, 2008 01:54 pm (UTC)
I won't go into a lot of detail about my actual grades, but they were a little below average on both the midterm and final. I was disappointed not just because I'd worked really hard, but because I had no reason to believe I wouldn't do better.

Sounds like the story of my MIT career. It doesn't seem to affect everyone, though. I wish I had figured out what made it not affect some people.
gregbo
Dec. 2nd, 2008 04:29 am (UTC)
(continued from last post)

The place where I lost most needed credit on the midterm was on some true/false questions. They were graded the same way the old SAT multiple choice were graded – you lost credit for incorrect answers. But in this case two points were deducted for each wrong answer, so instead of getting eight points for that section, I only got two, which was enough to put me under the mean. (Someday I intend to post something about why I think that the excuse given for this type of grading – that it discourages guessing, is a poor excuse. If it is so important for people not to guess, give an exam where one has to show work.) Two of the true/false I gave wrong answers to involved determining whether a pair of given regular expressions accepted the same language. There is an algorithm that can determine this that I knew of, but while I was taking the midterm I decided that it would take too long to iterate through all of the steps of it, so I needed to (try to) use some insight to answer those questions. (Arguably, the algorithm was not covered in CS181, but it was given in one of the textbooks that CS280a (analysis of algorithms) uses. So one might reasonably assume that SP didn't expect people to use that algorithm to answer the question. It just so happens that I was taking CS280a at the same time, and SP was teaching it also, otherwise I may not have been aware of that particular algorithm.)

Another side of this is not knowing how much preparation time is necessary to get a certain grade. For example, I probably spent as much study time on linear algebra (when I took it at the 'tute) as I did on signals and systems (the 6.003 of my day). But I got one letter grade higher in linear algebra than signals and systems. Arguably, I needed to spend more time on signals and systems, but I had no way of knowing how much less time I could've spent on linear algebra and still earned the same grade. That may be part of what sakky's trying to get at – that you can't really be sure how much preparation you need to do to earn a certain grade.


Edited at 2008-12-02 04:04 pm (UTC)
grark
Dec. 3rd, 2008 03:43 am (UTC)
What are CS interviews like? (I'm a current undergraduate student)
gregbo
Dec. 3rd, 2008 05:22 pm (UTC)
CS interviews tend to vary quite a bit, depending on the interviewer, the company, etc. Some interviews require a lot of detail – you'll be asked specific questions about what a command or system call does. Some require you to write code on the spot, and the interviewer may be looking for specific coding practices, such as if you write or type in a closing brace beneath an opening brace. Some interviewers may insist that you write the code on paper or a whiteboard instead of on a computer. There are also some who'll ask tricky puzzle questions.

If you follow the link of my "job search" tag, you'll see some excerpts of interviews I had up to about a year ago, when I landed my current job. There are also links to sites with more interview questions, the philosophy of interviews, etc.
( 16 comments — Leave a comment )