History of the CS enrollment roller coaster

I just read a report by Eric Roberts, A History of Capacity Challenges in Computer Science, which I was pointed to by a guest post on Mark Guzdial’s blog.  The report discusses the two previous rapid increases in CS enrollment and BS degree production (peaking in 1986 and 2004), comparing them to the current rapid growth.  He makes the argument that the current growth is more like the 1986 peak (triggered by the introduction of the PC) and less like the 2004 peak from the dot-com bubble.

Number of CS bachelor’s degrees per year, which trails total enrollment by a couple of years. [Copied from http://cs.stanford.edu/people/eroberts/CSCapacity/images/BSDegrees-1975-2014.png%5D

The crash in enrollments after then 1986 peak was a “capacity crash”—that is, it was not triggered by a loss of interest by the students nor by a lack of interest from industry, but by deliberate university policies to make CS unwelcoming to limit the demand.

The rapid growth of enrollment in CS poses a problem that academia is ill-equipped to handle, for two reasons:

1. The rate of growth is much faster than the rate at which universities respond.  Faculty growth in any department is generally limited by bureaucratic processes to a maximum of about 10% a year and generally only allocated after a 2–5-year delay, but enrollment growth has been 15–20% a year for several years in a row.
2. The PhDs to fill CS faculty positions are not available.  This is an unfamiliar problem for academic administrators, because most of the rest of academia has a huge buffer of under-employed PhDs (the “postdoc holding tank” in life sciences) that can be tapped to fill any new positions.  But in CS, and in some other engineering fields, the existence of attractive industrial jobs with more resources, better working conditions, and higher pay than academia means that there aren’t many people waiting for an academic position to open—in the earlier spurts of enrollment growth, there were as many as 7 faculty job openings per qualified candidate, and the current market seems to have 4 faculty job openings per qualified candidate.

Eric Roberts makes the case that we can’t know for sure whether the current rapid enrollment growth is like the dot-com bubble, but if we don’t do something to address capacity very quickly, we will trigger a capacity crash like the 1986 one.

I think that one point he missed, which is affecting UCSC strongly, is that the enrollment growth in this round is much broader than in previous ones—many fields of engineering are seeing rapid growth in enrollment, and all the departments in the Baskin School of Engineering at UCSC are seeing capacity problems.  The problem is most acute for Computer Science, but it is not a single-department problem as it was in the two previous peaks. (Note: Eric Roberts presents a slide which shows that engineering as a whole is not in crisis, while CS is, but I think that a lot depends on which engineering fields you look at—there is a glut of petroleum engineers, but shortages of electrical, robotics, and computer engineers.  UCSC is unusual in having mainly the fields in which there is high demand, due to deliberate planning to grow only those programs, rather than having the full complement of traditional engineering programs.)

He shows the Bureau of Labor Statistics estimates of jobs in various fields for the next few years:

Note that computing and engineering jobs make up huge fraction of the job market, despite the relatively small proportion of engineering and CS faculty in most universities.
Figure copied from http://cs.stanford.edu/people/eroberts/CSCapacity/images/BLSJobGrowth.png

Unfortunately, Eric Roberts’s paper does not offer solutions, merely historical perspective, but even that is valuable, as there are relatively few CS faculty who remember the enrollment problems of the early 1980s—too many faculty left for industry or took early retirement.  He points out another interesting challenge for newer CS faculty learning about the enrollment problems:

Much of the early history lies beyond the Google “event horizon.” In putting together this history, I was interested to discover that several relevant articles I remembered from the early 1980s were invisible online because they predate digital archiving for the journals in which they appear. Looking for evidence about faculty shortages in the 1980s becomes much harder when none of the references from, for example, The Chronicle of Higher Education, show up in Google searches.

I highly recommend reading the full report by Eric Roberts, as I’ve only touched on a couple of the highlights here.

Advice for CS students

Katrin Becker has collected advice from senior computer science students to those just starting out, which she has collected in her blog post Worth Sharing: Advice for New Students From Those Who Know (Old Students).   She summarizes the list as follows:

Top 10 list [Highlights]

• Start Assignments Early: they will take longer than you think.
• Make friends.
• Ask questions.
• Practice Programming. Try Stuff.
• Make sure Computer Science is really what you want to do.
• Choose options you will enjoy.
• Go to class.
• Don’t take too many CPSC courses in one semester.
• Plan your programs on paper before you start typing.
• Remember to have fun.

and goes on to give specific quotes from students for each of these categories (plus some other unsorted quotes).  Some of my favorites include

Be nice to people. You never know when they will be able to help (or hinder) you in the future.

You’ll enjoy it more if you treat your work as a full-time hobby, and not just a path to a job.

Working in groups tends to be harder, not easier.

Of course, a lot of the advice applies to any engineering major, and some to any college major.  Even the advice about programming applies to other design tasks.

Where PhDs get their Bachelors’ degrees

Last year I wrote about a study that looked at where CS PhD students got their bachelors’ degrees. Now Reed College has extended that question to other fields as well: Doctoral Degree Productivity.  Their point was to show how high Reed ranked on the standard they chose: the number of students who went on to get PhDs divided by the number of students getting bachelor’s degrees.  I quote the tables and accompanying text below, but I take no credit or blame for the data—this is directly from Reed’s site:

Undergraduate Origins of Doctoral Degrees

Percentage ranking of doctorates, by academic field, conferred upon graduates of listed institutions.

Rank	All Disciplines	Science and Math	Social Sciences	Humanities and Arts
1	Calif. Inst. of Tech.	Calif. Inst. of Tech.	Swarthmore	New England Conserv. of Music
2	Harvey Mudd	Harvey Mudd	Grinnell	Curtis Institute of Music
3	Swarthmore	Reed	Reed	Juilliard
4	Reed	MIT	Bryn Mawr	Cleveland Inst. of Music
5	Carleton	NM Institute Mining & Tech.	Spelman	St. John’s College
6	MIT	Carleton	Oberlin	Reed
7	Grinnell	Wabash	Wesleyan	Hellenic College-Holy Cross Greek Orthodox Sch. of Theology
8	Princeton	Rice	St. Joseph Seminary	Swarthmore
9	Harvard	Univ. of Chicago	Harvard	Oberlin
10	Oberlin	Grinnell	Pomona	Amherst

Percentage Ranking by Specific Fields of Study

Rank	Life Sciences	Physical Sciences	Psychology	Other Social Sciences*	Humanities
1	Calif. Inst. of Tech.	Calif. Inst. of Tech.	Univ. Puerto Rico – Aguadilla	Swarthmore	St. John’s, MD
2	Reed	Harvey Mudd	Wellesley	Reed	Reed
3	Swarthmore	Reed	Vassar	Harvard	Amherst
4	Carleton	MIT	Hendrix	Grinnell	Swarthmore
5	Grinnell	NM Institute Mining/Tech.	Pontifical Coll. Josephinum	Univ. of Chicago	Carleton
6	Harvey Mudd	Carleton	Grinnell	Bryn Mawr	Yale
7	Univ. of Chicago	Wabash	Swarthmore	Thomas More College of Lib. Arts	Thomas More College of Lib. Arts
8	Haverford	Rice	Barnard	Oberlin	Bryn Mawr
9	MIT	Univ. of Chicago	St. Joseph Seminary Coll.	Bard College at Simon’s Rock	St. John’s, NM
10	Earlham	Grinnell	Pomona	Wesleyan	Wesleyan
11	Harvard	Haverford	Reed	Amherst	Princeton
12	Cornell Univ.	Swarthmore	Wesleyan	Pomona	Bard College at Simon’s Rock

*Does not include psychology, education, or communications and librarianship.

Source: National Science Foundation and Integrated Postsecondary Education Data System. The listing shows the top institutions in the nation ranked by estimated percentage of graduates who went on to earn a doctoral degree in selected disciplines between 2001-2010.

All the schools listed are private schools except Univ. Puerto Rico—Aguadilla and NM Institute Mining/Tech., but seeing dominance by expensive private schools is not very surprising—grad school is expensive, and students who can afford expensive private schools are more likely to be able to afford expensive grad school and are less likely to need to work immediately after getting their B.S. or B.A. A PhD is not a working-class degree—it is prepares one for only a small number of jobs, mainly in academia or national labs, so for many it is just an elite status symbol.  What is more surprising is how poorly the Ivy League schools do on this list—perhaps those who get their elite status conferred by their bachelor’s institution see no need to continue on to get higher degrees.

Reed does not report numbers directly comparable with the ones in the Computing Research Association report, which reports only on computer science PhDs, where

Only one institution (MIT) had an annual average production of 15 or more undergraduates.   Three other institutions (Berkeley, CMU, and Cornell) had an average production of more than 10 but less than 15.  Together, these four baccalaureate institutions accounted for over 10% of all Ph.D.’s awarded to domestic students.   The next 10% of all Ph.D.’s in that period came from only eight other baccalaureate institutions (Harvard, Brigham Young, Stanford, UT Austin, UIUC, Princeton, University of Michigan, and UCLA). 

Note that five of the top producers of bachelor’s in CS who went on to get PhDs were public schools.  The CRA does not report PhD/BS numbers for individual institutions, probably because the numbers are too small to be meaningful for most colleges—you have to aggregate either across many colleges or across many fields before the denominators are big enough to avoid just reporting noise.  Reed did the aggregating across fields, while the CRA report aggregated across colleges, finding that research universities sent about 2.5% of their CS graduates on to get PhDs, 4-year colleges about 0.9% and masters-granting institutions about 0.6%.  They did have one finding that supports Reed’s analysis:

The top 25 liberal arts colleges (using the U.S. News and World Reports ranking) collectively enroll slightly less than 50,000 students per year in all majors and were the origins of 190 Ph.D. degrees between 2000 and 2010, collectively ranking ahead of any single research university.

Reed’s findings are also consistent with the NSF report that put the “Oberlin 50” colleges highest at over 5% of their science and engineering graduates going on to get PhDs, compared to about 3% for research universities.  The NSF report supports somewhat the analysis that socio-economic status is important in determining who goes on to grad school—private research universities match the Oberlin 50, but public research universities have only about half as large a fraction of their graduates go on to grad school.

I found out about this site from The Colleges Where PhD’s Get Their Start, which has a copy of the tables that probably came from an earlier, buggy  version of the site, because Lynn O’Shaughnessy wrote

I bet most families assume that attending a public flagship university or a nationally known private research university is the best ticket to graduate school. If you look at the following lists of the most successful PhD feeder schools for different majors, you will see a somewhat different story. Not a single public university makes any of the lists. The entire Cal State system, however, is considered the No. 1 producer of humanities PhD’s.

I could believe that the Cal State system had the largest raw numbers of students going on to get PhDs in humanities, as they are a huge 4-year college, enrolling about 438,000 students [http://www.calstate.edu/as/cyr/cyr13-14/table01.shtml], with about 76,000 bachelor’s degrees per year [http://www.calstate.edu/PA/2013Facts/degrees.shtml]. Are there any other colleges in the US graduating so many BS or BA students per year? But the fact remains that Cal State is not the flagship university of California, and the University of California probably has a much higher percentage of its alumni go on to get PhDs.

In fact, one of the big problems with these lists is the question of scale—most of the colleges that come up high on Reed’s lists (which means high on NSF’s lists) do so by having very small denominators—they don’t graduate many students, though a high percentage of those go on to get PhDs.  In terms of raw numbers of students who go on to get PhDs, the public research universities produce many more than the private research universities, and the liberal arts schools are just a drop in the bucket. Of the top 25 schools in terms of raw numbers who go on to get PhDs in science and engineering, 19 are public research universities and 6 are private research universities—of the top 50 only 17 are private research universities.

When you are looking for a cohort of similarly minded students, you get slightly higher enrichment at some very selective private schools, but there are actually more peers at a large public research university—if you can find them.

CS is not a foreign language

The Computer Science Teachers Association recently had a blog post about a recent development in high school CS teaching, Why Counting CS as a Foreign Language Credit is a Bad Idea.

When these policy makers look at schools, they see that computer science is not part of the “common core” of prescribed learning for students. And then they hear that Texas has just passed legislation to enable students to count a computer science course as a foreign language credit and it seems like a great idea.

But all we have to do is to look at Texas to see how this idea could, at the implementation level, turn out to be an unfortunate choice for computer science education. Here are the unintended consequences

1. If a course counts as a foreign language course, it will be suggested that a new course must be created.
2. If a new course is created, chances are that it won’t fit well into any of the already existing course pathways for college-prep or CTE.
3. This new course will be added to the current confusing array of “computing” courses which students and their parents already find difficult to navigate.
4. There will be pressure brought to ensure that that course focuses somehow on a “language”. For the last ten years we have been trying to help people understand that computer science is more than programming. Programming/coding is to computer science as the multiplication table is to mathematics, a critical tool but certainly not the entire discipline.
5. If this new course is going to be a “language” course, we have to pick a language (just one). And so the programming language wars begin.

I agree that computer science should count as a math or science credit, not a foreign language credit.  CS and foreign language use different mental skills. For one thing, the “grammar” taught in programming languages is deliberately very much simpler than natural language grammars—so much so that there is little transference between learning on and learning the other.  Also, foreign language courses include a lot of memory work (particularly vocabulary, but also conjugation and declension patterns in many languages), while beginning computer science courses are more about learning how to design and debug—that is, to develop problem-solving skills, not memory skills.

Foreign language instruction is also very important, and replacing it with computer science would not be helpful for the cultural understanding and ease of relations between different countries that is an important reason for teaching foreign languages. Decreasing foreign language to make room for computer science is short-sighted.

The most sensible classification for CS (if it needs to be classified in the narrow categories that guide secondary school administrators) is with math.  The reasons for learning CS are much the same as the reasons for learning algebra, both in terms of the underlying set of mental skills that one hopes would transfer to other fields (but usually don’t) and the usefulness as a base for future study in STEM fields.

I think that the claim “Programming/coding is to computer science as the multiplication table is to mathematics” is a bad analogy.  Closer would be “Programming/coding is to computer science as algebra is to mathematics”.  Programming is a much larger and more complicated skill than multiplication, and it underlies most of the rest of computer science. Multiplication is primarily a memory skill (multiplication tables and a simple algorithm), while programming is primarily a problem-solving skill.

I also agree with the concern that making CS a “foreign language” skill will put even more pressure on high schools to adopt a common programming language.  The College Board AP test has already blessed Java, which I think is a poor choice for a first teaching language (though I know that many computer science professors like it as a first language, for many of the same reasons I dislike it).  I prefer Scratch followed by Python, as I’ve explained before on this blog.

Standardizing the way computer science is taught would be a moderately bad thing, as there is no “one true path” that produces great programmers and computer scientists. I see value in having a variety of different pedagogical approaches to teaching programming (and computer science), as the first few programming languages one learns tends to color the way one things about programs for several years.  I believe that a diversity of different approaches to programming is important to the health of the computer science both as an academic pursuit and as an industry—and different initial programming experiences is as important to that diversity as different people are.

---

Note: I tried to post a comment on the original post, but I kept getting

Server error!

The server encountered an internal error and was unable to complete your request.

Error message:
Premature end of script headers: mt-comments.cgi

If you think this is a server error, please contact the webmaster.

Error 500

You’d think that the Association for Computing Machinery could keep a blog running, but I guess their disdain for mere “programming” extends to the programmers who set up their website.

Technical entitlement—is it a thing?

I learned a new buzzword yesterday: “technical entitlement”.  I encountered the phrase on  the blog On Technical Entitlement | Soshitech.com, though apparently Tess Rinearson originally wrote it in June 2012 and also published it on medium.com.

I’m the granddaughter of a software engineer and the daughter of a entrepreneur. I could use a computer just about as soon as I could sit up. When I was 11, I made my first website and within a year I was selling code. I took six semesters of computer science in high school, and I had two internships behind me when I started my freshman year of college.

Despite what it may seem, I’m not trying to brag—seriously. I’m just trying to prove a point: I should not be intimidated by technical entitlement.

And yet I am. I am very intimidated by the technically entitled.

You know the type. The one who was soldering when she was 6. The one who raises his hand to answer every question—and occasionally tries to correct the professor. The one who scoffs at anyone who had a score below the median on that data structures exam (“idiots!”). The one who introduces himself by sharing his StackOverflow score.

That’s technical entitlement.

“Technical entitlement” seems to be the flip side of “imposter syndrome”. In imposter syndrome, competent people question their own competence—sometimes giving up when things get a little difficult, even though an outside observer sees no reason for quitting.  “Technical entitlement” seems to be blaming those who have both competence and confidence—as if it were somehow deeply unfair that some people learned things before others did.

Certainly some things are unfair—as an engineering professor I’ve been able to provide opportunities for my son to  learn computer science and computer engineering that would not be available to a parent who knew nothing about those fields.  And some of the characteristics she lists would apply to my son—I can see him correcting his professors, and although he’d never introduce himself by sharing his StackOverflow score, he did include it in some of his college essays, as evidence that he was knowledgeable and interested in sharing what he had learned.

But Tess Renearson goes on to say

It starts with a strong background in tech, often at a very young age. With some extreme confidence and perhaps a bit of obliviousness, this blooms into technical entitlement, an attitude characterized by showmanship and competitiveness.

While my son has confidence in his abilities and “perhaps a bit of obliviousness”, neither showmanship nor competitiveness are big factors in his behavior.  I think that Ms. Renearson has confused a personality trait and stereotypical US male behavior (showmanship) with early technical education. I see the arrogance as a bad thing, but the early technical education (which she herself had) as a good thing.

The rest of her post goes on to talk about ways that Amy Quispe and Jessica Lawrence managed to increase participation (particularly by women) in tech events.  But the analysis there really addresses imposter syndrome more than it does “technical entitlement”.  She quotes Jessica Lawrence: ‘“There is,” she said, “an under-confidence problem.” But Ms. Renearson then says

Sound familiar? Yep, it’s exactly the kind of self-doubt that can arise when there are so many technically entitled people around.

Somehow blaming “technically entitled people” for the under-confidence of others seems to be imposing blame where none is warranted.

Now imagine someone starting out as a college student taking their first CS course. Imagine how the technical elite make them feel.

I can understand someone being intimidated when entering a new field if they are surrounded by people more skilled in the field—but that is hardly the fault of the those who are skilled.  Newcomers anywhere are going to feel out of place, even when people are trying to welcome them. The “technical elite” are not making the newcomers feel intimidated.

If Ms. Renearson’s point is that some of the tech communities are not sufficiently welcoming of newcomers, I agree.  I’ve seen snarky comments in places like Stack Overflow that offered gratuitous insults rather than assistance.

But Ms. Renearson seems to assume that anyone who is more experienced than her is automatically trying to put her down, and that this is the way that everyone should be expected to feel.  When one starts with that assumption, there is no remedy—no matter what those more experienced or more skilled do, they will be seen as threatening.

Perhaps she has not identified those who should be getting blamed precisely enough.  I don’t think that it is “The one who was soldering when she was 6” who is a problem, but those who refuse to give children an opportunity to learn (no public school in my county teaches computer science, except one lottery-entry charter) or who force students who’ve been programming for 6 years into the same classes as those who have never programmed, as many college CS programs do, providing no way for more advanced students to skip prerequisites.

Unfortunately, identifying the problem as being “technical entitlement” makes the problem worse not better, as it encourages public schools to suppress the teaching of technical subjects, rather than expanding them.

If she means to attack the arrogant culture of “brogrammers”, mean-spirited pranks, and other unpleasant culture that has emerged, then I support her, as I’m not happy with some of the culture I see either.  But don’t blame it on the kids who learned tech early, nor on the parents who taught them—the late-comers are more likely to be the arrogant bastards, since that arrogance is mainly a defense mechanism for incompetents.  The competent tech people are much more likely to be eager to share their enthusiasm with newcomers and help them join in the fun.