This is a beautiful visualization, but it is very hard to parse, and I have no idea as to what problem it was designed to solve.
With regards to this visualization in particular, it is hard to judge which majors are disproportionately represented in any individual field.
There are in some places gaps in the spacing which make it appear as if a particular area is underrepresented; for instance, observe the apparent gap in Div I majors between Engineering and Health/Medicine.
It is very hard to judge the relative distribution of fields for small majors, such as Geoscience.
Path density seems exaggerated when the paths are drawn between nearby cells; compare the apparent density of paths between Div I <-> Arts/Writing/Social Science/Govt/etc with Div I <-> Education/Health. This seems to be because of the thinning of lines in the middle of the figure.
In reading about this type of figure, I managed to find possibly the most incomprehensible infographic I've ever seen:
I have a feeling that for the coming year or so, these visualizations will be overused in much the same way animated web site intros were in the late 90's. Because right now they are still new enough that the casual viewer will inevitably be impressed by their pretty colors, even if they do not particularly clarify any information. I suspect a slight backlash by late 2013, then a convergence towards legitimate use thereafter.
They're already overused in genome biology. They do a beautiful job of representing interchromosomal rearrangements in cancer, but don't actually convey a whole lot of information.
That isn't much better though. With very few exceptions, it looks like regardless of major people tend to enter various fields in very similar proportions. If the data were displayed in a table, I have a feeling the numbers would tell a much different story.
Looks impressive, but a simple N x K contingency table would help understand the data better. Here, N=K=15 and entries of the table would be raw counts (e.g. number of major i in career j).
You could color the cells by a few different criteria, e.g. absolute scaling, row normalization, and column normalization with red for "lots" and blue for "few". Maybe toggle those three criteria via a radio button.
The advantage of such a visualization is that it allows you to see both trends (which rows/cols are more red, and which blue) and specific numbers. It is also immediately interpretable for a new viewer without any explanatory preamble. Sometimes the simple stuff is best.
I would say the advantage of the original visualization is the clear lines drawn between n and k. The table is very complex without filtering, but imo very clear when you mouse over the majors on the right.
I would say the real problem is that the subsets are very uneven; the orange group dwarfs the other two, which arguably makes it more difficult to understand the trends. Along that line of thought, however, I would argue that the Psych majors belong with the green group.
As a Williams alumnus: the major color groups correspond to the "divisions" of courses at Williams -- the administration groups all departments into three divisions, and requires that students take a certain number of classes from each division.
So the chart faithfully reproduces the decision of the Williams College administration to include Psychology in Division II (Social Sciences) rather than Division III (math and hard sciences).
Nice data visualizations. One thing I'd also like to see here is a representation for multiple majors (35% of Williams students are double majors).
Tangentially related, I was a little disappointed to see the slice for Computer Science is still so small. It was also small in mid-90's, and I believe it grew dramatically during the first dotcom boom, but it seems to have dropped back to the earlier levels.
Not that I think the CS program should get any easier just to attract more students (it was notoriously difficult for a while), but rather I think that CS can be the perfect science complement in a well-rounded liberal arts education.
I believe my year (2003) was the biggest CS class ever. Somewhere around 25-30; I forget the exact number.. I know it dipped in the couple years after that quite a bit, but not sure if it's picked up again.
Class of 2015 here! We have record numbers signed up for Algorithms in the spring, something like 30-40 and a lot of those people are planning to major.
It's a real department with seven profs. Lots of non-majors take CS classes, but even so, class sizes are relatively small (usually in the 10 to 30 range). That's really one of the major selling points for the LAC model of education.
At Bard College, we have recently begun offering CS intro courses that are juxtaposed with other disciplines. For instance, one of the most popular courses this semester was a CS called "Interactive Media" that had students learning Processing with the end result being they had to create a substantial piece of art (installation / visualizer / game) at the end of it. I also believe that next semester there is an intro course that is focused on using NLP (really basic stuff like N-grams, counting the number of word repetitions on a text) in conjunction with experimental poetry.
Its great because it gets the fundamentals of programming down while also allowing the students to attach it to their other interests. I love being able to talk to my art student friends about their projects and helping them with code. They are all really into it.
I think I know who you were and met you at Williams. I'm also a CS major and I graduated in the late 90's. It is indeed the case that the CS major grew absurdly huge in the late 90's, with my class being double the size of the previous year's class, and the class below me doubling in the size of my class. And with intro courses being yet more absurd. All of my memories of our profs involve them with big circles under their eyes, clutching mugs of coffee.
It says it does represent double majors, but probably not in any useful capacity that you were looking for: "Those with double-majors have two arcs on the left (one from each of their majors, each arc of 1/2 thickness) that combine into one resulting career choice."
I like how each bar has paths whose widths add up to the length of the bar (100%), but I don't like the circular approach to the visualization. It seems to imply that each arc on the circle is the same "type" as everything else, while in reality each subject of the agglomeration will have one major and one career (roughly speaking).
Wouldn't it make more sense to divide it into two linear segments, whose sub-segments map to each other? That clarifies the distinction, and it'd make it clearer that, e.g., majors can't map to majors.
The data may be interesting to some people, but what intrigued me was the data visualization techniques. I thought the groupings were good, that the use of color and of line widths was helpful, and that the ways the lines curved actually HELPED to locate them (as compared with straight lines or a simpler curve). However, I think the left and right sides should have been represented by two different curves (perhaps the base of a parabola) that did NOT meet to make a full circle -- because until reading the details I didn't realize that the left and right sides of what appeared to be a single circle were essentially unrelated. All in all, a nice visualization.
The article credits it, but fwiw, this is done via CIRCOS, an open-source piece of software specialized for visualizing things as these circular graphs: http://circos.ca/
I agree that is was a very attractive visualization, but it took a couple minutes to see the significance of the left and right sides. It was very cool once it made sense.
At a school like Williams, Biology+Chemistry ~= pre-med. There's essentially no engineering. And most humanities degrees don't map to specific career paths.
interesting, but probably suboptimal in a quantitative sense. As an immigrant, it is interesting how Law is a universal dumping ground in the US...whether you major in Arts or Econ or PolSci or English or Philosophy or Culture Studies....you end up in Law !! Why ?
The clearest trends were Biology & Chemistry...almost always end up in Healthcare.
Is "college education" a proxy for gradschool ? If so, nice to see some strokes ending up in that bucket even in these very difficult economic times.
Williams College's primary student body composition is affluent upper middle class white persons. It is #1 on the US News & World Report and has a fairly high number of cross admits with Harvard/Yale. So everyone is high achieving/extremely driven. The primary differentiator is that it disproportionately attracts students who are very much sold on the "study a broad liberal arts education for the sake of education and forsake career oriented life skills". I.E. the attitude that gets mocked daily on reddit. I.E. the sort of idealism that led Paul Graham to study Philosophy in college (http://paulgraham.com/philosophy.html) is omnipresent on campus. Almost everyone starts an idealist.
This continues at Williams for a few years, until people panic and realize they don't have too many bankable life skills to get a real job. However, everyone is, or is surrounded by affluent, successful, career-oriented people. So they look around to find one of those successful career paths that are still open to them. Unless they are majoring in math/sciences (Williams has a 10% math major population, which is very high compared to the US average), or have already begun the pre-med track, the only clear avenues that aren't blocked off by their past decisions are consulting->MBA, and law school.
Disclosure: I attended Williams. About 25% of students double major. I started off with a clear "1 major for me, 1 for life skills" path, and intended to major in CS and Music. At the end of freshman year I realized I was an exceedingly mediocre composer, and switched my second to Economics (if that seems like a cynical career move, at the time I thought economics was the path to becoming Hari Seldon: http://www.guardian.co.uk/books/2012/dec/04/paul-krugman-asi...)
I went to a science/technology magnet high school, and 10 years out more of my classmates are in one of: law, consulting, or finance than any other. Even people who did a BS and sometimes MS often ended up getting graduate degrees and going into consulting. Why? The real money in this country isn't in making things, but in managing the people who make things, or in providing professional services to the managers. Silicon Valley is an exception right now, but only because consultants have yet to figure out how to drive down engineering salaries by outsourcing social network development to India (although Google, Microsoft, etc, have tried ardently to keep salaries down by engaging in anti-competive practices like agreeing not to poach each others' engineers).
Heck, even the Valley isn't immune to this phenomenon. Taking your CS degree to Goldman Sachs or JP Morgan then lateraling to a VC fund after a few years is probably, adjusted for risk, the highest return career path in the Valley.
Williams is a (highly ranked) private liberal arts school[1]. That alone will significantly bias the results of this data. I'm sure a good chunk of those people chose Williams just so they could have a prestigious school to get their undergrad degree from for their application into law school.
>"whether you major in Arts or Econ or PolSci or English or Philosophy or Culture Studies....you end up in Law !!"
Law school has been the "go to" graduate option for the "soft" sciences for a long time, since there is no requisite pre-law undergrad. It's also led to a surplus of Law graduates with no job prospects.
>"Is "college education" a proxy for gradschool ?"
I believe that is actually teaching college courses. But I could be wrong.
I agree with you about law - this is a kind of "doubling down" on the "no math career path".
However, top humanities and arts students with very good grades and LSAT scores can still dramatically improve their earning prospects by attending an elite law school. I suspect that quite a few Williams students fit that description.
MBAs are a harder option, because while they don't strictly require any one major, it can be harder to gain the experience MBA programs value with a very unmarketable major.
>"However, top humanities and arts students with very good grades and LSAT scores can still dramatically improve their earning prospects by attending an elite law school."
Whole-heartedly agree. I'm an economics guy myself, so I didn't intend to denigrate the degree choice. Top students will succeed no matter what; but the law grad outlook (or lack of) is no joke.
"Has been" is probably the applicable tense. And I saw this dynamic a lot among many English/Classical Studies/etc. majors that I've know. But, today, probably in no small part because of this dynamic (plus the ability to outsource/automate a lot of low IP legal tasks), the situation is actually pretty bad for middle-of-the-road law school gards from middle-of-the-road schools. The top Harvard grads still mostly do fine but law as a doubling-down strategy after graduating with a not particularly employable liberal arts degree is no longer especially viable.
Economists end up in law all of the time because economics and law are so tightly knit, particularly antitrust/competition policy.
In my undergraduate thesis I looked at what causes schools to graduate more or less economics majors, and the most important factor is the existence of a business school at the university. If a school does not have a business school, like at a lot of top ranked liberal arts schools, students flock to the economics major.
As a former economics PhD student, I've seen first hand the breadth of fields that economists end up in. From professors in a law school, to FTC economists, to public policy and healthcare, economists will always end up in a diverse set of fields. That's the nature of economics -- it does not teach you a set of facts. Instead, it teaches you how to solve problems. Whether anyone wants to admit it or not, every industry has problems that boil down to raw economics.
As a graduate of (rival) Middlebury College, I am both impressed an un-surprised. Great visualization, but I'm not too surprised by the results. Accompanying percentages might make the results easier to digest.
I would be interested to see similar correlations for more specialized schools (Caltech / MIT for instance) as a contrast to liberal arts education.
Agreed. As a Bowdoin alum, this showed exactly what I'd expect from a great liberal arts college. But perhaps that's its value, for the parent concerned about their kid's choice of a "soft" major.
Pretty graph but how useful is it ? There is lots of data on it and it looks crowded and seems hard to interpret unless you use the right panel to filter down the data. Even then, no numbers mean that you only get a feeling of relative quantities. The interesting info for me is the "Division 1" graph showing large numbers of Arts/Literature people going into Technology/Engineering/medicine.
How are the majors grouped within each division, and how are the careers grouped? It looks like the careers and the majors are both grouped from least to most quantitative (see, for example, the fact that economics and psychology are towards the bottom of group 2). Is this done in some systematic way or just by eyeballing?
The categories are a little confusing. If you go into scientific research, say at a publicly funded institution, where does that put you? College Education? Technology? Engineering? Government? or just Other?
Every "ribbon" tapers to nothing in the middle, as if one were viewing a ribbon with a half-twist in it. Is this an intentional part of the visualization (to prevent the center from becoming overly crowded)?
Would be cool to also see majors displayed as equally-sized arcs in a separate viz, so that you could discern the relative percentages of each major that go into a given field.
It's cool that they share this data, but that design reminds of the bad examples Edward Tufte includes in his The Visual Display of Quantitative Information.
With regards to this visualization in particular, it is hard to judge which majors are disproportionately represented in any individual field.
There are in some places gaps in the spacing which make it appear as if a particular area is underrepresented; for instance, observe the apparent gap in Div I majors between Engineering and Health/Medicine.
It is very hard to judge the relative distribution of fields for small majors, such as Geoscience.
Path density seems exaggerated when the paths are drawn between nearby cells; compare the apparent density of paths between Div I <-> Arts/Writing/Social Science/Govt/etc with Div I <-> Education/Health. This seems to be because of the thinning of lines in the middle of the figure.
In reading about this type of figure, I managed to find possibly the most incomprehensible infographic I've ever seen:
http://circos.ca/intro/general_data/img/circos-car-purchase-...