I just completed recommending one of my 6.171 graduates for a master’s in computer science (I guess he isn’t that smart, since he wants to go back to school instead of working for Google). It was a great tour of just how painful Web applications can be.
Harvard University rejected my recommendation because the software insisted on me figuring out what exact department the kid was applying to and what degree. The menu offered about 300 choices, the first 10 of which started with “African-“. None of the CS choices worked and I eventually tried the African- degrees but they weren’t accepted either.
Many of the schools allowed the uploading of a letter, but never in HTML, the original format of my letter. Sometimes it had to be Microsoft Word. Sometimes it had to be RTF. Sometimes it had to be PDF.
Nearly every application required at least two extra keystrokes, e.g., “sign here to confirm the recommendation” would lead to a page with a button to “proceed to the submit page” that would lead to a page with a button to “submit the recommendation”.
University of California is an interesting institution. There was no commonality among procedures or software at four U.C. CS departments. One department insisted on 100 percent paper. The others each had their own peculiar Web software.
If you didn’t know better, the experience might lead you to believe that the average Google employee knew more about building software than the average professor of Computer Science…
Why would a professor of CS build a web app for the University? The tenured ones don’t care and it’s a waste of an untenured professor’s time. Generally speaking, that sort of thing is built by an outside vendor (or possibly by the internal IT department, completely unrelated to CS) as part of some sort of ill-conceived University-wide standardization effort
Google is not that different from UC. They have 4 different RSS readers with no interopability (Google Reader, one in each of iGoogle, Gmail and Google Desktop). Four different clients for Google Talk (Windows client, Flash client, Gmail and Orkut) and each one supports a different subset of features. Only Gmail is integrated with AIM, but doesn’t support voice chat*. The Windows client doesn’t do group chats, Flash doesn’t send files. Talk about chaos.
Furthermore, UC has an advantage, as the more complicated app process is a natural filter for people that don’t really want to go there, but are trying 20 diff. schools. Google, on the other end, doesn’t have an advantage from the lack of predictability.
Do you care what school they go to?
Of not then what about a central repository for recommendations, GRE scores, essays, resumes, etc. etc. It will have to in a very specific format (probably of a generic type). Do all validation checking in one place.Then just send the schools the appropriate login to access said information.
Schools can extract information automatically and convert it to whatever format they wish.
Alek, you’re not being fair to Google. All of those products are in beta, so you really can’t count them. Wait until they’re completed to give a true assessment. I’ll be back here in the year 3039 to see your updated overview.
It’s the same situation for CS faculty applications. Every school/department has their own online system, so the applicant has to type in the same information in a slightly different format for every school to which he applies. It takes far longer for the applicant to apply using the web than it would take to make paper copies of all the applications materials and send them in the mail.
Regarding the student “not being that smart”, I think that getting an MS and then going into industry may actually be the financially optimal path. I think the additional earning power of the MS outweighs the two lost years of wages and compounded earnings on investments
Getting a PhD in CS, however, is definitely financially irrational. A postdoc in CS pays better than postdocs in most fields, but still far less than the salary that a smart person with a BSCS and 6 years of experience would get. And the lost wages and compounded interest from 6 years of grad school result in a huge loss of net worth down the line.
Finally, you said “If you didn’t know better, the experience might lead you to believe that the average Google employee knew more about building software than the average professor of Computer Science…” That’s definitely true. The typical CS professor gets a BS, during which he learns the rudiments of software development. Then he goes to grad school, where he works smallish projects that consist mostly of research code that will probably never be used by anyone but him. However, having had no outside software development experience, he things that these little projects are actually quite big and is deluded into thinking that his grad experience has trained him as a software developer. Then he goes on to a faculty job, where he never has to develop any code again.
Until potential grad students start making decisions about what school to buy their degree from based on the quality of their school’s web apps, what incentive do they have to make it work well? Government web applications (the Federal Direct Loan Servicing website and Harris County Toll Road Authority website come to my mind) have the same problem. Just because a CS prof may know how to build a good web application doesn’t mean it isn’t still a lot of work. If you don’t have to do work, why should you? (Put less pessimistically, why not focus your energy on other, more important things, like the research you’re actually being paid to do?)
I went looking for a cast iron tub drop-in (ie: not claw-foot or stand-alone) that’s deeper than the stock sizes recently. You could pick any online vendor of bathtubs and find a ton of ways to improve the shopping experience.
Might be shooting fish in a barrel, and stuff that any undergrad could solve, but there’s certainly lots of good stuff to be gleaned from there.
Alex: Your idea of “Let’s build an organization that does a really bad job in every area where we don’t think it matters and is excellent in the places and ways that matter to our bottom line” sounds sensible. You would certainly not want sentimental ideas such as organizational pride to cause resources to be wasted on stuff like Web applications, teaching undergraduates, etc. However, I think it has turned out to be tough to manage people in this way. If you say to workers “excellence only matters occasionally” they hear “sloppiness and mediocrity are almost always acceptable, so I will go home at 3:30 pm and not bother to finish or check anything”.
This is why when you pull up at a high quality FBO in a $30,000 1970 Cessna 172 and say that you don’t need any gas, they bring out the red carpet, wash the windshield, hand you the keys for a crew car, and ask what else they can do to help. The management recognizes that it would be more profitable to put all efforts into serving the jet fuel customers, but it is hard to get staff to switch behavior.
Same deal at management consulting companies. They are obsessed with details in every corner of the company, even ones that customers will never see.
You might think that although these web application systems make life hell for letter writers, they are very nice to use for those reviewing huge piles of applicants. This is not the case. The product used by my department, ApplyYourself, is an HCI nightmare. The UI is a confusing mess of oddly-sized windows and popups, there are multiple places where a letter of recommendation or a comment by one of my colleagues might be found, and the system for rating applicants is bizarre and nonintuitive. We end up passing around lists of interesting applicants by email. One year I spent a fair amount of time making specific recommendations that were never implemented.
I linked your blog entry about poor information systems to the Andromeda Project’s source forge mailing list.
Also, the assumption you are making is that these information systems are built by computer science professors. However, I will bet you my loose change dish that some of these systems were written by undergraduates financed by the Federal Work Study Program and other government sponsored or university sponsored agencies. Think: Why would there be no commonality across departments? What market force creates this unique tension?
As an aside, at least one graduate school I wanted to apply to ended up off my interest list because I had literal vitriol and hatred for their application form. Many universities spend money sending fliers via postal mail, and all of these fliers advertise the website prominently! but don’t bother to improve their UI!
John: I never said in my posting that the professors had to build the applications themselves. They could have spec’d them for undergraduates to build, as you suggest, but I don’t think that is the case because in most cases the systems that I was using had .com addresses and seemed to be commercial vendors. I don’t think these commercial vendors would have had much luck selling to any customers besides university CS departments, though, because the user interface quality was so poor.
Phil: I misread your conclusion paragraph when you said “the experience might lead you to believe that the average Google employee knew more about BUILDING software than the average professor of Computer Science”. Sometimes sentences live and die on the tiniest calibrations. My apologies for shooting you down on it – I didn’t mean to detract from the otherwise stellar points you made.
Try finding a job using the time-saving web tools that power most big companies’ web hiring. Even though most Fortune 500 companies use Taleos, there is no way to share the same information across different companies. Instead, you’re asked to register, log in, fill out the same basic information, upload the same CV, and answer the same “Are you an axe murderer?” questions. Even if you weren’t an axe muderer, by the time you finish applying to ten companies, you will be.