Column: Hot-Shot Programmers Won’t Quickly Fix Health Exchanges
A recent news item reported that three programmers, all in their 20s, had whipped up an alternative to the U.S. health-insurance exchange system and claimed it was much more effective than the government’s system.
The legend of the 20-something coding genius is so powerful that a major news organization seemed to accept that it should be this simple to fix the exchanges. Of course it isn’t, and boasts such as this are masking the complexity of the challenge.
The Department of Health and Human Services has been mum on the details of the technological problems. But the picture that has emerged so far is that the failure of the Obamacare exchanges was largely the result of a lack of understanding of systems-performance issues, rather than insufficient programming skill per se. Software that works fine on a small scale can be a spectacular failure when writ large.
And expertise in performance comes from experience, especially experience with large-scale hardware and large-scale usage — exactly what most 20-year-old geeks lack. Many talented kids might write great apps for iPhones, but few have access to large-scale, distributed hardware, and even fewer have written apps that require such large systems.
In fact, youthful enthusiasm, normally a boon to the quality of projects, may be a negative in this case. The 20-something may be so enamored of the New, New Thing in programming languages and techniques that he might be fitting the problem to the tool, rather than vice versa, and end up developing something that is quite inefficient for the job at hand.
Many of those who worked on the health-insurance exchanges — HIX, for short — are probably in their 20s, too. This has been a standard in the industry, as the young are cheaper, in both wages and benefits, and are more willing to work long hours. (CGI Group Inc., the chief contractor on HIX, hires a large number of non-U.S. workers with H-1B visas, and most H-1Bs in the field are under 30.)
Young workers typically haven’t had the experience of seeing a project blow up in their face. Sadly, many who worked on HIX now have had such an experience, though they may not highlight this on their resumes.
A less experienced programmer may write great code, but have done so only in small-scale settings. Asked to work on a much larger scale, he may not realize the consequences of the manner in which he designs his code.
For example, an expert cited by Reuters found that a single mouse click in HIX triggers 92 separate file openings and network transactions. When an iPhone app opens a file, there is no noticeable delay, and those who worked on HIX may not have thought that 92 such events would be excessive, when multiplied by tens of thousands of users. But files being opened tens of thousands of times per second could present a big problem if not handled with a deep knowledge of computer systems.
Computer-science curricula (both in the United States and abroad) are to blame, too. A typical graduate is reasonably knowledgeable about computer programming, but knows shockingly little about computer systems. Most graduates couldn’t explain something as fundamental as how an operating system boots up, for instance. So they may leave school equipped to write iPhone apps (they can probably do that by the end of their second year of study) but with no inkling of such matters as the time overhead involved in accessing a file or transmitting a network message.
Indeed, very few curricula even require a course in networks, the backbone of modern information technology. Computer-science professors often admonish students not to equate the field with just programming, and every program requires a course in the large-scale behavior of algorithms. Yet little attention is paid to the computers themselves.
Still, performance issues aside, some howler-level bugs are undoubtedly lurking here and there in the HIX software, and, in this, the employers of programmers are culpable as well. Many studies have shown a vast range in ability among programmers.
Since computer-science curricula are jobs-oriented, they tend to attract a lot of students who lack a real excitement about the material, and who are rote-memory learner types, a potential disaster for a field that relies on creativity and abstraction. Yet many companies treat programmers as exchangeable commodities. And companies that depend on being the low bidder for government contracts may be especially prone to hiring cheap commodity programmers.
Any software problem can be fixed eventually. But it will be a rocky road for the health-care exchanges for some time to come — no matter what you hear from a few 20-something computer whizzes.
Norman Matloff is a professor of computer science at the University of California at Davis.