Saturday, October 8, 2011

The Internet And How We Consume Information

Okay, moving on from League of Legends to a more substantial topic. A few years back, the Atlantic writes about how the Internet affects our behavior patterns in an article titled Is Google Making Us Stupid?. The author, Nicholas Carr, argues that the flitting browsing pattern of most Internet users suggests a deterioration of deep reading and deep thinking; as people rely on machines to bring them relevant and interesting content, Carr writes, they lose the ability to penetrate dense information to find that content themselves.

I disagree with a number of points, and writing about it was interesting enough to make a blog post of it, so here goes...

First, I think the idea of Internet-induced ADD is a self-fulfilling prophecy and a scapegoat. Maintaining concentration is not an inherently easy thing to do; it's not like people didn't have trouble with that before the Internet came along. But it's easy to blame one's concentration issues on Internet usage affecting one's brain, so people do it. This issue is only exacerbated by misuse of research like the cited study of online research habits, which looks at how people browse a research site; when the author suggests that this study evidences a broader change in reading and thinking patterns as a result of Internet usage, he's speculating without basis, because the study examines Internet usage rather than broader reading and thinking patterns. Indeed, the researchers themselves are doing nothing but speculating in the quotation the author chose, when they conjecture that Internet users "go online to avoid reading in the traditional sense." Did they study how Internet users' offline reading behavior changed? No. They just looked at the common-sense skimming approach one uses on an online research site, and jumped to a wild conclusion about reading in general. This sort of confirmation bias is poor scholarship, but I would bet it's typical of research done on this topic.

Second, I find it misleading to examine Internet usage only through the lens of comparison to a bookworm. As Cracked pointed out in its article about proscribed student behaviors that are good for studying, the dichotomy in many cases is not between texting and essay writing, but between texting and nothing; in our case, the alternative to the Internet is not necessarily the library. The Internet caters to a wide audience, so online material is perforce required to be available in a form that is comprehensible to the lowest common denominator, but that does not preclude greater depth for those who seek it. That's why, in addition to Google, we have tools for deep search and exploration of the 'invisible Web' of private databases and dynamically generated pages.

Third, a brief diversion: I disagree with the suggestion that there is a clear financial incentive to promote shallow online browsing. For example, just recently my mother and I were discussing a customer intelligence company that is using the information it collects about people online to, among other things, promote prolonged browsing on particular websites for greater ad revenues per viewer. That seems to be a direct contradiction of the author's point. The Internet is moving past the era of page views and click-throughs; in order to attract continued consumer attention, content has reclaimed its rightful place as the driver of revenue.

Finally, the author's warnings about how Brin/Page's blithe assumptions about the benefits of AI characterize intelligence as reducible to a series of mechanical steps miss the mark by a long way. Intelligence can harness mechanized processes without itself being purely mechanical. Consider the analogy of a programmer with many languages at his disposal, at various levels of abstraction, with work below the level of abstraction in each case being done by a machine. Is the Python programmer more mechanical or less natively intelligent than the C programmer because his language relies more on interpretation by the compiler? Nonsense. And if you desire functionality from C that's not natively available in Python, you can always wrap a C library and bring it up to the appropriate level of abstraction. Similarly, harnessing artificial intelligence to manage the presentation of information for consumption does not inhibit reading, even if in some sense we are delegating the task of reading to the machine. Rather, it allows us to get more out of our reading at every level of engagement. Machine intelligence enables human intelligence, rather than replacing it.