OPINION

By Terry Calhoun

On a weekly basis, my three University of Michigan work-study students and I spend several hours scouring the Web for news items and for more substantial resources on a variety of topics related to higher education: IT, campus edge stories, residence life, campus planning, sustainability, you name it. If it relates to higher education, we try to find it.

Our life was made simultaneously easier and more complicated when Google News Alerts became available for us to use. I have set dozens of them and all of that useful information is just sitting there in email folders, ready for me to access when I have the time to go browsing. Unfortunately, a simple feed to a Web document of Google News Alerts results would not serve as a useful product for my readers, as they still produce lots of “false positives.”

In a recent Wired News story, “Man v. Machine in Newsreader War,” the current state of who or what d'es the best job of funneling the most pertinent and comprehensive news content to readers is put in terms of Human versus Machine:

• Think back to John Henry racing a steam drill and forward
  to Garry Kasparov trying to outmaneuver IBM's Deep Blue
  in 1997 to the Onion tweaking the genre with its “Accountant Battles Excel” story.

Just like John Henry, Garry Kasparov, and Wallace Peters, the writer of the Wired News story, Ryan Singel, is probably correct in his conclusion that human filters will be replaced by better automated news readers in the future. However, I think that future is still quite a ways off.

Let’s take a look at one of my regular Google News Alerts: “student center” university new. You can guess that I am trying to stay up to date on what’s going on with the planning, design, construction, and use of newer student centers on college and university campuses. Yet my Google Alert for March 13th found: “Pondexter Named Wade Trophy Finalist” and “Sports Group Offers Support to Disabled Athletes, Coaches.”

You can argue that my search terms in this example are pretty poorly done, and I would tend to agree. We’re still working – and probably will always be working – on refining the search terms. As it stands, though, we have to look back through a week’s worth of search results and filter out lots of unwanted links before we find a story about San Diego State University students voting themselves a fee increase in order to build a new student center.

In his article, Singel is contrasting companies like Digg, which use human filtering agents, with companies that use artificial intelligence (or as he puts it “artificial, artificial intelligence”) to produce filtering results. Google is an example of one of these. In my experience it takes a human being, using technology filters as a tool, to produce the best results.

When I think of the best daily or weekly email newsletters in higher education – one of the longest-running of which is my own SCUP Email News, which has been regularly transmitted since 1987 – the best share news blurbs and links written by cyborgs: intelligent human agents using intelligent search engines;

Limiting that list to newsletters that point to items outside their own Web sites, I’m talking about the whole string of Campus Technology newsletters like IT Trends, as well as Edupage, UB Daily, Academic Impressions, and so forth.

Of course I get them all, but I find it amazing how many pertinent news items my students and I can find that still don’t make their way into any of the other newsletters.

One weakness of machine searches is that the best they can provide to tell you the true theme of an article is (a) the title and (b) some brief quotation from within the article. These quotes are almost always taken from the very first few lines. Unfortunately for the fully-automated model, titles don’t always reflect content and the first paragraph in an article d'esn’t always tell the story of the entire article.

Human agents, like my students or myself, can either (a) abstract the content and write a presumably interesting brief description of it (like here in IT Trends) or take one or two paragraphs from throughout the article, selecting the lines that best represent the content, and reproduce them with the link (like in SCUP Email News).

We’ve also found that there are a lot of what we consider “news items” of pertinence that do not show up in Google Alerts. I don’t know for certain, but apparently not every news provider in higher education is indexed for Google Alerts. We’ve found that we do need the alerts to thoroughly cover a topic, but we also need to search in Google overall. Lots of campus-based student newspapers, for example, seem not to be indexed for alerts, although they are in Google itself.

So for now, the human touch is needed. I think it will be needed for quite some time to come. Perhaps things will change if, in 10 or 15 years, we can get flash memory chips with 10 terabytes in them, approximating the storage capacity of the human brain. But even then, I bet on the cyborgs. Already, “wearable” computing has moved from science fiction to reality, and we have all sorts of people sticking RFID chips into themselves.

I think that as the machine algorithms get better, so will our ability to slap “wetware” onto them. And in that there’s going to be a human filtering component to all news reporting, if not news gathering.