The Washington Post has a vignette about the "economic assistants" who work for the Bureau of Labor Statistics, checking the prices of all manner of goods and services nationwide. The work these part-timers do feeds into the nation's basic economic data (e.g., the Consumer Price Index) and thus is critical to a lot of people.
I've heard of these people's work before but the intriguing thing now is whether the job can be farmed out to a computer. (The Post itself seems ambivalent about this point, with the text of the article claiming that "as the Bureau looks for ways to modernize ... this century-old job could be in jeopardy," while the subheading on it says this is "still an in-person job.")
I'd say real people are going to be needed for a good long while yet. At the crux of the matter, as the subject of the article, Ms. Gaffney, points out, is that accuracy can only be guaranteed by ensuring that the "price" being quoted for any item across different sales portals is for a completely comparable product. The difference between bread crumbs and "Italian" bread crumbs, organic and non-organic apples, gasoline that's on special Tuesdays as opposed to the "regular price," or a shirt that's 30% cotton vs. one that's 50% -- these things affect prices.
So, could the finer points of this comparability be adequately handled by a computer checking bar codes, or scanning the inventories of stores? The Bureau official waffles a bit, but I'd say, absolutely not.
The problem is the same as for us poor consumers shopping online. If you're at all concerned about details of what you're buying, you'll be stymied at every turn online. Some retailers fully describe a garment: the mix of fibers in the fabric, for example -- but most do not. "Cotton, polyester, and rayon" is meaningless unless percentages are given. Looking for a laptop or tablet? Many offerings you'll see don't even bother to tell you how much memory it has. And when you account for plain old sloppy data entry ... a product may be shown as containing no fragrance although it does; the different volumes of a multi-volume book won't be distinguished from one another, and on and on.
In short, it's a world of inaccurate, outdated, and sometimes even deceptive, information out there. Is this the appropriate basis for statistics that are fundamental to our understanding of our economy? Hardly. It's difficult to see how the human element (Ms. Gaffney) could be dispensed with at all; certainly not in the near future.