Monday, July 7, 2014

Methodology Monday with Robots

From Mashable (hat tip to Tara Calishain at ResearchBuzz):

"Quarterly corporate earnings reports from the AP will soon be produced through a computer program that is able to take the key numbers from companies' results to create a story of 150 to 300 words, the media company announced in a blog post. . . . The AP isn't the only organization using journobots. The New York Times uses automation for some of its wedding announcements, while Automated Insights also provides recaps for fantasy football matchups. . . . the AP is looking at using automation on 'results stories for lower-audience sports.'"

(I omit mention of TV announcers as they may well already be robots, albeit rather excitable ones.)

The same thing is happening to genealogy, bit by bit, although not in exactly the same ways.

Look for information and a robot may offer you "hints." Some are wildly wrong, some may be helpful.

Enter the resulting data and your genealogy program will produce a "report." Granted, it reads like it was written by robots, but they are getting smarter all the time. And of course the report in any case is only as good as the data on which it is based.

Robots are getting better at distinguishing kinds of text -- for instance, in searching city directories, knowing enough to distinguish the "Jones" in "Ralph Jones" from the "Jones" in "Jones Street." They may be soon reading handwriting, a function that once required a human being, often an astute one.

Interestingly, many stages of automation involve a certain sacrifice of quality, much as cell phones give up in sound quality some of what they gain in portability. Robots make mistakes in indexing that humans would never make. We tolerate these foibles because on balance the robots usually make our lives easier, but the core of genealogy is not something that can be averaged out -- it's either the way things were back then, or it's not.

What's my point? Not to complain. It's just this: Within some lifetimes now begun, advanced genealogy will be the only genealogy requiring human involvement:
  • looking in physical places where robots don't know and can't go, either because the materials aren't digitized or because no one thought of "that stuff" as being genealogically relevant;
  • distinguishing bogus robotic "hints" from useful ones, and otherwise fixing robotic errors in their output;
  • resolving conflicting evidence;
  • analyzing and correlating complex collections of evidence properly; and
  • writing a coherent and convincing proof argument.
 Have a nice century.

"The Associated Press Now Automates Earnings Stories, No Humans Needed," Mashable ( : viewed 1 July 2014).

Photo credit: Neil Milne's photostream, "Cheery Robot Lazer Attack Nail Art" ( : viewed 1 July 2014), per Creative Commons.


Harold Henderson, "Methodology Monday with Robots," Midwestern Microhistory: A Genealogy Blog, posted 7 July 2014 ( : viewed [date]). [Please feel free to link to the specific post if you prefer.]

1 comment:

Eileen said...

What I'm wondering is if I could get my pedigree painted on my nails?