How Do We Know What We Know?
History as a "Collective Hunch":
Trudy and H-Bot
Reality is nothing but a collective hunch.
Trudy the bag lady in Lily Tomlin's The Search for Signs of Intelligent
Life
in the Universe
Trudy's line is funny but telling. It can illuminate how we view
history and conduct research in a web-mediated age.
Researchers at the Center for History and New Media at George Mason
University developed "H-Bot", which they describe as a History Software
Agent. This fascinating and talented 'bot is extremely competent at
guessing important dates, and identifying historical figures. It scored
82% on the 4th grade National Assessment of Educational Progress U.S.
History exam.
H-Bot is not concerned with "facts." Instead, it combs sites in the
Google database and arrives at a consensus opinion on when certain
events happened, or what a certain person might have done. In other
words, to H-Bot, historical facts are a collective hunch!
But H-Bot is not perfect. Some of its limitations are spelled out in
a
great article written by a co-developer.
For example, H-Bot has difficulty differentiating among people with the
same name. H-Bot can also reflect true uncertainty over certain dates,
such as the birth date of founding father Alexander Hamilton.
More interestingly for our purposes, it turns out that H-Bot is quite
susceptible to widely published pseudo-facts. If you ask it when the
aliens landed in Roswell, New Mexico, H-Bot "correctly" answers 1947. Of
course, not all pseudo-facts smack you in the face quite as vividly as
this one.
My favorite example of how H-Bot (or other seemingly careful
research) can go badly wrong occurs when different sites on the web tie
back to a single, suspect source. For example, I asked H-Bot "when were jews expelled from spain" and it replied 1491, though the actual year is
1492. The error had been cloned across the websites that H-Bot drew upon
that day. (You will likely get different results if you search now,
since the underlying Google corpus changes constantly. To
paraphrase Heraclitus
(1), you can't step
twice into the same H-Bot search.)
H-Bot search, November 15, 2006
When pseudo-facts are combined with cloned or plagiarized content, it
makes for some very authoritative-looking misinformation.
Mobile, Alabama to the Ohio River
Here's a simple question, drawn from our earlier investigation into
the Drinking Gourd song: How long would it take a slave to escape
from Alabama to the Ohio River? According to all these
sites (ca. early 2007), the answer was a year.
• "Since it took most escapees a year to travel from the South to the
Ohio..."
•
Madison (WI) Metropolitan School District (taken offline in February,
2007)
•
NASA
•
University of Wisconsin,
Stevens Point
•
Delta College, MI
•
University of California, San Diego
•
Utah Educational Network
•
New York State Council for the Social Studies
• "A slave who left...southern Alabama or Mississippi in the winter
would arrive at the Ohio river about a year later" (appears in the
following sites)
•
Iowa Public Television
•
Oklahoma Baptist University
•
University of Kansas
So if websites from NASA to the University of Wisconsin to Iowa
Public Television agree, then it must be true. H-Bot would probably love
these sources. But how do the authors of these websites know? And how to
account for this amazing unanimity of opinion?
100+ Years in the Life of a "Fact"
Factoids are stubborn things
With apologies to President John Adams, who said no such
thing. (Factoid: An invented fact believed to be true because of
its appearance in print – Webster's)
The estimate of a year's travel time is drawn from the account of
exactly one slave, first published in James Freeman Clarke's
Anti-Slavery Days in 1883. Clarke writes, "I was once, with my wife,
in Columbus, Ohio, and having a day to spare, we employed it in visiting
the public institutions. Among other places we went to the Penitentiary,
and were introduced by the warden to a colored man who had escaped from
Alabama. He had taken a whole year in coming from Alabama to
Cincinnati."
But Professor Loren Schweninger wrote me: "My guess is that the story
of a one-year journey to the North is apocryphal. A slave who…had some
idea of the terrain could make it to the North in a matter of weeks, as
a direct northerly route would be only a few hundred miles. A daunting
undertaking at that."
(2)
An account in Alexander Milton Ross's Recollections and
Experiences of an Abolitionist supports Professor Schweninger. Ross
tells of two slaves who left from Columbus, Mississippi, in the upper
watershed of the Tombigbee. They made it to the Ohio River in 17 days.
The distance from Columbus to Paducah is 256 miles, or just over half
that from Mobile to Paducah.
Desperate slaves on the run in hostile territory would aim to cover as much
territory as possible. The pair of escaping slaves described by Ross
managed 15 miles per day. The subject in the suspect Clarke account
covered roughly 1.7 miles per day (assuming a starting point of Mobile,
the distance to Cincinnati is 618 miles, divided by 365.) According to
the Educator's Guide, slaves managed to advance northward only
about 1.2 miles per day (440 miles, divided by 365.)
Factual or not, the Clarke account was cited in Larry Gara's The
Liberty Line and then used in a history book for juveniles by
Shaaron Cosner. As shown in the table below, this "fact" was transformed
into a conjecture, into a qualified fact and last, into a completely new
"fact."
| QUOTATION |
SOURCE |
COMMENTS |
| (We) "were introduced by the warden
to a colored man who had escaped from Alabama. He had taken a
whole year in coming from Alabama to Cincinnati." |
Clarke, James Freeman.
Anti-Slavery Days. A Sketch of the Struggle which Ended in the
Abolition of Slavery in the United States, 1883. p. 92 |
The original "fact." |
| "One fugitive took a year to go
from Alabama to Ohio..." |
Gara, Larry, The Liberty Line,
The Legend of the Underground Railroad, 1961, p. 46 |
Accurately recounts the original "fact." |
| "...it took one slave a year to get
from Alabama to Ohio." |
Cosner, Shaaron, The Underground
Railroad, 1991, p 71. |
Accurately copies the Gara account. |
| "We know one slave traveled for a
year from somewhere in Alabama to the Ohio and one may assume
that most slaves fleeing the Deep South needed at least a year
to reach the Ohio." |
Rall, Gloria D., The Planetarian,
1994 |
Starts with the original "fact" and then
conjectures that this proves a general case. |
| "Since it took most escapees a year
to travel from the South to the Ohio..." |
Educator's Guide to Follow the
Drinking Gourd, 1995 |
Drops the original fact, but keeps the conjecture.
|
| "Since it took most escapees a year
to travel from the South to the Ohio..." |
Madison (WI) Metro. School District
website, 1998-2007 |
Online version of the same (Page removed in
February, 2007.) |
| "It took slaves about a year to
escape from the deep south." |
Maryland Public Television website, 2004
to September, 2007 |
Dropped the qualifier "most." (Page
modified in September, 2007.) |
The transformation is now complete: the questionable account of one
slave now stood for all escapees from the Deep South, on many
important (and credible!) sites on the web.
|
