Spam was just driving out the last of the Moors

How do historical gaffs and errors work in a digital age? In this edition of my Techist blog, we take a look at the history that drives search engine results, and at how historical errors can take on a life of their own thanks to computational thinking.

You can read the blog here: https://techist.tumblr.com/post/183237579702/spam-was-just-then-driving-out-the-last-of-the

or read a full text version below without the pictures. Comments and feedback welcome!

---------------------------

A hunt for an out of print book destroyed by our campus flood last year ended with a most delightful addition to my library. I was already a fan of Nehru’s Glimpses of World History, but the version I now have looks to have been made in the printing equivalent of an illicit car chop shop. Page numbers are scattered randomly across the pages, and the text is full of scanning errors, including the intriguing suggestion that spam might be a driver in world history.

Historical errors like these are beguiling, and a little window into the soul of an old discipline that has something to contribute to new disciplines such as computer science.

The fact and feeling of historical errors are arguably a persistent part of human history, but why and how they happen changes over time, as do forms of response. We might assume, understandably, that the most removes a text is from an event or exemplar text, the more likely it is to contain errors. Such textual entropy is possible, but changes can also result from spelling and grammar not being stable, wilful editorial intervention or customary practices of textual improvement, subtle bias and outright hatred, and the copying or dissemination process itself.

These kinds of changes are still with us, despite the measures of mocking, rage and digital ditch dying doled out in response to public expressions of history. If you want to get a taste of it, read the somewhat predictable response to Sam Fallon’s argument in The Chronicle that professors of history need to lay off the pedantry in response to the rich fields of historical error that Donald Trump seems to be ploughing, and get more in touch with their figurative side. Or you could peruse moviemistakes.com, in which no historical error is too small to engender critical delight. There are just under 280 mistakes listed for Titanic, for example, but it is soundly outscored and out voted by the likes of Blackhawk Down and 500.

Outrage cycles can fuel the afterlife of historical gaffs and misinformation by leaders like Trump and our own Prime Minister. The latter was right to elevate the role of crab walking in history making. If a crab slips up, it is unlikely to land the task of getting back to its burrow because crabs track their strides compulsively.

We are computationally compulsive, too. The fabled Bowling Green Massacre has a Google trend profile but our Prime Minister’s ill-founded suggestion that Cook circumnavigated Australia does not. It all comes down to the numbers. More people talked about the former than the latter, and that generates tracking, even if the traces mask belief and mockery alike.

But the suggestion that Captain Cook did circumnavigate Australia does pop up at number five with the auto complete function of Google search. Fortunately, the results that come up make it clear that Cook didn’t circumnavigate Australia, at least in my browser search. And there’s the rub.

Search engines can vote up on popularity, just like the error spotters in the movie mistakes site. The more people that get something wrong, the more likely you are to find it. In this way, historical mistakes can be like spelling errors, search engine designers have worked to accommodate. If I misspell ‘Holocaust’ as ‘Holocost’, for example, Google replies first with ‘Showing results for Holocaust’, but it still offers me the smaller option of ‘Search instead for holocost’ if I really want it. The results acknowledge the mistakes that people most commonly make.

If, however, I type in the rather legendary historical error ‘Francis Drake circumcised the world’ then I get hundreds of results discussing the blooper, but no ‘did you mean?’ message. The error has taken on a life of its own because so many people have discussed it. This makes me wonder why search engine design does not include the most common historical errors as well as the most common spelling errors. Because it is hard, and fraught, no doubt, as the Ngram for “Hitler did not die” shows us. Publishing something in a book does not make it right, and it takes human intelligence to determine whether some things are right and whether we can even be certain about some things in the past.

But popularity is not our only problem, and indeed ranking results by popularity has arguably given way to an age of optimised results. As Lucas Introna and Helen Nissenbaum argued in their now classic paper ‘Shaping the Web’, results reflect the interest of paid advertisers, and people who pay enough or know enough about how to influence the ordering of results. To this we now have to add personalisation: my search results will depend on the profile Google has built for me using those cookies I agreed to in visiting site after site. No wonder the ABC comes up as a source when I search on Captain Cook.

In the world of personalised searches, it is highly feasible for two people to search on the same historical topic and to get two different results. A savvy promulgator of error will know that, and know how to optimise results for people I teach, and people who don’t study history at school or university. And I will wonder, over and over again, how those folks can be so na?ve or misled. The view from my window on the past is fine, but I hadn’t realised there was more than one window all along.

History of another kind is driving the results that I get. My own. The Auditing Alogrithms team at Northeastern University have found that just over 11% of search results are personalised, with factual queries least personalised, and those connected with shopping or politics most likely to be personalised. They have even tracked cases where Google personalises country borders in areas of political dispute. So some history queries will generate more personalised results than others, with the most disputed likely to be the most personalised.

Sam Wineberg is right to ask the question Why Learn History (When It’s Already on your Phone). But the answer is not just about school students. It is about all of us, and the histories that we make every day, likely without knowing. Time to uncover the logic of history—errors and all—in the twenty-first century.