BETA
This is a BETA experience. You may opt-out by clicking here

More From Forbes

Edit Story

Public Crowdsourcing Vs Researchers In Libraries: Image Searching Lessons From World War II

Following
This article is more than 4 years old.

Getty

In today’s world of Google Images and Google Street View, it can be hard to imagine there was once a time when even the US Government struggled to find imagery of the rest of the world it needed for critical wartime planning. Something as simple as needing photographs of a major harbor to plan a military landing was an enormous exercise in research, involving scouring libraries, placing inquiries to countless photographic archives and even, for a brief moment in 1944, launching a massive wartime crowdsourcing initiative that failed quite spectacularly. Looking back upon this era, there are many lessons we can apply to today’s image-saturated world in which intelligence agencies increasingly dubiously claim they can simply mine Twitter and Instagram for the images they need.

It can be hard to describe in detail all of the ways the web has changed how we access information. Perhaps its greatest contribution has been the way in which the answer to nearly every question is now available with a few keystrokes or increasingly just from speaking to our smart speakers and smartphones.

Today we don’t even think twice about typing a location, celebrity, product, animal, plant or anything else imaginable into Google Images to see what it looks like. Instead of keyword searching for a strange plant and clicking fruitlessly on search result after search result hoping to find a picture of it instead of just another textual mention, Google Images allowed us to separate images from text and search them on their own merits, though still through the dimension of textual captions and context.

To imagine what the world was like before Google Images, try searching Google Books for images of a historical place. You could likely search for hours just to find a handful of actual images. FoxTrot comic Bill Amend has noted on several occasions the difficulty of drawing everything from motorcycle officers to the Apollo rocket in the days before Google Images.

Yet, this was the position the US Government was in during World War II when the military desperately needed to gather imagery of critical infrastructure and ports of entry across Europe. Contained within the nation’s libraries, private archives and homes were likely a treasure trove of vital photographic intelligence, but how to sort through all that material to find what it needed?

The Government took a number of approaches to the wartime effort of searching for imagery.

The first was to reach out to news agencies, commercial photographic archives and even Hollywood to acquire any available commercial imagery. Yet, merely acquiring imagery is not enough: it must be triaged to determine its intelligence value and to confirm the subject matter it depicts. As Yale historian Robin Winks put it, initial efforts lacked the necessary subject matter experts such that staff "confused pictures of Hawaiian beaches with the Solomon Islands."

In parallel, the Government turned to professional researchers scouring the nation’s great university libraries. In one highly successful effort, two Yale researchers went "rooting through Italian engineering publications in the Yale library, looking for photographs of hydroelectric plants and docks." This effort was so successful that "by the end of the war the Pictorial Records unit at the Yale library grew to six people, who selected, analyzed and photographed 17,780 strategic photographs,” with the Government concluding that “future work of this kind might best be contracted out to able scholars who knew their own turf.”

In short, the Government was recognizing that simply throwing non-specialists at triaging imagery would not end well. Instead, by leaving such work to subject matter experts who deeply understood their areas of expertise, vastly better results could be obtained.

Sage advice three quarters of a century later when we are all today our own researchers and turn less and less to professional expertise to help us navigate the treacherous online information waters.

In contrast to these professionalized efforts, in 1944 the OSS Pictorial Records unit issued a call to the American public to, as Winks put it "scour their albums, attics and cupboards for snapshots they had taken in Europe and to send these to Pictorial Records with the thought that some might reveal harbors, railway stations and the like." Unfortunately, this three-quarter-century-old experiment in crowdsourcing failed miserably. The Government received "hundreds of thousands of photos ... showing equestrian monuments and sea gulls" with less than 1% of the images actually being useful, of which the majority were taken by professional news photographers.

Even with the rise of television in the decades after, US intelligence agencies cataloged television news programming primarily through textual descriptions, treating it as "radio with pictures."

Putting this all together, it can be hard to imagine there was once a time when we “saw” the world primarily through text, rather than imagery. The US Government’s experiences during World War II remind us that for all its touted benefits, the actual real-world utility of crowdsourcing for production insights can be severely limited. Rather than allowing the government to see the world through the total holdings of the nation’s libraries, photographic archives and civilian tourist images, the government’s ultimate realization was that professional photographs were the most useful and having trained scholars catalog those images, rather than outsourcing their cataloging to non-specialists, yielded the best results.

Technology and instant search has helped the Internet make images more readily accessible and searchable, but the question remains whether the deluge of tourist images on social media and the ability to harness the public at large to help catalog those images are merely repeating the same mistakes we made three quarters of a century ago.

In the end, is crowdsourcing image collection and analysis the future? Or is it merely another failed enterprise that will give way back to expert collection and analysis when the results really matter? Or is our reliance on that crowdsourcing at the very root of our struggles with misinformation and disinformation and "fake news" today? Perhaps the nascent world of deep learning image recognition will help codify the vast knowledge of our professional image researchers and begin, after three quarters of a century, to realize that World War II dream of looking across the world’s information?