Stories

The IRE Resource Center is a major research library containing more than 27,000 investigative stories.

Most of our stories are not available for download but can be easily ordered by contacting the Resource Center directly at 573-882-3364 or rescntr@ire.org where a researcher can help you pinpoint what you need.

Search results for "machine learning" ...

  • Aggression Detectors: The Unproven, Invasive Surveillance Technology Schools Are Using to Monitor Students

    In response to mass shootings, some schools and hospitals have been installing devices equipped with machine learning algorithms that purport to identify stressed and angry voices before violence erupts. Our analysis found this technology unreliable. Our goal was to reverse-engineer the algorithm, so we could see for ourselves if it actually worked as the company advertised. (One salesperson suggested to us that the device could prevent the next school shooting.) We purchased the device and rewired its programming so we could feed it any sound clip of our choosing. We then played gigabytes of sound files for the algorithm and measured its prediction for each. After this preliminary testing, we ran several real-world experiments to test where the algorithm could be flawed. We recorded the voices of high school students in real-world situations, collected the algorithm's predictions and analyzed them.
  • Artificial Unintelligence

    A guide to understanding the inner workings and outer limits of technology and why we should never assume that computers always get it right.
  • Message Machine

    “Message Machine,” a news application, takes an innovative approach to decode how the presidential campaigns were shaping fundraising appeals and other communications to potential voters. Going into the election cycle, there were reports that the presidential campaigns were gearing up to use "data science" -- sophisticated quantitative analysis and statistics -- to "microtarget" messages as never before. They had little interest in explaining what they were doing, however, for obvious strategic reasons. When a couple we knew told us they'd received similar emails simultaneously from the Obama campaign, each asking for donations, but in language that differed in subtle, but important ways, we set out to reverse engineer how the campaign was altering its tone and content to specific audiences. ProPublica News Application Developer Jeff Larson created a system to automatically gather tens of thousands of campaign emails and analyze how they were targeted. He used the same sophisticated techniques, such as machine learning and natural language processing, used by the campaign. There were a huge number of stories done after the campaign about the “geniuses” at Obama For America, but ProPublica was virtually alone in providing real-time, deep analysis of the operation while the campaign was still happening.