METERI was struck by the title of a working paper published by the National Bureau of Economic Research (NBER): How to speak when a machine is listening: corporate disclosure in the age of artificial intelligence. So I clicked and downloaded, like one does. And then he started reading.
The document is an analysis of the 10-K and 10-Q filings that US public companies are required to file with the Securities and Exchange Commission (SEC). The 10-K is a version of a company’s annual report, but without the glossy photos and hype – a corporate nerd’s delight. Has, he says a guide, “The data of everything and the kitchen sink that you can spend hours reviewing, from the geographic source of income to the maturity schedule of the bonds that the company has issued.” Some investors and commentators (including yours) find the 10-K impenetrable, but for those with the necessary stamina (large companies can have 10-Ks that span several hundred pages), that’s the kind of thing they like. . The 10-Q filing is the quarterly little brother of the 10-K.
The observation that triggered the research reported in the paper was that “mechanical” (that is, machine-generated) discharges from 10-K and 10-Q corporate presentations increased from 360,861 in 2003 to approximately 165 million in 2016, when the 78% of all downloads. have been activated at the request of a computer. A great deal of research in AI is now devoted to evaluating how good computers are at extracting actionable meaning from a data tsunami like this. The stakes are high in this, because machine-read report output is the raw material that can power algorithmic traders, robot investment advisers, and quantitative analysts of all kinds.
The NBER researchers, however, analyzed the supply side Tsunami: How companies have adjusted their language and reports for maximum impact with algorithms that read their corporate disclosures. And what they found is instructive for anyone wondering what life would be like in an algorithm-dominated future.
The researchers found that “increasing the number of machine readers and artificial intelligence … motivates companies to prepare presentations that are easier to process and analyze using machines.” So far so predictable. But there is more: “Companies with high expected machine downloads manage textual sentiment and audio emotion in ways tailored to machines and AI readers.”
In other words, machine readability, measured in terms of the ease with which an algorithm can analyze and process information, has become an important factor in business reporting. Therefore, a table in a report can have a low readability score because its format makes it difficult for a machine to recognize it as a table; but the same table could receive a high score for readability if you made effective use of tagging.
The researchers contend, however, that companies now go beyond machine readability to try to adjust the sentiment and tone of their reports in ways that can induce algorithmic “readers” to draw favorable conclusions about the content. They do this by avoiding words that appear as negatives in the criteria given to the text reading algorithms. And they are also adjusting the tones of voice used in standard quarterly conference calls with analysts, because they suspect that those on the other end of the call are using voice analysis software to identify vocal patterns and emotions in their comments.
In a sense, this kind of arms race is predictable in any human endeavor where whoever has better technology can gain an advantage in the marketplace. It’s a bit like the war between Google and so-called “optimizers” trying to figure out how to play. latest version of search engine page ranking algorithm. But on another level, it’s an example of how digital technology is changing us, as Brett Frischmann and Evan Selinger argued in their book. Reengineering of humanity.
After writing the last sentence, I looked up publication information in the book and found myself trying to log into a site that, before I was admitted, required me to solve a visual puzzle: on the image of a road. intersection divided into 8 x 4 squares I had to click on all the squares showing traffic lights. I did so and was immediately presented with another similar puzzle, which I also solved diligently, like an obedient monkey in a laboratory.
And the purpose of this absurd challenge? To convince the computer that hosts the site that I was not a robot. It was an inverted turing test in other words: instead of a machine trying to fool a human into thinking they are human, I was called upon to convince a computer that I was human. It was being redesigned. The road to the future has taken a curious turn.
What i’ve been reading
Why I write about Joan Didion is a absolute gem from the archives of London Magazine.
Charles Duhigg has written an article in the New Yorker called How Venture Capitalists Are Warping Capitalism, in which he uses WeWork as a case study in the madness of the 2020s. Basically, it’s Boo.com of our days, but with contemporary twists of madness and greed.
The problem with Dom
What Dominic Cummings never understood: Impatience is no substitute for politics. Thus argues a perceptual piece at PoliticsHome by Sam Freedman.
Digsmak is a news publisher with over 12 years of reporting experiance; and have published in many industry leading publications and news sites.