American intelligence analysts are biased, and therefore make lousy decisions — even the spooky agencies admit that. The spy guys’ new hope for introducing some objectivity: Get the analysts to start playing a videogame.
“A Serious Game could provide an effective mechanism for exposing and mitigating cognitive bias,” theIntelligence Advanced Research Projects Agency announced Tuesday. Iarpa, the blue-sky science-and-technology division of the intelligence community, is looking to gather potential developers for this “Sirius” initiative next month in Washington.
“When an intelligence problem invokes [analysts'] biases, analysts may draw inferences or adopt beliefs that are logically unsound or not supported by evidence,” Iarpa notes. “Cognitive biases in analysis tend to increase with the level of uncertainty, lead to systematic errors, filter perceptions, shape assumptions and constrain alternatives.”
Does it ever. The litany of blown intelligence calls is nearly endless.
Even as Iraqi tanks moved closer to the Kuwaiti border in July 1990, few in the CIA believed Saddam would invade. In 2007, the intelligence community’s consensus view was that Iran had given up its nuclear weapons program. Enh, not quite.
Early last year, the top military intel official in Afghanistan lamented that he knew next to nothing about the people and cultures of Afghanistan. Do we even need to mention that whole Iraqi WMD thing?
Sirius is one of a whole bunch of Iarpa efforts to overcome biases — and reach more-accurate conclusions. Iarpa started work in December 2009 on a computer system that could replicate -– and then outdo -– human decision-making. The agency launched a project a few months later to let algorithms pick the most objective analysts.
With the Sirius project, Iarpa is hoping gamers will be able to study — and unlearn — all sorts of different prejudices. The agency is looking to axe everything from “anchoring bias” (relying too much on a single piece of evidence) to “confirmation bias” (only accepting facts that back up your pre-made case) to “fundamental attribution error” (attributing too much in an incident to personality, instead of circumstance).
Iarpa is hoping that “social scientists, computer scientists, statisticians, and gaming and virtual-world experts, as [well as] universities and companies from around the world will participate in this research.” Given the, um, uneven state of journalism these days, let’s hope they let a few reporters in on the fun, too.
Posts
