The Guantánamo Files: Judging Detainees’ Risk, Often With Flawed Evidence

 Relations between guards and detainees at Guantánamo range from cooperative to dangerous.

WASHINGTON — Said Mohammed Alam Shah, a 24-year-old Afghan who had lost a leg as a teenager, told interrogators at the prison at Guantánamo Bay, Cuba, that he had been conscripted by the Taliban as a driver before being detained in 2001. He had been caught, he said, as he tried to “rescue his younger brother from the Taliban.”

These articles are based on a huge trove of secret documents leaked last year to the anti-secrecy organization WikiLeaks and made available to The New York Times by another source on the condition of anonymity.

Editors and reporters who worked on these articles will be answering select questions about the coverage of the material.

In order to view this feature, you must download the latest version of flash player here.

Military analysts believed him. Mr. Shah, who had been outfitted with a prosthetic leg by prison doctors, was “cooperative” and “has not expressed thoughts of violence or made threats toward the U.S. or its allies,” according to a sympathetic 2003 assessment. Its conclusion: “Detainee does not pose a future threat to the U.S. or U.S. interests.”

So in 2004 Mr. Shah was sent back to Afghanistan — where he promptly revealed himself to be Abdullah Mehsud, a Pakistan-born militant, and began plotting mayhem. He recorded jihadist videos, organized a Taliban force to fight American troops, planned an attack on Pakistan’s interior minister that killed 31 people, oversaw the kidnapping of two Chinese engineers, and finally detonated a suicide bomb in 2007 as the Pakistani Army closed in. His martyrdom was hailed in an audio message by none other than Osama bin Laden.

The Guantánamo analysts’ complete misreading of Abdullah Mehsud was included among hundreds of classified assessments of detainees at the prison in Cuba that were obtained by The New York Times. The unredacted assessments give the fullest public picture to date of the prisoners held at Guantánamo over the past nine years. They show that the United States has imprisoned hundreds of men for years without trial based on a difficult and strikingly subjective evaluation of who they were, what they had done in the past and what they might do in the future. The 704 assessment documents use the word “possibly” 387 times, “unknown” 188 times and “deceptive” 85 times.

Viewed with judges’ rulings on legal challenges by detainees, the documents reveal that the analysts sometimes ignored serious flaws in the evidence — for example, that the information came from other detainees whose mental illness made them unreliable. Some assessments quote witnesses who say they saw a detainee at a camp run by Al Qaeda but omit the witnesses’ record of falsehood or misidentification. They include detainees’ admissions without acknowledging other government documents that show the statements were later withdrawn, often attributed to abusive treatment or torture.

A Growing Wariness

Written between 2002 and 2009, the assessments reflect a growing wariness on the part of Guantánamo analysts. Early on, the reports are just a page or two and often sanguine in tone. By 2008, after scorching publicity about released detainees who joined Al Qaeda and the dwindling of the prison population to hard-core detainees, the assessments are decidedly more cautious.

For every case of an Abdullah Mehsud — someone wrongly judged a minimal threat — there are several instances in which prisoners rated “high risk” were released and have not engaged in wrongdoing. Murat Kurnaz, a German resident of Turkish ancestry, was judged in a 2006 assessment to be a member of Al Qaeda who fell into the most dangerous category: “high risk” and “likely to pose a threat to the U.S., its interests and allies.”

Nonetheless, American authorities, under pressure from both Germany and Turkey, overruled the analysts and sent Mr. Kurnaz home to Germany three months later. He did not join the global jihad but instead became a prominent critic of Guantánamo, writing a book and making countless media appearances to denounce the American prison.

Among the most revealing of the leaked documents is a 17-page guide for analysts, evidently prepared by military intelligence trainers, on how to gauge the danger posed by a detainee. It lists major clusters of detainees, including the so-called Dirty 30, who were the bodyguards of Mr. bin Laden, as well as the large group of accused Qaeda operatives captured with Abu Zubaydah, an important terrorist facilitator, at two guesthouses in Faisalabad, Pakistan, in 2002. It lists nine mosques associated with Al Qaeda, in Quebec, Milan, London, Yemen and Pakistan.

The guide shows how analysts seized upon the tiniest details as a potential litmus test for risk. If a prisoner had a Casio F91W watch, it might be an indication he had attended a Qaeda bomb-making course where such watches were handed out — though that model is sold around the world to this day. (Likewise, the assessment of a Yemeni prisoner suggests a dire use for his pocket calculator: “Calculators may be used for indirect fire calculations such as those required for artillery fire.”)

A prisoner caught without travel documents? It might mean he had been trained to discard them to make identification harder, the guide explains. A detainee who claimed to be a simple farmer or a cook, or in the honey business or searching for a wife? Those were common Taliban and Qaeda cover stories, the analysts were told.

And a classic Catch-22: “Refusal to cooperate,” the guide says, is a Qaeda resistance technique.

Yet the guide appears to be the product of years of experience at trying to turn bits of evidence of varying reliability into a conclusion. Notably, it cites as a cautionary tale the early misjudgment about Abdullah Mehsud, the Pakistani suicide bomber, who had claimed he was forced to join the Taliban. He was “an example,” the guide says, “of a detainee who successfully applied the conscription cover story as a means to secure his release from U.S. custody.”

Scott Shane reported from Washington, and Benjamin Weiser from New York. Reporting was contributed by Charlie Savage from Washington, and William Glaberson, Andrew W. Lehren and Andrei Scheinkman from New York.

 

Critic’s Notebook: Busy Job of Judging Video-Game Content to Be Ceded to Machines

That was then. This is now. Starting on Monday the ratings board plans to begin introducing computers to the job of deciding whether a game is appropriate for Everyone, for Teens or for Mature gamers (meaning older than 16). To do this the organization has written a program designed to replicate the ingrained cultural norms and predilections of the everyday American consumer, at least when it comes to what is appropriate for children and what isn’t.

Faced with an explosion in the number of games being released online, the board plans to announce on Monday that the main evaluation of hundreds of games each year will be based not on direct human judgment but instead on a detailed digital questionnaire meant to gauge every subtle nuance of violence, sexuality, profanity, drug use, gambling and bodily function that could possibly offend anyone.

The questionnaire, to be filled out by a game’s makers (with penalties for nondisclosure), is like a psychological inquest into the depths of all the things our culture considers potentially unwholesome.

Offensive language, for instance, is broken down into six subcategories: minor profanities, epithets, scatological vulgarities, racial obscenities, sexual vulgarisms and a final category devoted to one particular three-letter word that refers to both a beast of burden and, colloquially, to a part of human anatomy. (Interestingly, the survey does not ask about religious slurs, perhaps because those are relatively rare in popular Western culture.)

The sexuality category is fairly straightforward — basically, if there’s sex, how much of it can you actually see? — as are the sections on gambling and drugs. And surely every consumer is grateful that the portion on bodily functions makes a point to ask separately about flatulence sounds, “whimsical depictions of feces,” realistically depicted feces and the “act of human (or humanlike character) defecation visually depicted.”

This is the sort of stuff that major international corporations generally like to know about before they offer a mass-market media product for sale to the world.

As the Supreme Court prepares to release its potentially groundbreaking decision on a California law that intends to regulate the sale of video games (an announcement is expected any week now, and potentially as early as Monday), it may bear remembering that Justice Potter Stewart found it impossible to reduce obscenity to a definition, declaring, “I know it when I see it.”

And yet in this digital age it is inevitable perhaps that a group that is paid to sort creative entertainment endeavors into neat and tidy categories based on content would eventually start outsourcing its mission to computers. After all, major companies, including banks, credit-rating agencies, Amazon.com, Netflix and Google, have made it their business to reduce consumer behavior to an algorithm. For video games, the publishers’ answers to the questionnaire will determine the rating.

The rating board was founded in 1994 by the video-game industry’s main trade organization in an effort to mimic Hollywood’s popular system for rating big-studio films. Major retailers require a video game to have a board rating before they will stock it, just as major theater chains refuse to screen films that have not been rated by the Motion Picture Association of America. Perhaps even more important, the big global game console makers — Microsoft, Nintendo and Sony — won’t allow a game on their systems that has not been rated.

And yet the big growth in video games these days is on systems in which the games are not rated by age category: mobile phones and social networks like Facebook. Instead, those systems are under the exclusive control of the companies that run them, and Apple and Facebook have not developed detailed age-rating schemes of their own.

So the game-rating board is trying to come up with a system that allows it to quickly, reliably and consistently rate the thousands of games that are appearing each year on digital devices. In an interview, the board’s president, Patricia Vance, refused to comment on any negotiations with Apple, Facebook or other companies about adopting the ratings.

Last year the board rated about 1,600 games, of which up to 30 percent were only online. A few years ago there were almost no online games to review. By contrast, the motion-picture board considers around 850 movies a year. (And most video games are far longer than the average film.)

The human game classifiers who will soon be ceding work to machines don’t actually play the games. Instead they watch a DVD submitted by a game’s publisher that is supposed to include the raunchiest and most violent scenes, if any. So they have not been experiencing games as they are played.

This week the board intends to start using the new computerized system for games released online on systems controlled by Microsoft (Xbox Live), Sony (PlayStation Network) and Nintendo (Wii and DSi Shop). That’s about 650 games a year. The board says that publishers’ answers to the digital questionnaire will determine a game’s rating and that a human won’t review it until after the game is out the door.

As stated in a draft of the board’s news release, “All games rated via this new process will be tested by E.S.R.B. staff shortly after they are made publicly available to verify that disclosure was complete and accurate.”

What could possibly go wrong?

Major retail games — the ones advertised on television and stocked by Wal-Mart — will still be evaluated by real people. For now.