Late last year, a Palestinian man was arrested by Israeli police, who found out that he had posted an image of himself on Facebook next to a tractor, a work vehicle weaponized by Palestinian lone wolf attackers to lethal affect in recent years, and captioned it with an a threat of a slaughter, presumably of Jews. What perfect pre-crime profiling.
Profiling is something we do all the time. As soon as we see a person (or an animal), we immediately, automatically, subconsciously make a profile in our head – are they a friend or a foe, will they try to kill us, and are they stronger or weaker than us? This is an evolutionary mechanism designed to protect ourselves from possible threats. The science of predictive profiling seeks to analyze data on criminals, identify recurring patterns, then apply them to people who haven’t transgressed yet and obtain threat assessments on them, predicting which of them are most likely to commit such crimes. This information can be utilized by security forces, who use the pre-crime assessments to investigate, detain, arrest and/or incarcerate them, thus foiling the crimes rather than dealing with their post factum implications.
Since social media services are rife with data about people, researchers and security forces scour them and glean information for their predictive profiling, and act upon that data. This method is problematic: scientists make mistakes, which might lead to false positives and false negatives; profiling algorithms are not objective, but rather embody their creators’ biases; spying on innocent people harms their right to privacy; holding people for crimes they might commit hurts their right to freedom and their presumption of innocence; being unjustifiably handled by security forces may turn non-criminals into criminals.
Haaretz reported a year ago that in wake of the 2015 so called “lone-wolf intifada,” Israel arrested hundreds of Palestinians, partly by analyzing their social media posts. This seems like a natural development for the army, police and Shabak, as it’s considered borderline impossible to gather intel on lone-wolf attackers. On the other hand, it’s safe to presume some of those arrests, if not all of them, raised one or more of the aforementioned problems.
The Palestinian who had arrested because of his tractor post was released a few hours later. While interrogating him, police learned that he had written “تصبحهم” (tesbechhum), Arabic for “good morning to you”. Alas, Facebook’s machine translation misplaced a tiny point diacritic and erroneously read the word as “تضبحهم” (yidbachhum), which means “slaughter them”. Don’t blindly trust technology.
In this episode of the Lex Cybernetica, the Hebrew University of Jerusalem Cyber Security Research Center’s podcast, we will talk about predictive profiling, pre-crimes and how they balance with human rights, with our guests Adv. Lila Margalit, who talks about how profiling clashes with human rights, Dr. Simon Perry, who’s co-writing a study about predicting terrorism through social networks, and Adv. Eli Bachar, who talks about how online profiling is about ideas rather than race or color, and Lex Cybernetica’s host, Ido Kenan.
Our podcast is now on Stitcher and iTunes! You can listen to it via Android or IOS as well: