{"id":38338,"date":"2017-03-28T12:10:54","date_gmt":"2017-03-28T19:10:54","guid":{"rendered":"http:\/\/world.edu\/?p=38338"},"modified":"2017-03-28T12:10:54","modified_gmt":"2017-03-28T19:10:54","slug":"suing-federal-government-free-research","status":"publish","type":"post","link":"https:\/\/world.edu\/suing-federal-government-free-research\/","title":{"rendered":"We\u2019re suing the federal government to be free to do our research"},"content":{"rendered":"

\"ethics<\/a><\/p>\n

Many apps and algorithms that feature prominently in our lives are, essentially, black boxes: We have no idea how they accomplish what they do; we just know they work. Or at least we think we do. Most recently this became apparent when The New York Times revealed that Uber used a system<\/a> it called \u201cgreyballing\u201d to show certain users phantom cars and prevent them from getting rides through the Uber app.<\/p>\n

In essence, these users \u2013 often government officials or others Uber feared might interfere with the company\u2019s services \u2013 were shown a completely different (and deliberately false) view of Uber\u2019s data, without their knowledge or consent. Given the ubiquity of the Uber service \u2013 and the controversy that surrounds Uber<\/a> in general \u2013 this article immediately raised questions about the fairness and legality of Uber\u2019s practices. For example, why should people trust Uber\u2019s surge prices if the app and data can be manipulated arbitrarily, at will by Uber?<\/p>\n

Unfortunately, platforms like Uber are almost always closed: Users, regulators and policymakers typically have no way to know whether, or to what extent, apps and algorithms are behaving in questionable ways. The greyballing was only discovered because its existence was leaked to the media by current and former Uber employees.<\/p>\n

But algorithms are widely used in many contexts, including selecting job applicants when hiring<\/a>, estimating borrowers\u2019 creditworthiness<\/a>, determining where police departments should deploy patrol officers<\/a> and even setting bail for criminal suspects<\/a>.<\/p>\n

In all of these cases, the output of the algorithm can have enormous consequences on people\u2019s lives, and yet we often lack a basic understanding of whether they may be biased, unfair or discriminatory. It\u2019s a concern shared by<\/a> government<\/a> agencies<\/a> and scholars<\/a> alike.<\/p>\n

Running up against the law<\/strong><\/p>\n

\"linkedinTS\"<\/a>
\nOur own research seeks to identify these problems through study and scholarship, without needing to wait for whistle-blowers to spill the beans. We leverage real users and fake accounts created by us to compile data from online services<\/a>. Using these data, we try to tease out how black-box algorithms work: What data do they use? How do user characteristics affect the output of the algorithm<\/a>? Does the system do things that a normal person might find questionable<\/a>? Of course, companies view their algorithms and data sets as proprietary, and are loath to open them up to outside scrutiny, especially when it may reveal embarrassing things<\/a> (like attempts to manipulate law enforcement).<\/p>\n

In effect, our goal is to increase the transparency<\/em> of black-box algorithms (like those used by Uber) that affect our daily lives, and to make algorithms and their human designers accountable<\/em> to social and legal norms. We\u2019re developing ways for researchers, regulators and policymakers to measure these systems and identify instances of unfairness and discrimination.<\/p>\n

But there is a significant legal barrier<\/a>, one that we think effectively ensures the American public is, and will stay, in the dark about how these computerized systems work, and whether they\u2019re fair and equitable to everyone. It\u2019s called the Computer Fraud and Abuse Act<\/a> (CFAA). It\u2019s the country\u2019s main \u201canti-hacking\u201d law, originally passed in 1986 and broadened in 1996. And we\u2019re among a group of scholars and news organizations who have sued to overturn key provisions that block researchers like us from investigating these crucial elements of modern American life.<\/p>\n

Letting companies make the rules<\/strong><\/p>\n

Among other things, the CFAA imposes civil and criminal penalties on anyone who \u201cexceeds authorized access\u201d to any computer. This might seem relatively benign and vague, but the vagueness is part of the problem. Some courts and prosecutors have taken the position that it \u201cexceeds authorized access\u201d to do anything contrary to a website\u2019s or company\u2019s Terms of Service. These are the long screens of legalese users must agree to \u2013 usually without having read<\/a> a word of them<\/a> \u2013 before using a website or a piece of software.<\/p>\n

\"<\/a>

Facebook\u2019s Terms of Service include provisions raising privacy concerns. Screenshot of Facebook.com<\/p><\/div>
\nUnfortunately, many Terms of Service contain what we consider egregious claims and limitations on users.<\/p>\n