Privacy

The concept of contextual integrity is a philosophy that dictates how information handling and privacy interact. The Barocas-Nissenbaum paper, Big Data’s End Run around Procedural Privacy Protections, describes it as saying information privacy is determined by standards of the given situation. This is to say that the rules are not always the same. Based on specific scenarios privacy will increase or decrease. In the paper, two different medical scenarios provide a look at two situations in which a lot of privacy is necessary. But still, certain “key actors” need access to said information, in each situation. To put it simply, different situations control the manner in which information is passed. The rules that dictate how information should be passed in certain situations are called informational norms. For instance, for a student, there are informational norms that dictate the way grades are communicated. I would not want everyone to be able to access my grades, and I assume most students would also want to same privacy. That being said, both the teacher of the class and I need access to the grades, we are the key actors. Contextual integrity goes even further than that though. Before going to college, I, like many other students signed the FERPA waiver. This gave me the control of the access to my grades, my transcripts, and my disciplinary records. Under the contextual integrity framework, a new question would appear. If the FERPA waiver did not exist, would it deter students from applying to college. But in a completely different situation, such as if I were to set a high score in a video game, I would want everyone to see that. Meaning the access to this information should be less restricted. These are just two examples to show that informational norms change based on the specific situation.

Looking at the Wu & Zhang paper through the lens of contextual integrity, more questions arise. If a picture of a person’s face could be used to accurately determine their criminality, how would people react? A fair amount of people would probably want access restricted to photos of them online. In fact, many people would probably guard pictures of themselves like treasures. This is because the potential amount of information that could be gathered from just an image would be a terrifying prospect. However, the findings from the paper are monumentally flawed. The cardinal error in their findings was that the algorithm does not analyze a person’s facial structure but instead, it analyzes whether they are smiling. In addition, they fail to understand that the origin of the photo would have an effect on the algorithm. Getting all the criminals’ pictures from one place and all the innocent people’s pictures from a different place would teach the algorithm to recognize the differences between the types of photos.

The results of the Kosinski et al. paper, Private Traits and Attributes Are Predictable From Digital Records of Human Behavior, through the same lens produce similar results of mild hysteria in the general public. But only temporarily, only until realize that the alternative is a life without access to any information. The public now knows that Facebook has these capabilities, the ability to deduce who you are as a person based on your online activity, and not much has happened. There was a brief moment of mass paranoia, followed by a modest exodus from Facebook. But after some conciliatory remarks and actions made by Facebook, including the addition of some rather weak privacy settings, people just forgot. I think there is a far more intriguing question to ponder. If Facebook is able to make an accurate profile on us based on our likes and then sell that information to advertisers, should we be able to cut out the middleman and sell our information directly to advertisers?

In the paper What if Everything Reveals Everything?, the authors, Ohm & Peppet, harp on about the concept of everything revealing everything, which is essentially a roundabout way of saying that in a technological age anything can be discovered about a person through machine learning. Through the same lens as before, this concept entails that privacy is increasingly difficult to maintain because everything is now sensitive data. How would we even begin to protect our information if all of it needs to be kept under lock and key? It would require an infinite expansion of the set of informational norms we currently have. Lawmakers will need to quickly start expanding what is considered sensitive information so that privacy can be maintained. The key to this happening is through the use of inference. To put it plainly, people need to start using common sense when it comes to the protection of data. The golden rule needs to enter the digital age. Privacy as a whole would be much less complicated if people just thought about whether they would want their information used the way they are using other people’s information.

Contextual integrity as a framework allows for more situational flexibility than differential privacy does. With contextual integrity, each situation dictates the privacy etiquette, whereas differential privacy requires anonymity always. The two frameworks overlap, but differential privacy demands far more strict adherence to secrecy. The easier and more useful of the two to implement is differential privacy because it is a much more concrete concept.

Related posts:

When i joined LinkedIn,i gradually got to see many law students’ profile then i realised that most of the law students have done many internships, from NGO to Khaitan & Co. and even they have many…

The growing popularity of streamers who have taken advantage of YouTube proves that making money online is entirely achievable. We can thus believe that everyone can easily generate income online…