User:LBird BASc/sandbox/ATK/Seminar6/Truth/gsdddsbss

This Wikibook chapter explores how the truth about the Facebook–Cambridge Analytica data scandal is presented/portrayed in different disciplines and why these contrasting truths create an interdisciplinary issue: when the meta-knowledge integrated from various disciplines is influenced by knowledge contradictions between the disciplines themselves.

Introduction and Case Study: The Facebook-Cambridge Analytica Scandal

This scandal came out in 2018 when Cambridge Analytica was reported to have been harvesting data from Facebook profiles in order to sway the 2016 United States presidential campaign in favour of Republican candidate Ted Cruz. Upon further investigation, Cambridge Analytica was found to have influenced more than 200 elections worlwide, including Donald Trump's 2016 campaign, resulting in the scrutiny and revisiting of policies regarding data usage and protection. It was revealed that in total the data of up to 87 million Facebook profiles were exploited for mainly political reasons. How exactly did Cambridge Analytica accomplish this extraordinary feat?

The truth about the acquisition of the data

Context

In 2013, the data scientist Aleksandr Kogan, in the context of his academic work linking Facebook profiles to personality traits, developed a personality quiz app called "thisisyourdigitallife", which paid people to answer a 120 question personality test. However, in order to be paid for completing the survey, users had to log on to their Facebook profile, and allow the app to access to their Facebook activity: Kogan was trying to deduce personality traits based on Facebook activity, and to verify his hypotheses, he would use the personality test the people took to confirm whether his prediction was accurate. Later that year, Kogan was contacted by the firm SCL Group, a parent company of CA, with a business proposal: SCL would pay Kogan and provide him extra money, which he would use as an incentive to get more people to take the survey, and in exchange Kogan would provide the data gathered to SCL. The reason for SCL interest's was that the app in question not only gathered data on the individual who answered the survey, but on all of their friends's data as well, such as likes, posts, comments, location, and even private messages. An estimated number of 270 000 people took the survey, amounting up to the data of up to 87 million people and the extremely specific personality profiles of the people that took the survey being thus shared with SCL.

Law

Under 2013 Terms of Service, Facebook's API (Application Programming Interface) Version 1 allowed developers to access the data of an individual in question and their friends given that the user consented, and before signing up for Facebook, users accepted these terms and conditions: when in 2013 Kogan's app launched, its operations were completely legal. However, Facebook changed its API version in April 2014, limiting the data scope new developer apps could have, but giving pre-existing apps a one year enforcement delay to redesign their software to match the new API: Kogan had legally the right to keep harvesting data under the 2013 API for one more year. In 2015, thisismydigitallife ceased its operations as it no longer fit Facebook's terms and conditions, but was allowed to keep all the data it had harvested. The data of up to 87 million people had been obtained by Kogan under entire legality. Under the initial agreement between Kogan and Facebook with the regards to the application, Kogan explicitely states his possible intention to sell, to which Facebook knowingly agreed, although it conflicted with sections 3.3, 3.9, 3.10, and 7.4 of Facebook's Platform Policy. This means that the data gathered by Cambridge Analytica was legally obtained: Kogan, under the agreement signed by Facebook, was allowed to sell the information he had gathered, and that is what he did.

Ethics

From ethical perspective it was unrighteous to gather users' personal data without an actual consent. Personal information should be strictly secured, because easily accessible intimate information can be used for manipulating people, therefore, we have data protection policies or when personal data is exposed for third parties people should be aware and give consent for that. Indeed it is why Facebook was fined £500,000,000 by the Information Commissioner's Office in the UK: Facebook supposedly breached British Privacy Law section DPP1 (Data Privacy Principles), as the means of collection of data was done unfairly.

Conclusion

We clearly see that the data of 87 million people gathered by Cambridge Analytica was legally obtained and sold by Kogan. However, Facebook had a security hole which did not secure users' privacy.

The truth about the exploitation of the data

Context

CA's job as a political consulting firm, once they had acquired the data, was to provide information to their employers in regards to the best way to target individual Facebook profiles in order to swing the elections for their employer's side.

The OCEAN Model

Psychology

To do this, CA first had to match a personality profile to each of their 87 million users they had acquired the data on: they decided to use an OCEAN (The Big Five Personality Traits) model in this purpose. This model attempts to quantify the presence of 5 different traits in a person: their openness (willingness to explore, curiosity, flexibility), their conscientiousness (self-discipline, goal-orientation, planning skills), their extraversion (leadership, social confidence), their agreeableness (empathy, kindness, cooperative, generosity), and their neuroticism (irritability, insecurity, anxiety).

Using Machine Learning, they were able to match each person to their corresponding model, and use this information to manipulate people's opinions on a particular political party. For example, on the topic of creation of jobs, CA could draft a different Republican Party advertisement for two swing voters that scored differently on the OCEAN model: a person scoring high on the O part would receive an ad on how more jobs allows people to expand their knowledge and grow, whilst a person scoring high on the N part would receive a different ad detailing how more jobs allows people to receive security and emotional stability. This micro-targeting of voters is much more efficient as it appeals specifically to someone's personality, rather than a mass distribution of one ad that is less likely to appeal in large numbers.

Machine Learning/ Data Science

A simplified artificial neural network

To convert the raw data CA had into individual OCEAN profiles for all 87 million Facebook profiles, CA used a process called data mining, which attempts to predict information based on current one. CA began by developing a training data set (the initial Facebook and personality data obtained through the survey responses), on which they could then base the target variable data set (the Facebook data they wanted to make predictions on) on. Using artificial neural networks, the CA team linked the Facebook data info of the 270 000 people that took the personality test to their personality attributes through a supervised learning method that determined the mathematical connections between these two pieces of information. This method essentially works by continually adjusting the weight of the neural connections to optimize the prediction process. Once this training data set perfected, they applied the technique to the target variable set. To obtain a near perfect predictive process, CA used an ensemble model: they used 253 different types of training sets, which resulted in individual OCEAN personality predictions at an extremely high confidence level. This is how CA figured out the personalities and extremely private information of 87 million people.

Once the psychographic information discovered, CA took advantage of Facebook's extremely lenient Advertisement Policy, which allows political parties to target extremely specific voter groups through microtargeting. To fabricate their ads, CA used another algorithm which created different advertisements for each one of the 87 million people, appealing to their OCEAN traits. Then they sent out these individual ads through Facebook.

Ethics and Politics

The scandal is a major problem in the realm of politics since it was an invasion of privacy, which was used for political means. Democracy is a political system, which enforces the freedom of its citizens. Social media and data analytics firms are being used more frequently in electoral campaigns and targeted advertisements, which in many cases is starting to undermine democracy as a concept. For example, in the case of Cambridge Analytica scandal, it was carried out using Facebook's business model. In the Trump Campaign citizens’ data was used without authorisation and for personal gains, however Cambridge Analytica deleted all their data before the 2016 election which raises questions about the actual effects on the campaign. Nevertheless, from an idealist perspective of a normative truth the use of personal data was both unethical and manipulative. Furthermore, the issue extends beyond the Trump campaign as new investigations are being carried out regarding its effects on Brexit and EU campaign. Overall, this raises questions about the need of governments to pass laws for greater regulation of online political advertisements and data use.

Law

In legal terms, the exploitation of the data in political campaigns does not have any significant legal consequences. Data driven political campaigns have existed for a long time, however the extent of micro-targeting and profiling populations has become a lot more privacy invasive. The majority of countries lack the regulation and laws to protect such personal data as well as transparency on how it is gathered and used. From the perspective of the discipline of law and the empirical truth gathered from the American Law, the use of data in Cambridge Analytica scandal was not a legal issue. There is no single data protection legislation, but rather many small sector-specific laws on state and federal levels. However, in Europe steps have been taken through the General Data Protection Regulation from May 2018, showing how data exploitation is seen more of an issue with legal consequences.

The truth about the presentation of the scandal

The scandal is presented subjectively compared to what happened in 2012. Then there was a similar case when an Obama campaign took an advantage of Facebook users data. Around 1 million of supporters signed up for an Obama campaign's Facebook app, where users gave permission to access their and their list of friends personal data. The issue was identical, as programmers could get data of users' friends without their actual consent. Thus, the campaign gathered data of around 200 millions of users, most of them were not aware of their data being given to the campaign. . However, media perspective was very different on Obama and Trump campaigns. As Cambridge Analytica caused a huge outrage and Obama's breach did not cause any scandal it was even complimented. Also, Facebook was not fined for the security holes of giving personal information without actual consent of that person, thus, it did not solve the bug of exposing personal information from one's friends list. Therefore, it seems media is being biased on different campaigns even though the privacy of people was violated in both cases.