Legal and Technical Issues of Privacy Preservation in Data Mining

Abstract

To paraphrase Winograd (1992), we bring to our communities a tacit comprehension of right and wrong that makes social responsibility an intrinsic part of our culture. Our ethics are the moral principles we use to assert social responsibility and to perpetuate safe and just societies. Moreover, the introduction of new technologies can have a profound effect on our ethical principles. The emergence of very large databases, and the associated automated data analysis tools, present yet another set of ethical challenges to consider. Socio-ethical issues have been identified as pertinent to data mining and there is a growing concern regarding the (ab)use of sensitive information (Clarke, 1999; Clifton et al., 2002; Clifton and Estivill-Castro, 2002; Gehrke, 2002). Estivill-Castro et al., discuss surveys regarding public opinion on personal privacy that show a raised level of concern about the use of private information (Estivill-Castro et al., 1999). There is some justification for this concern; a 2001 survey in InfoWeek found that over 20% of companies store customer data with information about medical profile and/or customer demographics with salary and credit information, and over 15% store information about customers’ legal histories.

Targeting vulnerable populations: The ethical implications of data mining, automated prediction, and focused marketing

“…Further, Google uses data mining to develop thorough profiles of the online users of its services to determine buying preferences and habits and other personal details. In a similar fashion, Facebook records every click and "like" by users of its site, captures browsing histories of users, and buys externally collected information concerning its users, all of which allow for the establishment of more holistic user profiles (Tufekci, 2018). Computer applications can even listen in to ascertain what people are watching on television (Maheshwari, 2017), and both Google and Facebook can track individuals based on online visits to sex-oriented websites (Warzel, 2019).…”

“…Once personal information is collected, individuals throughout much of the world have virtually no control over the buying and selling of that information. When personal data are sold to other parties, the person who essentially generated that information has no jurisdiction over how it is used (Hess, 2017;Tufekci, 2018). Both Google and Facebook run "instant digital auctions" for every advertisement placement whereby they consider competitive bids from multiple potential companies that advertise and sell their product or service to a defined group of online users of the Google and Facebook systems (Metz, 2017).…”

“…Indeed, through the use of facial recognition systems, databases containing millions of images of people's faces have been created that can allow for analysis of moods and other characteristics (Metz, 2019). Further, because substantial percentages of the global population are connected and wired into internet-based searches and applications, it is virtually impossible to "opt out" of the databases that contain personal knowledge and related predictive capacity (Tufekci, 2019). Exacerbating the potential abuse of automated prediction by marketers is the stark realization that up to 40% of the population has some form of "internet-based" addiction (Alter, 2017).…”

“…For example, as Tufekci (2019, p. A23) states, leaked documents from Facebook indicated that the company told advertisers that it could "predict when younger users, including teenagers, were feeling 'insecure,' 'worthless' or otherwise in need of a 'confidence boost.' " Facebook's ability to draw these subtle inferences apparently resulted from monitoring photos, posts, and other data generated by social media accounts (Tufekci, 2019). Whether Facebook or other technology companies ultimately permit advertisers to target an individual who might be in a psychologically vulnerable state of mind, the disconcerting fact is that the capacity exists.…”

“…Additionally, computational inference as a machine-based analytical tool is not infallible (Tufekci, 2015), where mined data can create "overand inaccurate personalization" (Pan, 2016, p. 250). Not surprisingly, personalization can lead to errors in judgments about individuals that can result in such outcomes as discriminatory targeting (or exclusion) or the misidentification of a person into at-risk or otherwise vulnerable group (Tufekci, 2019).…”