Lemi Baruh

Big data analytics and the limits of privacy self-management

This article looks at how the logic of big data analytics, which promotes an aura of unchallenged objectivity to the algorithmic analysis of quantitative data, preempts individuals’ ability to self-define and closes off any opportunity for those inferences to be challenged or resisted. We argue that the predominant privacy protection regimes based on the privacy self-management framework of “notice and choice” not only fail to protect individual privacy, but also underplay privacy as a collective good. To illustrate this claim, we discuss how two possible individual strategies—withdrawal from the market (avoidance) and complete reliance on market-provided privacy protections (assimilation)—may result in less privacy options available to the society at large. We conclude by discussing how acknowledging the collective dimension of privacy could provide more meaningful alternatives for privacy protection.

Age differences in privacy attitudes, literacy and privacy management on Facebook

Privacy has been identified as a hot button issue in literature on Social Network Sites (SNSs). While considerable research has been conducted with teenagers and young adults, scant attention has been paid to differences among adult age groups regarding privacy management behavior. With a multidimensional approach to privacy attitudes, we investigate Facebook use, privacy attitudes, online privacy literacy, disclosure, and privacy protective behavior on Facebook across three adult age groups (18-40, 41-65, and 65+). The sample consisted of an online convenience sample of 518 adult Facebook users. Comparisons suggested that although age groups were comparable in terms of general Internet use and online privacy literacy, younger groups were more likely to use SNSs more frequently, use Facebook for social interaction purposes, and have larger networks. Also, younger adults were more likely to self-disclose and engage in privacy protective behaviors on Facebook. In terms of privacy attitudes, older age groups were more likely to be concerned about privacy of other individuals. In general, all dimensions of privacy attitudes (i.e., belief that privacy is a right, being concerned about one’s privacy, belief that one’s privacy is contingent on others, being concerned about protecting privacy of others) were positively correlated with engagement in privacy protective behavior on Facebook. A mediation model demonstrated that amount of disclosure mediated the relationship between age groups and privacy protective behavior on Facebook. Finally, ANCOVA suggested that the impact of privacy attitudes on privacy protective behavior was stronger among mature adults. Also, unlike older age groups, among young adults, considering privacy as a right or being concerned about privacy of other individuals had no impact on privacy protective behavior.

Press-Party Parallelism and Polarization of News Media during an Election Campaign The Case of the 2011 Turkish Elections

The aim of this article is to examine press-party parallelism during the 2011 national elections in Turkey. The article reports findings from a content analysis of 9,127 news articles and editorial columns from fifteen newspapers regarding the trajectory of press-party parallelism over the course of the twelve-week national elections campaign period. We focus on two indicators of press-party parallelism: (1) respective “voice” given to the two leading parties, calculated as the ratio of news that quoted sources from the incumbent Adalet ve Kalkınma Partisi (AKP) to the leading opposition party Cumhuriyet Halk Partisi (CHP) and (2) news articles’ tones toward AKP and CHP. The newspapers that were content analyzed were first categorized into three groups based on survey data regarding the voting intentions of their readers: (1) a group of “conservative” newspapers whose readers intended to vote primarily for AKP, (2) a group of “mainstream broadsheets,” and (3) a group of “opposition” newspapers with a readership base intending to vote for CHP. The findings suggest that over the course of the election campaign, internal pluralism in both conservative and opposition papers declined in terms of voice given to respective parties and tone of news coverage.

Captive But Mobile: Privacy Concerns and Remedies for the Mobile Environment

We use the legal framework of captive audience to examine the Federal Trade Commission 2012 privacy guidelines as applied to mobile marketing. We define captive audiences as audiences without functional opt-out mechanisms to avoid situations of coercive communication. By analyzing the current mobile marketing ecosystem, we show that the Federal Trade Commissions privacy guidelines inspired by the Canadian “privacy by design” paradigm fall short of protecting consumers against invasive mobile marketing in at least three respects: (a) The guidelines overlook how, in the context of data monopolies, the combination of location and personal history data threatens autonomy of choice; (b) the guidelines focus exclusively on user control over data sharing, while ignoring control over communicative interaction; and (c) the reliance on market mechanisms to produce improved privacy policies may actually increase opt-out costs for consumers. We conclude by discussing two concrete proposals for improvement: a “home mode” for mobile privacy and target-specific privacy contract negotiation.

Communicating Turkish-Islamic identity in the aftermath of the Gaza flotilla raid: Who is the “us” in “us” versus “them”?

On May 31,2010, Israeli Defense Forces raided the ship Mavi Marmara, part of a six-vessel flotilla aiming to break the Israeli naval blockade of the Gaza Strip and to deliver supplies to Gaza. Using comments posted on Turkish online discussion forums in the aftermath of the raid that resulted in the death of nine passengers, this article analyzes how the incident was appropriated to negotiate between Turkishness and Islam as two alternative, yet coinciding forms of collective identity. Particularly, the article will compare different discursive strategies that were utilized in “general-interest” and “Islamic-leaning” online discussion groups. A deductive thematic analysis of 585 posts in general-interest and Islamic-leaning forums found significant differences in how metaphors of the body—blood, sacrifice, and martyrdom—as well as in-group/out-group comparisons were used in order to support a territorial-based nationalism versus a religion-based identity. The analysis also discusses the rhetoric that enabled discussants in general-interest forums to negotiate the tensions between the two collective identities.

Online text classification for real life tweet analysis

In this paper, we study multi-class classification of tweets, where we introduce highly efficient dimensionality reduction techniques suitable for online processing of high dimensional feature vectors generated from freely-worded text. As for the real life case study, we work on tweets in the Turkish language, however, our methods are generic and can be used for other languages as clearly explained in the paper. Since we work on a real life application and the tweets are freely worded, we introduce text correction, normalization and root finding algorithms. Although text processing and classification are highly important due to many applications such as emotion recognition, advertisement selection, etc., online classification and regression algorithms over text are limited due to need for high dimensional vectors to represent natural text inputs. We overcome such limitations by showing that randomized projections and piecewise linear models can be efficiently leveraged to significantly reduce the computational cost for feature vector extraction from the tweets. Hence, we can perform multi-class tweet classification and regression in real time. We demonstrate our results over tweets collected from a real life case study where the tweets are freely-worded, e.g., with emoticons, shortened words, special characters, etc., and are unstructured. We implement several well-known machine learning algorithms as well as novel regression methods and demonstrate that we can significantly reduce the computational complexity with insignificant change in the classification and regression performance.

When Sharing Less Means More: How Gender Moderates the Impact of Quantity of Information Shared in a Social Network Profile on Profile Viewers' Intentions About Socialization

This study summarizes the results from a 2 (low vs. high information) × 2 (female vs. male profile) experiment that investigates the impact of quantity of information shared on a Social Network Site (SNS) profile on viewers’ intentions to pursue further interactions with the profile owner. Quantity of information had no statistically significant effect on intentions to further socialize online. The two-way interaction between information quantity and profile gender was such that for male profiles more information increased profile viewers’ intentions to further socialize with the profile owner, whereas for female profiles the opposite was the case. The three-way interactions among quantity of information, profile gender, and profile viewers gender underline a tendency for male profile viewers to respond more positively to higher information shared by profiles from their own gender. For female viewers, this effect, although in the same direction, was smaller.

An Online Causal Inference Framework for Modeling and Designing Systems Involving User Preferences: A State-Space Approach

We provide a causal inference framework to model the effects of machine learning algorithms on user preferences. We then use this mathematical model to prove that the overall system can be tuned to alter those preferences in a desired manner. A user can be an online shopper or a social media user, exposed to digital interventions produced by machine learning algorithms. A user preference can be anything from inclination towards a product to a political party affiliation. Our framework uses a state-space model to represent user preferences as latent system parameters which can only be observed indirectly via online user actions such as a purchase activity or social media status updates, shares, blogs, or tweets. Based on these observations, machine learning algorithms produce digital interventions such as targeted advertisements or tweets. We model the effects of these interventions through a causal feedback loop, which alters the corresponding preferences of the user. We then introduce algorithms in order to estimate and later tune the user preferences to a particular desired form. We demonstrate the effectiveness of our algorithms through experiments in different scenarios.

Mathematical model of causal inference in Social Networks

In this article, we model the effects of machine learning algorithms on different Social Network users by using a causal inference framework, making estimation about the underlying system and design systems to control underlying latent unobservable system. In this case, the latent internal state of the system can be a wide range of interest of user. For example, it can be a users preferences for some certain products or affiliation of the user to some political parties. We represent these variables using state space model. In this model, the internal state of the system, e.g. the preferences or affiliations of the user is observed using users connections with the Social Networks such as Facebook status updates, shares, comments, blogs, tweets etc.

Online Anomaly Detection With Nested Trees

We introduce an online anomaly detection algorithm that processes data in a sequential manner. At each time, the algorithm makes a new observation, produces a decision, and then adaptively updates all its parameters to enhance its performance. The algorithm mainly works in an unsupervised manner since in most real-life applications labeling the data is costly. Even so, whenever there is a feedback, the algorithm uses it for better adaptation. The algorithm has two stages. In the first stage, it constructs a score function similar to a probability density function to model the underlying nominal distribution (if there is one) or to fit to the observed data. In the second state, this score function is used to evaluate the newly observed data to provide the final decision. The decision is given after the well-known thresholding. We construct the score using a highly versatile and completely adaptive nested decision tree. Nested soft decision trees are used to partition the observation space in a hierarchical manner. We adaptively optimize every component of the tree, i.e., decision regions and probabilistic models at each node as well as the overall structure, based on the sequential performance. This extensive in-time adaptation provides strong modeling capabilities; however, it may cause overfitting. To mitigate the overfitting issues, we first use the intermediate nodes of the tree to produce several subtrees, which constitute all the models from coarser to full extend, and then adaptively combine them. By using a real-life dataset, we show that our algorithm significantly outperforms the state of the art.