Dana Turjeman

Research

Publications

1. Turjeman, Dana and Fred M. Feinberg (2020), “Our Data Driven Future: Promise, Perils and Prognoses”. Available on SSRN: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3440726

Nowadays, most of our activities and personal details are recorded by one entity or another. These data are used for many applications that fundamentally enrich our lives, such as navigation systems, social networks, search engines, and health monitoring. On the darker side of data collection lie usages that can harm us and threaten our sense of privacy. Marketing, as an academic field and corporate practice, has benefited tremendously from this era of data abundance, but has concurrently heightened the risk of associated harms.

In this paper, we discuss both the great advantages and potential harms ushered in by this era of data collection, as well as ways to mitigate the harms while maintaining the benefits. Specifically, we propose and discuss classes of potential solutions: methods for collecting less data overall, transparency of code and models, federated learning, identity management tools, among others. Some of these solutions can be implemented now, others require a longer horizon, but all can begin through the advocacy of Marketing Research. We also discuss possible ways to improve on the benefits of data collection – by developing methods to assist individuals pursue their long-term goals while advocating for privacy in such pursuits.

Working Papers

2. Turjeman, Dana and Fred M. Feinberg, “When the Data Are Out: Measuring Behavioral Changes Following a Data Breach”.

Invited to revise and resubmit, Marketing Science. Available on SSRN: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3427254

As the quantity and value of data increase, so do the severity of data breaches and customer privacy invasions. While firms typically publicize their post-breach protective actions, little is known about the social, behavioral, and economic aftereffects of major breaches. Specifically, do individual customers alter their interactions with the firm, or do they continue with "business as usual"? We address this general issue via data stemming from a matchmaking website, one for those seeking an extramarital affair, that was breached. The data include de-identified profiles of paying male users from the United States, and their activities on the website since joining, and up to 3 weeks after, the disclosure of the data breach. A challenge in making causal inference(s) in the setting of a massive and highly publicized data breach is that all users were informed of the breach at the same time. In such cases of "information shock", there is no obvious control group. To resolve this problem, we propose Temporal Causal Inference: for each group of users who joined in a specific time period, we create an appropriate control group from all users who had joined prior to it. This procedure helps control for, among other elements, potential trends in both individual and temporal site usage that broadly fall under the rubric of "normal" usage trajectories. Following the construction of suitable control groups, we apply and extend several causal inference approaches. We adapt Causal Forests, among other forest-based methods) into Temporal Causal Forests, to better align 'temporal' inference settings. The combination of Temporal Causal Inference and Temporal Causal Forests methods allows us to extract insights regarding the homogeneous (average) treatment effect, along with nontrivial heterogeneity in responses to the data breach. Our analyses reveal that there is a decrease in the probability of being active in searching or messaging on the website, and a notable increase in the probability of deleting photos, ostensibly to avoid personal identification. We investigate several potential sources of heterogeneity in response to the breach announcement, and conclude with a discussion of both managerial consequences and policy considerations.

3. Fan, Ying, Yeşim A. Orhun and Dana Turjeman (alphabetical order), “Heterogeneous Actions, Beliefs, Constraints and Risk Tolerance During the COVID-19 Pandemic”.

NBER Working Paper No. 27211: https://www.nber.org/papers/w27211

During a pandemic, an individual's choices can determine outcomes not only for the individual but also for the entire community. Beliefs, constraints and preferences may shape behavior. This paper documents demographic differences in behaviors, beliefs, constraints and risk preferences across gender, income and political affiliation lines during the new coronavirus disease (COVID-19) pandemic. Our main analyses are based on data from an original nationally representative survey covering 5,500 adult respondents in the U.S. We find substantial gaps in behaviors and beliefs across gender, income and partisanship lines; in constraints across income levels; and in risk tolerance among men and women. Based on location data from a large sample of smartphones, we also document significant differences in mobility across demographics, which are consistent with our findings based on the survey data.

4. Dooley, Samuel, Dana Turjeman, John P. Dickerson, and Elissa M. Redmiles, “Field Evidence of the Effects of Pro-sociality and Transparency on COVID-19 App Attractiveness”. Available on SocArXiv: https://osf.io/preprints/socarxiv/gm6js/

COVID-19 exposure-notification apps have struggled to gain adoption. Existing literature posits as potential causes of this low adoption: privacy concerns, insufficient data transparency, and the type of appeal used to pitch the pro-social behavior of installing the app. In a field experiment, we advertised CovidDefense, Louisiana's COVID-19 exposure-notification app, at the time it was released. We find that all three hypothesized factors - privacy, data transparency, and appeals framing - relate to app adoption, even when controlling for age, gender, and community density. Specifically, we find that collective-good appeals are effective in fostering pro-social COVID-19 app behavior in the field. Our results empirically support existing policy guidance on the use of collective-good appeals and offer real-world evidence in the on-going debate on the efficacy of such appeals. Further, we offer nuanced findings regarding the efficacy of transparency - about both privacy and data collection - in encouraging health technology adoption and pro-social COVID-19 behavior. Our results may aid in fostering pro-social public-health-related behavior and for the broader debate regarding privacy and data transparency in digital healthcare.

Research in Progress

5. Tian, Longxiu and Dana Turjeman (alphabetical order), “Privacy Preserving Data Fusion: Leveraging Full Customer Records while Maintaining Anonymity” (dissertation version is available upon request)

Data fusion - the combination of multiple datasets - is a powerful technique to make inferences that are more accurate, generalizable, and useful than those made with any single dataset alone. However, when data fusion involves user-level data, the technique poses a privacy hazard due to the risk of revealing the identities of users. To preserve user anonymity while allowing for a robust and expressive data fusion process, we propose a privacy preserving data fusion (PPDF) methodology based on variational autoencoders (VAE), a nonparametric Bayesian generative modeling framework estimated in adherence to differential privacy (DP) - the state-of-the-art theory for privacy preservation. PPDF does not require the same users will appear in both datasets when making inferences on the joint data, and explicitly accounts for missingness in each dataset by leveraging additional variation in the other to correct for sample selection. Moreover, PPDF is model-agnostic: it allows for inferences to be made on the fused data, without the analyst specifying a model a priori. PPDF does so without the original datasets ever coming in contact on a single machine or model. We undertake a simulation to showcase the quality of our proposed methodology, and describe a planned fusion of a large customer dataset from a match making website with a detailed, anonymous survey.

6. Turjeman, Dana and A. Yeşim Orhun, “Information Preferences on Information Collection”

We explore preferences to learn about privacy risks before adopting an online product or service. When people click “I’ve read and agree to the privacy policy”, they nearly always have not. This phenomenon is commonly described as “the biggest lie on the internet”. Several reasons have been proposed in the literature: Information Overload (length and readability of the policies), Digital Resignation (the feeling that one cannot do anything about them), and the common belief that the policies can change any time, among other reasons. We show that another reason for not attending to information about privacy, even if it is available and accessible to the customer, is active information avoidance. In study 1 we show that 20% of participants chose not to know whether a social platform they use collected their personal data, even when the answer is merely “yes” or “no”. When asked why they choose not to know, 78% of those who avoided the information stated they did so for reasons such as “the answer terrifies me” or because “ignorance is a bliss”. These people’s decisions were driven mainly by a desire to avoid information to manage their anticipated emotional responses and are referred to as “active information avoidance”. The rest (22%) of the people who did not want to be informed claimed low interest as the reason (e.g., they are not sharing information on the social platform, or they do not care about privacy). These people’s decisions were driven by an evaluation of the value of information: if the information would not change the way they behaved, it was redundant. Therefore, in Study 2 we focus on teasing apart active information avoidance from a perceived low value of information (“I do not want to know about privacy even though, or because, I care”, vs. “I do not care about privacy”). Teasing apart these two reasons requires knowing individual preferences for privacy. To determine the actual value of information at the individual level, we first measure individual preferences towards privacy - this informs us with the individual part worth of each attribute, and from this – estimate the value of actively remaining ignorant about the attribute. We then contrast this with the willingness to pay for information on each attribute compared to other attributes. This allows us to measure the willingness to stay ignorant (i.e., avoid information) about privacy (relatively to other attributes). Understanding what drives people to actively seek or avoid information on privacy risks (such as privacy policies and data breach announcements) may improve the way these privacy risks are presented, so as to avoid mistrust, consumer harm, and false sense of privacy.

7. Elissa Redmiles, Turjeman, Dana, et al. (in multiple group projects) - “Descriptive Ethics for Coronavirus Contact Tracing Apps”. See more information about the project, output, and media mentions in Microsoft's webpage here and in Dr. Redmiles' webpage here.

The COVID19 pandemic spread across the world in late 2019 and early 2020. As the pandemic spread, technologists joined forces with public health officials to develop apps to support COVID19 response, including contact tracing apps.

For these technological solutions to benefit public health, users must be willing to adopt these apps. In free-choice democratic societies, users have the choice whether or not to install these apps.

This project is a descriptive ethics based examination of people’s preferences regarding COVID apps in order to better align COVID-related technology with human values. The findings from this work are designed to aid the development of policy and technology to address COVID19.

Research in (Less) Progress

8. Turjeman, Dana, A. Yeşim Orhun and Dan Ariely, “Useful Sharing: The Role of Accountability in Financial Decisions” - one study conducted, more to come when possible.

Dissertation

Preface and introduction appear below. Want the full dissertation? Just email me: turji@umich.edu.

Turjeman, Dana (2021) - Introduction of Dissertation.pdf