Search

Consumers regularly encounter privacy policies explaining if and how their personal information will be collected, stored, used, and shared. Consumer advocacy groups and regulators who argue for the provision of such assurances typically assume that consumers prefer transparency in data practices. Indeed, past research supports the notion that most consumers favor restrictions on the gathering and use of personal information (Nowak and Phelps 1992) and that when such protections are not communicated via a strong privacy assurance, consumers may respond defensively (Lwin, Wirtz, and Williams 2007). In fact, a recent study reports that 9 in 10 Americans avoid companies they do not believe protect their privacy (National CyberSecurity Alliance, 2016). In an effort to comply with these demands and build consumer trust, organizations regularly provide assurances to consumers about data management practices in the form of privacy policies. In this paper, we document an ironic effect of privacy policies on consumer behavior: although privacy policies are often intended to assure consumers that a firm's use of their personal data will be legitimate, they can instead raise the salience of privacy concerns and paradoxically decrease rather than increase perceived security. Specifically, we propose that due to these decreased perceptions of security, consumers often react to these assurances as if they were warnings, leading to lower purchase intent in the presence (vs. absence) of a privacy policy. In a series of experiments, we examine how the presence (vs. absence) of a privacy policy affects consumers' perceptions of the security of their personal data and their interest in purchasing a product. Study 1 aimed to show that despite the opposite outcomes they are designed to produce, both assurances and warnings can decrease consumer interest in a product. We predicted that instead of promoting a sense of confidence that personal data will not be misused, a privacy policy would ironically decrease perceived security and purchase intent (relative to a condition in which no privacy policy is provided). Data for this study were collected from participants recruited on Amazon Mechanical Turk (AMT) who were randomly assigned to one of six conditions in a 2 (Privacy Policy: Absent vs. Present) × 3 (Prime: Warning, Neutral, No Prime) between-participants design. As a test of the effect of a warning on perceived security and purchase intent, participants in the Warning Prime condition read an article warning them of the potential abuses of geolocation data. As a control condition, participants in the Neutral Prime condition read an article discussing innovative technology in cars. As an alternate version of the control condition, participants in the No Prime condition did not read an article. Following the prime manipulation, all participants were asked to imagine that they were considering a new technology system for their car and shown a description of a new product (called ViewTech), followed by detailed descriptions of various product features. As a test of the effect of a privacy policy on perceived security and purchase intent, participants in the "Privacy Policy Present" condition viewed a privacy policy immediately before the feature descriptions, whereas participants in the "Privacy Policy Absent" condition did not view a privacy policy at all. The privacy policy was designed to make explicit information that was already more subtly conveyed through the description of product features. For example, the privacy policy stated that the exact location of car and mobile devices will be collected and stored. This information may be reasonably and easily inferred from the product features, which include the ability to pinpoint your car's exact location and retrace routes, to stream real-time route data to any mobile device for a live-updating map, and to receive directions to your parked car on your mobile device. Next, purchase intent was measured by asking participants to indicate their level of interest in purchasing the product using a sliding scale (0=Not at all interested; 100=Very Interested). Perceived security was measured by asking participants how comfortable they would be with their data being collected and stored by ViewTech; participants indicated their response using a sliding scale (0=Not at all comfortable; 100=Extremely comfortable). Not surprisingly, the results indicated that participants felt less secure in the warning prime condition relative to both the neutral and no prime conditions. More central to our theorizing, the privacy policy seemed to have a similar effect on perceived security as the warning prime. Specifically, there was a main effect of the privacy policy manipulation, such that those who saw a privacy policy felt less secure than participants who did not see a privacy policy. Average interest in purchasing the product showed an identical pattern in that there was a main effect of both the prime and the privacy policy and no interaction. Study 2 aimed to show that this effect influences downstream consequences, such that the presence versus absence of privacy policies can change consumers' personal willingness to pay for a product (without affecting their perception of the product's value in the marketplace). We also aimed to test whether the effect is robust to the timing of exposure to the assurance and to provide some more direct evidence of the process by analyzing the extent to which privacy concerns influenced consumers' decisions. Data for this study were collected from participants recruited on AMT who were randomly assigned to one of four conditions in a 2 (Privacy Policy Before Descriptions: Yes vs. No) in a 2 (Privacy Policy After Descriptions: Yes vs. No) between-participants design. As in study 1, all participants were asked to imagine that they were considering a new technology system for their car and shown a description of a new product, followed by detailed descriptions of various product features. The product description and feature descriptions were identical to those used in study 1, as was the privacy policy. Depending on their condition, participants viewed the privacy policy immediately before the feature descriptions, immediately after the feature descriptions, both before and after the feature descriptions, or not at all. Participants then indicated their purchase intent using the same scale as in study 1 and were asked to briefly explain the rationale for their answer. Next, participants estimated the retail price of the package (on a sliding scale from $1 to $10,000) and indicated their own willingness to pay in a text box. Perceived security was measured using the same scale as in study 1. Since no significant difference between the before vs. after conditions was observed on any of the dependent measures, they were collapsed into a single condition in the analysis. Consistent with our predictions, participants who did not see a privacy policy felt more secure than participants who saw a privacy policy. Average interest in purchasing the product showed an identical pattern and was higher among participants who did not see a privacy policy than among those who did. Moreover, participants who did not see a privacy policy were willing to pay more for the product than participants who saw a privacy policy. Perceived security mediated the effect of the privacy policy on consumers' interest in purchasing the product. Importantly, our manipulation did not affect expected retail prices; these did not differ regardless of whether the privacy policy was absent or present. Viewed in light of the significant difference on the other measures, this null result suggests that the presence or absence of a privacy policy selectively influences the extent to which an individual consumer personally desires a product, but not the consumer's perception of the objective value of the product in the marketplace. Study 3 aimed to provide further insight into the mechanism underlying this effect by showing that in joint evaluation, where privacy concerns are made salient by exposing all participants to a company with a privacy policy, the effect observed in earlier studies and in separate evaluation would be attenuated. Data for this study were collected from participants recruited on AMT who were randomly assigned to one of three conditions. In the first condition, participants were asked to imagine that they needed to buy some new clothes and had found some items they liked on the website of a retailer with whom they were previously unfamiliar. Participants were then asked to rate on two different sliding scales how interested they would be in making a purchase from this retailer and how comfortable they would be with the way their data are collected and stored by this retailer (0 = Not at all; 100 = Extremely). The second condition was identical except that in addition to the shopping scenario previously described, participants were told that before proceeding to check-out, they noticed a privacy policy, and they were shown the privacy policy on the following screen. Participants in the third condition were told to imagine that they had found items they liked on the websites of two online retailers that they were not previously familiar with, and that before proceeding to check-out, they had noticed a privacy policy on one site, but could not find a privacy policy on the other site. They were then shown the same privacy policy as participants in the separate evaluation condition. After viewing the privacy policy, participants completed the same set of dependent measures as in the separate evaluation conditions, except that they completed one set while thinking about the retailer with the privacy policy and the second set while thinking about the retailer without the privacy policy. Consistent with our predictions, the effect observed in our previous studies was replicated in separate evaluation, but attenuated in joint evaluation. Specifically, interest in purchasing from a retailer with a privacy policy was lower in separate evaluation than in joint evaluation. However, interest in purchasing from a retailer without a privacy policy was higher in separate evaluation than in joint evaluation. This boundary condition to the effect observed in earlier studies illustrates how organizations might present a privacy policy without losing consumer trust. From a theoretical standpoint, our research contributes to the literature on persuasion by highlighting parallels between the unintended ways in which consumers react to warnings and assurances and showing that although most consumers claim to want transparency in privacy practices, they may react negatively to it. By documenting how consumers respond to privacy assurances, we complement prior research that has focused on how consumers respond to warnings (Stewart and Martin 1994), including advertisement disclaimers (Barlow and Wogalter 1993; Herbst et al. 2011; Johar and Simmons 2000), disclosures of conflict of interest (Cain, Loewenstein, and Moore 2010), and product warning labels (Argo and Main 2004). Our findings have important implications for how companies' transparency in data practices can affect consumer trust. Specifically, they suggest that the observed preference to transact with organizations that lack privacy policies is not likely driven by an active aversion to assurances, but rather by a passive ignorance of potential risk when privacy concerns are not made salient. This interpretation is consistent with the notion that privacy concerns (and the lack thereof) are context-dependent (Acquisti, Brandimarte, and Loewenstein 2015; Acquisti, John, and Loewenstein 2013, John, Acquisti, and Loewenstein 2011) and often remain dormant until consumers are prompted to think about them (Marreiros et al. 2016). [ABSTRACT FROM AUTHOR]

The provision of assurances for self-adaptive systems presents its challenges since uncertainties associated with its operating environment often hamper the provision of absolute guarantees that system properties can be satisfied. In this paper, we define an approach for the verification of self-adaptive systems that relies on stimulation and probabilistic model-checking to provide levels of confidence regarding service delivery. In particular, we focus on resilience properties that enable us to assess whether the system is able to maintain trustworthy service delivery in spite of changes in its environment. The feasibility of our proposed approach for the provision of assurances is evaluated in the context of the Znn.com case study. [ABSTRACT FROM PUBLISHER]