Privacy and Human Behavior in the Age of Information

Dated Jan 30, 2015; last modified on Tue, 28 Sep 2021

Privacy and Human Behavior in the Age of Information. A. Acquisti; L. Brandimarte; G. Loewenstein. Science, Vol. 347, Issue 6221, Page 509-514. Jan 30, 2015.

Incomplete and asymmetric information.

People rarely know what info others, firms and govts have about them or how the info is used.

Some privacy harms are tangible, e.g. financial costs associated with identity theft. Other privacy harms are intangible, e.g. strangers being aware of one’s life history.

Privacy typically involves trade-offs, e.g. privacy of consumers purchases protects them from price discrimination but excludes them from targeted offers and ads.

  • [Experiment: Online Shopping] When the search engine only displayed merchant prices, participants purchased from those offering lowest price. When the search engine included privacy protection practices in addition to prices, majority of the participants paid ~5% premium to buy from privacy-protecting merchants.

People tend to be uncertain about their own privacy preferences

[Experiment: Participants asked sensitive and potentially incriminating questions] When the questions were followed by assurances of confidentiality, participants had lower divulgence! Their privacy concerns went up.

[Westin’s Experiment: Used broad privacy questions in surveys to cluster people as privacy fundamentalists, pragmatists, and unconcerned] The Privacy Paradox: most people, when asked directly, consider themselves privacy fundamentalists.

  • “One might care deeply about privacy in general but, depending on the costs and benefits prevailing in a specific situation, seek or not seek privacy protection”

    • But the paradox persists even under high correspondence between expressed concerns and behavioral actions. [Experiment] Within the subset that had the highest degree of concern of strangers knowing their sexual orientation, political views and partner’s names, 48% publicly revealed their sexual orientation, 47% their political orientation and 21% their current partner’s name.

    • Privacy decision-making is also affected by misperceptions of those costs and benefits, as well as social norms, emotions, and heuristics. Any of these factors may affect behavior differently from how they affect attitudes.

Preference uncertainty is also evident in studies that estimate monetary valuations of privacy.

  • [Experiment] Subjects valued protection against errors, improper access, and secondary use of personal information between $30.49 and $44.62.

  • Attempts to pinpoint exact valuations that people assign to privacy may be misguided. [Experiment: Some subjects given $10 anonymous gift card, others given $12 traceable gift card. Later presented with option to switch] Of the $10 card holders, 52.1% chose to keep the $10 card. Of the $12 card holders, 9.7% switched to the $10 card. Maybe people value privacy more when they have it than when they do not.

Also, people have a powerful desire to be public, share and disclose. This may trump the desire for privacy.

Context-Dependence of People’s Concern About Privacy

Depending on the situation, people exhibit anything ranging from extreme concern to apathy about privacy.

Usually, we may be more comfortable sharing secrets with friends, but at times we may overshare with a stranger on a plane

[Study] Social network users over time increased the amount of personal information revealed to their friends while simultaneously decreasing the amounts revealed to strangers

People use cues to judge the importance of privacy, sometimes in unexpected ways:

  • Presence of government regulation reduces consumer concern and increase trust; people infer the existence of some degree of privacy protection

  • [Experiment] Subjects revealed more personal and incriminating info on an unprofessionally designed website with a ‘How Bard R U’ banner than on a site with formal interface. The latter was even deemed safer.

  • One’s culture and other people’s behavior, either through the mechanism of descriptive norms (imitation) or via reciprocity. If others are revealing information, people seem to reason unconsciously, doing so oneself must be desirable or safe.

Privacy concerns are often a function of past experiences.

Surveillance can produce discomfort and negatively affect worker productivity. However, privacy concern is adaptive.

[Experiment] Installation of monitoring tech in households made the families initially change their behavior, especially w.r.t. conversations, nudity and sex. Yet, if they accidentally walked naked in front of the sensors, they showed less concern about repeating the behavior.

Malleability and Influence

Disclosure of personal information powers industries like online social networks to behavioral advertising. Some entities have an interest in promoting disclosure.

Use of default settings. Sticking to default settings is convenient, and people often interpret default settings as implicit recommendations

Malicious Interface Design. Websites frustrate or even confuse users into disclosing personal information

Commercial entities choose not to ring alarm bells avoid raising privacy concerns. When they do, it’s usually a personalized ad with an irresistible deal.

[Study] 62% incorrectly believed that the existence of a privacy policy implied that a site could not share their personal information without permission

[Study] Control can reduce privacy concern. Participants provided with greater explicit control over which info could be published ended up sharing more sensitive information with a broader audience

Increasing the transparency of firms’ data practices doesn’t quite work:

  • Most internet users do not read privacy policies
  • Few users would benefit from doing so; most policies are in lawyer-speak

The feeling of being observed and accountable can induce people to engage in prosocial behaviors or (for better or for worse) adhere to social norms

  • [Experiment] Disinhibition Effect. Darkness caused participants in an experiment to cheat in order to gain more money.
  • [Experiment] Anonymity can lead to prosocial behavior. Higher willingness to share money in a dictator game, when coupled with priming of religiosity

Conclusions

Uncertainty and context-dependence imply that people cannot always be counted on to navigate the complex trade-offs involving privacy in a self-interested fashion

Malleability implies that people are easily influenced in what and how much they disclose.

Informing/empowering the individual is unlikely to provide adequate protection. (See transparency and finer control)

To be effective, privacy policy should protect real people - who are naïve, uncertain and vulnerable - and should be flexible enough to evolve with the complexities of the information age.