Privacy Decision-Making in Digital Markets: Eliciting Individuals’ Preferences for Transparency

We run two treatments, where we compare a high general probability of data loss against a low probability of data loss. Within a treatment, we analyze subjects’ ambiguity preferences regarding a data loss. We also consider an additional optional treatment, where the general setting is compared to a framed setting.

2021-08-19

2021-12-31

Ambiguity preferences of subjects for a data loss (ambiguity aversion or ambiguity seeking)

Ambiguity preferences are measured by probability equivalents of subjects' choices in an Ellsberg-type experiment and their deviations from the ambiguity-neutral probability.

Change of ambiguity preferences for a change in the ambiguity-neutral probability of a data loss.

Experiments are run online. Treatments are randomized at the session level. Participants will be recruited from the student subject pool of the University of Passau. Each subject participates in only one treatment (between-subject design). In all treatments, subjects are fully informed about the timeline of the experiment and the consequences of their actions.

Participants of the experiment are recruited from the University of Passau’s subject pool PAULA via the ORSEE platform (Greiner, 2015). As the experiment is conducted in German, participants must be proficient in the German language. The experiment is computerized with the online survey tool LimeSurvey and sessions are conducted with the video conferencing software ZOOM. Subjects attend the ZOOM sessions anonymously and with their camera and microphones turned off (unless the experiment leads to a disclosure of an individual’s test result). Participants receive a flat payment of 5 EUR for their participation in the experiment. Additionally, they receive 30 CENTS per correctly answered question in a logic test and 20 EUR for opting to participate in an Ellsberg-urn choice task that puts their test result at risk of being disclosed to the other participants in an experimental session.

Subjects give their consent to the experiment’s privacy policy before the start of the experiment and may exit at any time during the experiment. Subjects are fully informed about the experimental procedure. All steps of the experiment are explained in the experimental instructions.

At the beginning of the experiment, before any information is revealed, every participant independently chooses their personal decision color. The subject’s personal decision color may decide over the disclosure of the test result for the respective subject in the final stage of the experiment. After subjects have chosen their decision color, an audio recording of the experimental instructions is played to the participants in the Zoom conference while the textual instructions are shown on the subjects’ computer screens. After reading the instructions, subjects answer several comprehension checks.

Before subjects can proceed to the data collection stage, they indicate whether they want to participate in the Ellsberg-urn choice task. Those subjects that opt-out of the task only perform it hypothetically and, therefore, avoid the risk of having their test result disclosed. After subjects have decided whether to participate in the choice task, all subjects move to the logic test and then perform the choice task. Then, all participants answer several manipulation checks and fill out a questionnaire on privacy attitudes, information avoidance behavior and demographics.

In the choice task, subjects repeatedly choose between a situation of risk, where a loss of their personal data will occur with a known probability, and a situation of ambiguity, where a data loss will occur with an unknown probability. Subjects repeatedly choose between two bags filled with chips of different colors to operationalize the choice between a risky and an ambiguous option. Subjects are told about the composition of colors in the risky bag and do not know the composition of the ambiguous bag. If a subject is randomly selected at the end of the experiment, his or her choices determine from which of the two bags a chip will be drawn. If the color of the drawn chip matches the personal decision color, a data loss occurs for this subject. Overall, subjects make nine such decisions, with the risk of data loss increasing in the risky bag with every decision.

In the last stage of the experiment, one participant is randomly chosen by a draw from a lottery. If this participant has opted to perform the choice tasks, for this participant, one of nine decisions in the choice task is determined to be outcome relevant by a second draw from another lottery. For the randomly selected participant, the randomly determined decision is implemented. All random draws in the experiment are done live by the experimenter in the Zoom meeting. If a subject’s test result is determined to be disclosed, he or she has to turn on the web camera, and a photo is taken. Subsequently, the subject’s name is requested and verified with an official identification document. The name and photo, together with the subject’s test result, are then displayed to all the other participants in the Zoom conference.

Randomization by computer in office

Experimental sessions

No

Observations on the subject level are assumed to be independent, because subjects decide only once without interacting with other participants in the experimental session before their decision. Thus, the number of clusters equals the number of observations.

We schedule data collection aiming at 90 observations per treatment. This corresponds to six experimental sessions each, with 15 participants in a single session.

90 oberservations per treatment.

For our primary and secondary outcome we aim to detect an effect size of at least d = 0.4.