Privacy Risks and Ambiguity Preferences in the Internet of Things

We run two between-subjects treatments where we vary the general likelihood of a privacy risk that consumers encounter when they buy an internet-connected security camera. We employ one treatment with a ten percent and one treatment with a fifty percent privacy risk. Probabilities are stated to subjects in the description of the decision situation in the online experiment.

2022-06-07

2022-07-31

Ambiguity preferences of subjects when buying an IoT device which involves privacy risks. This is measured by subjects' decision to purchase either an internet-connected security camera whose manufacturer is transparent about the percentage of hacked cameras of its brand in the past or a security camera whose manufacturer is not transparent about this percentage.

The study is run as an online experiment with a scenario-based hypothetical decision environment. Treatments are randomized at the individual subject level. Participants will be recruited from Clickworker.com. Each subject participates in only one treatment (between-subject design).

Participants of the experiment are recruited from Clickworker.com and are blind to the research question of the study. As the experiment is conducted in German, participants must have indicated Germany as their country of residence and be proficient in the German language. The experiment is computerized with the online survey tool LimeSurvey. Participants receive a flat payment of about 3 EUR for their participation in the experiment. Subjects are informed that all information provided by them in the experiment will remain anonymous.

In the experiment, subjects are instructed to imagine they plan to buy an internet-connected smart-security camera for their home. They are told how high experts estimate the percentage of smart security cameras that have been hacked in the past and that they have the choice between two security cameras. Both cameras are equal in terms of price and features but differ with respect to the information their manufacturers give about the percentage of hacked security cameras of that brand in the past, i.e., the probability of privacy risk associated with the purchase of a camera. In the scenario description, subjects are informed that camera A transparently states the percentage of cameras of its brand that have been hacked in the past during their lifetime, while camera B does not state any information on that percentage. In two between-subjects treatments, we vary the general likelihood of a hack between fifty percent and ten percent. See the attached document for a screenshot of the translated description of the decision situation as displayed to subjects in the experiment.

In addition to the two treatments with likelihoods of a hack of ten and fifty percent, a third treatment with a likelihood of one percent may be considered depending on subjects' self reported perceptions of the treatments' risk likelihoods (measured in the experiment on a seven-point Likert scale).

The experiment contains two attention checks. Subjects who fail at least one attention check are automatically screened out of the experiment without finishing it and without being paid.

Randomization for treatment assignment by using random.org

Individual participant

No

Observations on the subject level are independent because subjects decide only once without interacting with other subjects before their decision. Thus, the number of clusters equals the number of observations.

We schedule data collection aiming at 1000 subjects.

500 oberservations per treatment.

Based on a chi-squared test with a statistical power of 0.8 and a significance level of 0.1, with the number of observations, we can detect a minimum effect size of Cohen's d = 0.12