Changing Robots into Strangers during Privacy Permission Requests

On this page, videos and photos of the pilot studies and experimental study of my master thesis are presented. The full title of my master thesis is: ''Agents presenting themselves as Strangers during Privacy Permission Requests: Effects on Disclosure and Privacy Awareness of Children''. The results of the study can be found in the thesis text.

Note: This research is part of the master Interaction Technology at the University of Twente. Some materials are therefore in Dutch.

Paper Abstract: Young children are prone to create strong bonds with robots and are very trusting of companies up until the age of 12. They give out a lot of information about themselves, even though they seem privacy aware. In recent papers, it became clear that the embodiment of a robot can facilitate higher levels of information disclosure, which creates serious privacy concerns as no regulations exist on this. Furthermore, not much research about this phenomenon has been done in relation to children, even though children are in contact with robots from a very young age and create strong bonds with them: this can make them more vulnerable to disclose information. These gaps in knowledge led to the experimental study (a 2x2 between-subject design) of this report in which 79 children, aged 8-12 years old, conversed with an embodied conversational agent for five minutes. The agent was either a Furhat robot or a Google Home Mini device. At several points in the conversation, the robot requested personal information in the form of a privacy permission request. Children's compliance on this determined an information disclosure score and their understanding of the content of the request determined a privacy awareness score. Besides the different levels of agent embodiment, a ``stranger presence'' within the agent could occur during a request. For the Furhat, this meant a change in voice and face, for the Google Home Mini a change in voice only. The stranger symbolised the company behind each service, as the people behind a company are often strangers to its users. Next to this, children are taught about not complying to strangers from a young age and they indicate that they seek privacy from strangers online as well. This novel approach of a stranger presence might therefore trigger their knowledge on stranger danger and lower their information disclosure. The results show that the level of agent embodiment and presence or absence of a stranger did not lead to significant differences in information disclosure and privacy awareness. Observations, children's comments and post-questionnaire measurements are used to go into further depth on these results.

Keywords: Child Robot Interaction, Stranger Danger, Embodiment, Information Disclosure, Privacy Awareness, Conversational Agents

First pilot study: identifying privacy sensitive variables

In the first pilot, a group of children (N=12) were asked about their online behaviour and privacy policy knowledge. Furthermore, they were asked to rate 16 variables on whether or not they would give this information to an application or website. The survey consisted of two parts.

Second pilot study: recognising auditory and visual changes

In the second pilot study 4 children aged 9-12 years old were send two media files: one was an audio file from the Google Home robot and one was a video file from the virtual Furhat interface. Each robot told a short part from a fairytale. In the middle of each fragment, the robot had a different voice or different voice/face. The children were only told to watch the fragments and comment on everything that they observed. Their parents made a small video or wrote in text what their children said about the fragments.

Visiting the primary schools

I made a presentation to introduce my research to children from ''groep 5'' up and until ''groep 8''. After this, I handed out information forms to give to their parents.

Research experiment

86 children were given parental consent to join the experiment, out of which 79 completed the full experiment. There was a 2x2 design, varying in the level of embodiment (high embodiment (Furhat robot), low embodiment (Google Home)) and ''stranger'' presence (Stranger/No Stranger). This lead to four conditions: GH NS, GH S, F NS and F S. This section contains the materials that were part of this research experiment.

Debriefing

Due to COVID-19 restrictions, the researcher was not allowed to visit the schools again to give a debriefing on the experiment. Therefore, she created a video that could be shown in the classroom, without her having to be present.