UI Dark Patterns and Where to Find Them: A Study on Mobile Applications and User Perception (CHI2020)

Linda Di Geronimo, Larissa Braz, Enrico Fregnan, Fabio Palomba, Alberto Bacchelli (University of Zurich, Switzerland)

Methodology (Classification)

- By following a taxonomy defined by Gray et al., we manually classified 240 trending Google free applications (US Play store).

- Each app was first used and recorded for ten minutes. The researchers recording the usage, perfomed a list of tasks that was kept similar among applications (create account, log out, log in, close and reopen app, etc.)

- A total of 40 hours of recording was then manually inspected by two researchers in pair

- An initial training phase with a third researcher was performed to improve accuracy

- 200 hours of classification were necessary to label Dark patterns in the 40 hours recordings of apps usage

Main Results (Classification)

- 95% of the 240 apps contained at least one or more Dark Patterns

- An average of 7,4 Dark Pattern for each app

- Nagging, False Hierarchy and Preselection were among the most common Dark Patterns (video for examples)

Methodology (User Experiment)

- We selected five Applications containing Dark Patterns from our dataset

- We recorded 30 second video of usage of these applications (in the video only one type of Dark Pattern was present)

- Each user will only see three videos: two randomly selected from the five containing Dark Patterns, and one without Dark Pattern. The NoDP case was the same among all users

- We would then ask users if they spotted any malicious design after watching the first video. We expected them to learn after the first case

Main Results (User Experiment)

- 55% of our users did not spot Dark patterns, 20% were not sure

- We found that the more users are knowledgable on the topic of Dark Patterns the more easily they will spot them

- Parents seemed to be particularly interested about the topic

Materials

All the supplementary material are available here. At the moment, only a portion of the recordings are available for two main reasons: the size of each video is very high (circa 200gb), and many portions had to manually anonymised to preserve authors' privacy. his last process will take a long time and resources. However, if you would like to do further research on the videos, please contact us (hilindig@gmail.com, bacchelli@ifi.uzh.ch) and request a password for the following shared folder: link. Please take in consideration the following agreement before requesting the data.

Acknowledgements

L. Di Geronimo gratefully acknowledges the support of the Digital Society Initiative of the University of Zurich. A. Bacchelli, E. Fregnan, L. Braz and F. Palomba gratefully acknowledge the support of the Swiss National Science Foundation through the SNF Projects No. PP00P2 170529 and No. PZ00P2 186090. L. Di Geronimo also thanks Professor Moira Norrie for inspiring her on this topic.