The computer science sub-area, which is part of the Institute of Computer Science at Heidelberg University, tackles the challenge of recognizing dark patterns in online interactions with the aim of warning users of dangers early on.
Software solutions for consumer protection are being used in many contexts: Spam filters protect users from fraudulent e-mails and ad blockers can suppress excessively aggressive advertising such as pop-ups.
Dark patterns can strongly influence consumer behavior, which is, for example, often exploited by online shopping portals. Software solutions similar to ad blockers which are able to protect users from this influence, or at least to point out potential dangers on websites, do not exist yet.
The informatics section researches to what extent dark patterns on websites can be recognized and classified. For this purpose, existing methods for recognizing patterns on websites and in texts are evaluated and tested for their practical suitability.
The challenge for Computer Science is twofold: First, a modeling language for dark patterns is required that can describe the structure and content of certain patterns. You can distinguish between the display of patterns, e.g. based on their arrangement on a website on the one hand, and on the other hand, the content-related aspects such as the textual content, which is directly visible for the user. In a second step, the developed modeling language supports a suitable pattern recognition to compare patterns with a reference pool and thus identifies problematic patterns.
Various AI-based classification techniques are used to detect dark patterns, which users can employ to make dark patterns visible and, by classifying the patterns in categories, make the associated problems more tangible. A particular challenge here is the heterogeneity of today's websites, which makes it necessary to classify patterns based on properties that are available on every website. Text-based approaches are therefore particularly promising since text can be identified on almost any website.
Finally, the knowledge gained is implemented in a Dark Pattern Detection App, which is intended to protect users from the influence of dark patterns. The measures that the app is able to take include highlighting dark patterns for the user and providing information about them. Thus, users understand why certain displays were classified as problematic. A second approach is the conversion of dark patterns into harmless representations. In this context, it is necessary to observe the requirements in the legal framework to avoid unlawful editing of works.