Li et al. 2024 – A Comprehensive Study on Dark Patterns
Citation
Li, M., Wang, X., Nie, L., Li, C., Liu, Y., Zhao, Y., Xue, L., & Said, K.S. (2024). A Comprehensive Study on Dark Patterns. ACM FSE (extended version). arXiv:2412.09147
Core Contribution
Introduces the Dark Pattern Analysis Framework (DPAF) with:
- A taxonomy of 68 dark pattern types across 6 categories
- Evaluation of 8 detection tools (45.5% coverage of taxonomy)
- Analysis of 4 datasets (44% coverage, 5,561 instances)
- Industry validation survey (173 participants, 87%+ approval)
Key Findings
Taxonomy: 68 types in 6 categories (Interface Interference (32%), Forced Action (18%), Social Engineering (15%), Sneaking (15%), Obstruction (10%), Nagging (1%)
Detection gap: Tools detect only 31 of 68 types. 37 types (54.5%) remain undetectable.
Dataset gap: Existing datasets cover 30 of 68 types. 38 types (56%) lack training instances.
Harm framework: Adopts OECD classification (H1 (autonomy), H2 (financial), H3 (privacy), H4 (psychological/time), H5 (competition distortion), H6 (trust erosion)