Abstract

This research explores the textual and visual detection of dark patterns in e-commerce websites using Large Language Models and image recognition. It builds on Arunesh Mathur’s taxonomy from the Dark Patterns at Scale paper, published in 2019. The study has two main outcomes: First, the development of an open-source Chrome plugin to identify dark patterns on websites, and second, the analysis of a dataset of websites using a multimodal approach for dark pattern detection on a dataset of 256 e-commerce websites. This analysis reveals current manipulative trends across various dark pattern categories and offers insights for designers advocating for increased awareness to counter certain long-standing manipulative practices in e-commerce.