Into the Dark: The Fault of Machine Learning Methods in Detecting UX Dark Patterns

Thumbnail Image

Files

Chu_Duc_2025.pdf (312.47 KB)

URL

Journal Title

Journal ISSN

Volume Title

Perustieteiden korkeakoulu | Bachelor's thesis
Unless otherwise stated, all rights belong to the author. You may download, display and print this publication for Your own personal use. Commercial use is prohibited.

Authors

Date

2025-05-08

Department

Major/Subject

Data Science

Mcode

SCI3095

Degree programme

Aalto Bachelor’s Programme in Science and Technology

Language

en

Pages

30+4

Series

Abstract

The world of design and the world of science has inextricably been intertwined, but it has rarely been truly studied. Especially in the smaller field of UX Design, there is one specific subject that should be discussed: dark patterns. Dark patterns are tricks used in digital spaces such as websites and apps that make the user do something they did not intend to do. Some examples of dark patterns are forcing users to give personal information to access free content, i.e., forced continuity, and tricking users into expecting certain outcomes but producing a different undesirable outcome, i.e., bait and switch. The consequences of these deceptive design tactics can range from loss of customer trust to financial loss to leakage of confidential personal information. There have been many serious attempts at categorizing dark patterns into taxonomies and building machine learning methods to detect them automatically. However, research into machine learning methods to detect these patterns is still in its infancy, requiring much more delicate insight and analysis. In this thesis, a state-of-the-art literature review is conducted on dark pattern taxonomies and the machine learning models used to detect these patterns. The shortcomings of the dark pattern taxonomies are their low comprehensiveness, narrow scope, potential of missing future updates and low usability for detection purposes. The shortcomings of the machine learning models consist of the usage of vague taxonomies, failure to detect dynamic dark patterns, failure to detect non-textual dark patterns, overfitting and the lack of fully autonomous models. Other dark pattern detection models that does not use machine learning are also analyzed to provide insights for future machine learning models. Some future directions are then suggested for both dark pattern taxonomies and machine learning models. The paper contributes to the dark pattern detection field by not only deepening the understanding of dark pattern taxonomies and detection tools, but also providing insights for future models and research.

Description

Supervisor

Korpi-Lagg, Maarit

Thesis advisor

Sahlsten, Jaakko

Keywords

automatic detection methods, dark patterns, deceptive patterns, machine learning, UX design

Other note

Citation

Permanent link to this item

https://urn.fi/URN:NBN:fi:aalto-202506104351

Collections

[kand] Perustieteiden korkeakoulu / SCI