Chapter 21: How deceptive patterns can be more harmful when combined

In 2021, Lior Strahilevitz, a law professor at the University of Chicago, and Jamie Luguri, a PhD in experimental social psychology, came up with a clever way to investigate the quantitative impact of deceptive patterns. They created an online survey that had deceptive patterns in the final section. The main part of the survey was about privacy, but these were all decoy questions, acting as a prelude to the deceptive pattern that would appear near the end.

Strahilevitz and Luguri created a number of different versions of the survey: a control version without any deceptive patterns; a ‘mild’ dark pattern version; and an ‘aggressive’ deceptive pattern version. You can see the control and mild versions in the figures below. I haven’t included the aggressive version here because it’s so bulky. If you’re interested to read more, you can refer to their 2021 paper, ‘Shining a Light on Dark Patterns’.1

The control condition in the study contains textual information about the offer to the user, with the benefit and cost in bold. Below the text are two options: ‘Accept’ and ‘Decline’.
Image depicting a question from the control condition of the Luguri and Strahilevitz study (2021).

In their study, the control condition asks the user if they want to sign up for a ‘data protection and credit history monitoring’ service that costs $8.99 a month after a free trial period. The user can then either accept or decline without any funny business – there are no deceptive patterns at work.

The mild deceptive pattern makes things a bit more difficult for the user, as shown in the figures below.

The ‘mild dark pattern’ condition of the study contains the same text and emphasis as the control condition, but the first option has been changed to ‘Accept and continue (recommended)’ in bold, and the second is ‘Other options’
Mild condition 1/3, Luguri and Strahilevitz (2021).
If the user selects ‘Other options’ another two options are offered: the first is ‘I do not want to protect my data or credit history’; the second is ‘After reviewing my options, I would like to protect my privacy and receive data protection and credit history monitoring’.
Mild condition 2/3, Luguri and Strahilevitz (2021).
If the user chooses not to protect their data, a final set of options are presented under the heading ‘Please tell us why you decided to decline this valuable protection’. They are: ‘My credit rating is already bad’; ‘Even though 16.7 million Americans were victimized by identity theft last year, I do not believe it could happen to me or my family’; ‘I’m already paying for identity theft and credit monitoring service’; ‘I’ve got nothing to hide so if hackers gain access to my data I won’t be harmed’; and ‘Other’ with a text box to complete. A final option beneath this is ‘On second thought, please sign me up for 6 months of free credit history monitoring and data protection services’.
Mild condition 3/3, Luguri and Strahilevitz (2021).

To summarise, in the mild deceptive pattern condition the user is initially given the same paragraph of text, but they can then pick either ‘Accept and continue (recommended)’, which is highlighted in bold; or they can pick ‘Other options’. Picking ‘Other options’ takes them to another step with more radio buttons. If they try to opt out again (‘I do not want to protect my data or credit history’), they are taken to yet another step. This design contains a range of different deceptive patterns: visual interference, trick wording and obstruction. The aggressive deceptive pattern condition is similar, except it has even more steps and lays on even more pressure. The aggressive version also has a countdown timer that forces users to dwell on the subsequent pages, so they can’t just skip through quickly.

The researchers deployed this survey to 1,963 participants. The scale of the impact was staggering...

Buy the book to

Since 2010, Harry Brignull has dedicated his career to understanding and exposing the techniques that are employed to exploit users online, known as “deceptive patterns” or “dark patterns”. He is credited with coining a number of the terms that are now popularly used in this research area, and is the founder of the website He has worked as an expert witness on a number of cases, including Nichols v. Noom Inc. ($56 million settlement), and FTC v. Publishers Clearing House LLC ($18.5 million settlement). Harry is also an accomplished user experience practitioner, having worked for organisations that include Smart Pension, Spotify, Pearson, HMRC, and the Telegraph newspaper.