,
13 tweets,
16 min read
Read on Twitter
Dark patterns are everywhere. In a year-long study we just released, we found dark patterns on over 1,200 sites, with about 200 sites being straight up deceptive.
Paper and data: webtransparency.cs.princeton.edu/dark-patterns/
NYT article by @jenvalentino based on our study:
nytimes.com/2019/06/24/tec…
Paper and data: webtransparency.cs.princeton.edu/dark-patterns/
NYT article by @jenvalentino based on our study:
nytimes.com/2019/06/24/tec…
@jenvalentino Deceptive practices are common in physical stores—mattress stores are always going out of business—so what's new? The scale, sophistication, and variety. We found 22 third-parties that offer "dark patterns as a service". The psychology research behind nudges has been weaponized.
@jenvalentino This is a paradigm shift for regulators: it is no longer enough to make rules for specific types of businesses that are known to be deceptive, like car dealerships or funeral homes. That's why the DETOUR act that seeks to rein in dark patterns is timely. warner.senate.gov/public/index.c…
@jenvalentino That said, many of the practices that we uncovered are already illegal in the US and the EU, and we hope that enforcement agencies will take immediate action. Our list of websites and dark patterns is available for download here: webtransparency.cs.princeton.edu/dark-patterns/…
@jenvalentino Our key innovation is a bot that automatically visits thousands of e-commerce sites, finds product pages, completes the shopping flow, and saves all the data, allowing us to scan for dark patterns in a mostly automated way using machine learning. 
@jenvalentino Of course, it took a huge amount of effort to develop the bot and analyze the data. Lead authors @aruneshmathur and Gunes Acar deserve major kudos; the other coauthors are undergrad Michael Friedman, first-year PhD scholar Elena Lucherini, @jonathanmayer @ineffablicious, & me.
@jenvalentino @aruneshmathur @jonathanmayer @ineffablicious We were inspired by many excellent pieces of previous work including:
A Purdue paper on dark patterns, led by @graydesign colingray.me/wp-content/upl…
A recent U Michigan study of 200 e-commerce websites carolmoser.com/wp-content/upl…
And of course @harrybr's darkpatterns.org.
A Purdue paper on dark patterns, led by @graydesign colingray.me/wp-content/upl…
A recent U Michigan study of 200 e-commerce websites carolmoser.com/wp-content/upl…
And of course @harrybr's darkpatterns.org.
@jenvalentino @aruneshmathur @jonathanmayer @ineffablicious @graydesign @harrybr Not all dark patterns are illegal, but they are nonetheless problematic because they are intended to prey on our cognitive limitations and weaknesses. In our paper we mapped each pattern to the corresponding cognitive bias from the psychology literature.
@jenvalentino @aruneshmathur @jonathanmayer @ineffablicious @graydesign @harrybr The most fun—and disturbing—part of the study was catching sites red handed at deception. In many cases we were able to find JavaScript that generates fake social proof messages. We've saved all those scripts.
@jenvalentino @aruneshmathur @jonathanmayer @ineffablicious @graydesign @harrybr If a site tells you that 7 people bought an item in the last hour, there's a good chance that that number was generated with a JavaScript call to Math.random(). Try refreshing the page!
@jenvalentino @aruneshmathur @jonathanmayer @ineffablicious @graydesign @harrybr Traditionally the law has been the main line of defense against deception. But maybe technical measures can also be effective. Many browsers now come built-in with protections against creepy online tracking; why not dark patterns as well?
@jenvalentino @aruneshmathur @jonathanmayer @ineffablicious @graydesign @harrybr Among the reactions to our study, I didn't expect this. Using dark patterns is not just unethical, it's also shortsighted. In addition to pissing off your users and regulators, you may be pissing off your own employees as well.
@jenvalentino @aruneshmathur @jonathanmayer @ineffablicious @graydesign @harrybr We do think the actual number is much higher; in this first study we focused on one single method of detecting dark patterns based on automated analysis of textual messages in the UI. We make clear in the paper that our numbers are underestimates.
