Chapter 2. The rise of deceptive patterns
When I started working on deceptive patterns, I was a little naive. I thought they might be eradicated if we could name and shame the companies that use these practices. Or at the very least, perhaps we could encourage UI and UX designers to use a code of ethics that would reduce the number of deceptive patterns in existence.
This approach didn’t work. In fact, things have become a lot worse since then. Deceptive patterns are everywhere now – there’s even a tip line that takes reports from concerned users and relays them to policymakers and enforcers around the world.1 The fact that we need a tip line at all means there’s clearly more to do.
To be fair though, deceptive patterns didn’t appear overnight. Deception is part of being human – in fact, it’s so common in the animal kingdom that we can even think of deception as a feature of life itself.2 The cover of this book features a Venus flytrap (Dionaea muscipula). This plant releases a scent that mimics the bouquet of fruits and flowers. Insects are attracted, and when they touch its sensory hairs inside the jaws, it snaps shut and traps the prey. This image is intended to be emblematic of unscrupulous tech companies who trick and trap their users using deceptive patterns.
Many historical stories and myths revolve around deception, such as ‘taking the King’s shilling’. In the 18th and 19th centuries, Britain spent a lot of time at war. But a career in the army or navy during wartime was not very attractive. With volunteers short on the ground, press gangs emerged to aggressively encourage recruitment, offering a shilling for every man who joined up. As the story goes, the act of receiving the coin was seen as a binding agreement, so unscrupulous recruiters would slip the coin into a sailor’s pocket or tankard of beer. When it entered their possession, the deal was done, and the men would be forced into naval servitude. Myth or not, the analogy with deceptive patterns is a strong one. Whether it’s clicking an ambiguously labelled button in a user interface or receiving a drink containing a hidden coin, it’s obvious that there’s a problem with the definition of consent if a person has no recourse after such a small, unintentional act.
It’s useful to think about what makes commercial deception and manipulation different today versus the pre-internet era. There are some aspects of modern technology that have acted as an accelerant or a catalyst, intensifying and spreading these practices.
The rise of metrics-driven culture
The idea of being driven by metrics dates back a long way: there’s archeological evidence of accounting records from Mesopotamia, 7,000 years ago. Crude as it may have been then, human beings have got better at measuring things over time, and we’re now fanatical about measuring things accurately.
What’s changed is that the barrier to measuring things is now much, much lower. You don’t need to be particularly clever or have a lot of capital to start measuring anything and everything you do in a business environment, and to start using data analysis to inform your business decisions.
In fact, metrics-driven management can be quite easy. You work out what metrics matter to your business, then you reward your teams for pursuing them using management techniques like performance-related pay, target metrics, bonuses and promotions. Of course, rewarding people for meeting a goal is almost the same as punishing them for not meeting it. In countries with less stringent labour standards, some companies use a management technique called ‘stack ranking’. This involves rating employees according to their performance on various measures, arranging them in rank order and then getting rid of the lowest performers. If an employee’s healthcare or immigration status is tied to their continued employment, this creates an enormous pressure on employees to do anything they can to hit their targets.
The web has also made it much easier to build and optimise deceptive patterns. With that in mind, I’d attribute the rise of deceptive patterns in software to the following general factors.
Easier tracking
Before the internet, it wasn’t easy to observe people without them being painfully aware of being watched. The traditional observation method was to send researchers to a store and have them stand there with a clipboard.3
But field researchers are costly and can only look at one thing at a time. Today, all you need to do is add a snippet of JavaScript4 to your website to get in-depth tracking that observes every conceivable behaviour of every user of your product simultaneously, and have it recorded into a huge database in the cloud. Business owners have also noticed another advantage to online tracking. Despite it being more invasive than ever before, people don’t feel anywhere near as worried about their privacy being invaded – because they don’t feel human eyes on them. All that tracking happens behind the scenes, out of sight and out of mind.
Then you’ve got the data processing. Before the internet, it was paperwork. Thousands of pieces of paper. Getting all the clipboards together, transcribing notes and recording them in a ledger. Doing calculations by hand to work out how many people did what, when, and how that impacted the company’s net income. Today, all of that calculation happens in the blink of an eye. Anyone can do it, using web-based software products like Google Analytics, Adobe Analytics, Mixpanel, Hotjar, or Amplitude.
These tools can give a wide manner of different insights: which ads or channels are driving traffic online, which pages are most effective at persuading users to take actions, the step in a series of pages at which users give up because they’re confused or frustrated, and more. All of these insights are then looped back into the...
These tools can give a wide manner of different insights: which ads or channels are driving traffic online, which pages are most effective at persuading users to take actions, the step in a series of pages at which users give up because they’re confused or frustrated, and more. All of these insights are then looped back into the design process, where changes are made to the product to boost conversion rates: the proportion of people who complete an action compared to those who do not.
Easier A/B testing
A/B testing5 was first used commercially in the early 20th century, but in those days it was an awkward, painstaking process.6 You could do it with newspaper ads: you’d run one version of your ad with a coupon, and another version with a different coupon. The version that won was the one that got the most coupons used. In those days, all the work was done by people; coupons delivered back to the agency were manually sorted and tallied by admin staff. It was a load of work and, of course, if your business wasn’t all about advertising general consumer products to the masses, you were stuck.
The limitations of the physical world mean you can’t apply the same kind of A/B testing to physical products and services as to digital without a great deal of cost and uncertainty. For example, if you have a shop on the high street, you can’t change the store layout from one customer to the next. Perhaps if you were Cobb from the film Inception, you’d be able to click your fingers and rearrange your shop floor at a whim. In the digital world, Inception-like remodelling is trivially easy. You can make two versions of a page or feature and easily find out which performs better. For example, version A of a page might say ‘20 other people are looking at this item’, while version B of the same page might say ‘Only 2 items left in stock’. Your A/B testing software then deploys version A to a random sample of users and version B to another. After the test is complete, your A/B testing software will automatically calculate statistics for you, telling you if either of the designs performed significantly better than the other on the measured conversion rate (purchases completed, for instance). You don’t even need to understand the statistics, as the results are usually dumbed down into simple sentences for you. No magic, cement, bricks or PhD needed. In fact, creating an A/B test today is as simple as signing up to a product like VWO or Optimizely free of charge and filling in a few forms.
A/B testing doesn’t judge whether a particular design is actually better or worse for the user – it just provides statistics as to whether design A or B performed better on your chosen metric. This means A/B testing opens a door towards deceptive patterns, because when a business tests a deceptive pattern against a more neutral pattern, typically it’s found to perform better on the chosen metric. Why? Because tricking or trapping users can be more effective than persuading them; and also because persuasion is frequently combined with deception, which means the overall page has two shots at capturing the user. It can start out by trying to persuade the user to complete the desired action. Then, if the user isn’t successfully persuaded, the deceptive pattern has a chance to get them to complete the desired action through nefarious means. Imagine some persuasive content followed by a preselected checkbox, for example. Some users will be persuaded by the content and will be happy with the default. Others won’t be persuaded and also won’t notice the preselected checkbox, so they’ll end up being tricked into opting into something they didn’t want.
When a deceptive pattern wins an A/B test, it’s often a direct source of revenue, with statistics to prove its effectiveness. In a metrics-driven environment, it can be very hard for employees to push back against this and encourage a more user-friendly – but less profitable – approach.
Copycat design
It was Oscar Wilde who said, ‘Imitation is the sincerest form of flattery that mediocrity can pay to greatness’. Some tech companies have been very successful in driving up conversion rates by using deceptive patterns. In response, others have copied them. This isn’t at all surprising. If you saw a competitor successfully making money for years without any legal or regulatory consequences, then why wouldn’t you copy them?