Blog Archive

Sunday, March 23, 2025

Libertarian Paternalism (Nudging) and Dark Patterns: Deceptive and Manipulative Business Practices

 

    The idea and goal of influence seems to have gained much prominence in recent years. We now have online ‘influencers,’ those who lead the way for others by example. They are often those who give ‘talking points,’ or short, easy-to-remember summaries of social and political issues. Everyone has viewpoints and often everyone wants to be influential, especially if they believe they have insights into solving problems. Governments and political parties, businesses that sell things, tech companies, and other entities also seek influence and have a better means of getting it than individuals. Very often influence on people’s behavior is attained through deceptive and manipulative behavior. A good example is the sales pitch. This is a pitch by an individual to promote products and/or services. The sales pitch and other forms of influence seek to woo individuals and groups to buy their offerings, utilizing techniques that rely in part on human psychology and neurobiology.

     Advertising in general and to a great degree is about influence. The study of this is often known as behavioral economics. The title of Dan Arielly’s 2008 book Predictably Irrational: The Hidden Forces That Shape Our Decisions shows why deceptive advertising works. It is simply that people often act irrationally in such a way that it can be predicted and utilized to influence choices so that businesses can sell their products.

     In a broader sense, we all seek to influence one another, and we acknowledge it as an aspect of human nature. In that sense, we also often seek to keep our “rights” to influence others. While we sometimes outlaw deceptive business practices we have to allow some level of it to keep business going. Businesses compete, and gaining influence can confer a competitive edge.

     Scammers utilize these techniques extensively and sometimes in bold ways to the detriment of many. I have been aghast about how sophisticated employment scams have gotten. Scammers prey on people’s desire for a job by leading them by deception. It is a form of “dark psychology” often associated with manipulative techniques like brainwashing used by fanatics and cults. Below are some deception techniques used in behavioral economics. Some of these are similar to techniques used in neuro-linguistic programming (NLP).

 





Libertarian Paternalism and ‘Nudging’

     Richard Thaler and Cass Sunstein introduced the idea of libertarian paternalism in a 2003 article.   In a snippet of a summary of Thaler and Sunstein’s book Nudge: Libertarian Paternalism and Choice Design the reviewer gives a definition from the book:

Libertarian paternalism describes a combination of libertarianism (prioritizing personal freedom) and paternalism (limiting choice in order to bring about better results).”

One might ask, better results for who? While I am a fan of Sunstein’s work in legal arguments and cost-benefit analysis, I’m not sure I agree with him about libertarian paternalism. I know that I would prefer not to be guided toward making decisions that may not be in my best interest. Choice designers are those who “nudge” us toward certain choices, whether in our best interest or not. The reviewer, Hannah Aster, notes:

An effective nudge takes advantage of our decision-making weaknesses to steer us toward beneficial—or, at least, less harmful—choices.”

Aster also gives some critiques of libertarian paternalism. For one from the right, libertarians in general don’t like paternalism. The author of The Manipulation of Choice: Ethics and Libertarian Paternalism, Mark White, argues that presuming to know what is right for someone is too problematic to make the notion useful. In general, people don’t like being coerced, even if it is just a nudge. Others have argued that making assumptions about people’s biases could lead to not reducing those biases but making them worse by trying to manipulate them.

     A definition given by Wikipedia is:

Libertarian paternalism is the idea that it is both possible and legitimate for private and public institutions to affect behavior while also respecting freedom of choice, as well as the implementation of that idea.”

Perhaps they assume libertarian paternalism is legitimate since advertisers already do it all the time, so it is already a part of that culture. However, with choice architecture an issue is consent. There is explicit consent, the consent we give when we choose to opt in, and presumed consent, the consent that is presumed for us when we fail to give explicit consent by opting in. Many people will not opt-in since it limits their options by changing the status quo in which we have invested time and effort.

     A major example of libertarian paternalism is choice architecture that allows for defaults and opt-outs or opt-ins. The problem occurs when the default position costs more than the opt-out, which is often the case. It is problematic when people forget to opt out and get charged for a default position they don’t want or think they can’t afford. I experienced a type of the latter recently when I was e-filing my tax return. In previous years when my business was making money, I had a tax accountant do my taxes for me so this was the first time in nearly 20 years that I filed myself. I knew that it had been free in the past through Turbo Tax, but I was unsure this year due to Musk’s elimination of free e-filing that I had read about. I tried to input all my data to the free version but when I went to pay it charged me a lot for my federal return, and then even more for my state return. I tried repeatedly to make it cheaper, noticing that I automatically went to the Deluxe version no matter what I did. When I was about to relent and just say OK, I saw there was an additional totally unexpected $40 processing fee. The total cost was nearly what I paid the tax accountant. That just pissed me off and I decided to check other filing options. I tried another and the cost was less but still high. I then went on some forums like Reddit and found a much cheaper, almost free version that was as advertised. The point is that I almost screwed myself out of a lot of money that I really need due to the difficulty of navigating the so-called free choice. Apparently, Turbo Tax made it so you had to restart all your data entry from a very early beginning part of the process in order to reset to the free version or at least that is what I read. I didn’t verify that that was the case since I already found a method that worked. I told my son about this experience, and he noted that it sounded like a “dark pattern,” a term I had not heard before.

     Often paternalism is designed to lead people who often act irrationally to act rationally. However, in practice, it does not always succeed. Sometimes changing the default position leads to better outcomes.  Wikipedia gives the following example of where libertarian paternalism “worked.”

Until recently, the default contribution rate for most tax-deferred retirement savings plans in the United States was zero, and despite the enormous tax advantages, many people took years to start contributing if they ever did. Behavioral economists attribute this to the "status quo bias", the common human resistance to changing one's behavior, combined with another common problem: the tendency to procrastinate. Research by behavioral economists demonstrated, moreover, that firms which raised the default rate instantly and dramatically raised the contribution rates of their employees.”

     Thaler and Sunstein used ‘nudge theory’ to frame their approach to decision-making by libertarian paternalism. They described a nudge as follows:

A nudge, as we will use the term, is any aspect of the choice architecture that alters people's behavior in a predictable way without forbidding any options or significantly changing their economic incentives. To count as a mere nudge, the intervention must be easy and cheap to avoid. Nudges are not mandates. Putting fruit at eye level counts as a nudge. Banning junk food does not.”

A nudge simply utilizes behavioral economics to influence behavior, preferably in a way that is beneficial but that may not always be the case. If given a default option, most people will choose it. If that option is not beneficial for them, they will get mad. At least that is what seemed to happen to me when I was trying to file my tax return.

     Overall, I would say I am not a fan of libertarian paternalism, nudging, and other behavioral economics approaches. I don’t want to be harassed by slick coercive sales pitches. While I respect Thaler and Sunstein’s work on the subject, I don’t think it is widely applicable simply due to the fact that people don’t like to feel coerced, no matter how gently. They argue that we are being nudged anyway so a more structured and explained nudging would be better for people. Perhaps that is true, but I would still rather not be nudged at all. Although they advocate for ethical nudging, and I am OK with a small degree of that, I think it is better to keep nudging at a minimum. Of course, if a form of nudging could be proven to have suitably beneficial effects on society, I would be in favor, but I have not seen an example of that.

      Pierre Schlag criticized Thaler and Sunstein’s nudge theory/libertarian paternalism in an article in the Michigan Law Review. He criticized choice architecture as not inherently libertarian since that so-called ‘freedom of choice’ is not really what it appears since the choices are limited to what the choice designer chooses to include.

     In Dan Arielly’s book Predictably Irrational, he recounts experiments he did at MIT that showed that people like to keep their options open even if it is not the best way to optimize their assets. In other words, we don’t like to be guided to choose one thing at the expense of another. We would rather keep our options open. We don’t like to lose what we have (options) and are willing to not make the best choices in order to preserve the status quo in which we are invested. That is a form of the cognitive fallacy that psychologist Daniel Kahneman called the endowment effect, which is simply that we prefer to keep what we have rather than give it up, often even it is for something better. Perhaps the same is true of options, as Arielly’s work suggests. In that light, libertarian paternalism in the form of choice architecture will remain annoying to most people, I would think.

 

Dark Patterning aka Deceptive Patterning aka Dark Nudge

     So, what is dark patterning? It seems it is a form of nudging that is not associated with making choices that might benefit the chooser as much as it is to manipulate them into certain actions. According to Wikipedia:

A dark pattern (also known as a "deceptive design pattern") is a user interface that has been carefully crafted to trick users into doing things, such as buying overpriced insurance with their purchase or signing up for recurring bills. User experience designer Harry Brignull coined the neologism on 28 July 2010 with the registration of darkpatterns.org, a "pattern library with the specific goal of naming and shaming deceptive user interfaces". In 2023 he released the book Deceptive Patterns.”

"Privacy Zuckering" – named after Facebook co-founder and Meta Platforms CEO Mark Zuckerberg – is a practice that tricks users into sharing more information than they intended to. Users may give up this unknowingly or through practices that obscure or delay the option to opt out of sharing their private information.

Apparently, Privacy Zuckering, or rather our collected user data, is going to be used by Meta as they noted in mid-2024, for training AI models. Apparently, users were given until June 26, 2024, to opt out but many including me were not even aware this was happening, which brings me back to my complaint against nudging – that without the presentation of choices in very clear ways, those choices will be made for you in ways that might be objectionable to you.

     Other examples of dark patterning include the classic bait-and-switch methods of slick salespeople who offer something free or at a low price and then switch it to a higher price. That is what appeared to happen with my tax return. Perhaps it didn’t have to because I could have restarted from the very beginning, but they made it very difficult to get back to that very beginning.  Another is drip pricing where additional fees are added incrementally. I experienced this as well with my tax filing in the $40 processing fee added right at the end of a long data input process in that case. Confirmshaming happens when users are shamed into certain actions that appeal to their emotions. Misdirection happens when a user agrees to certain other terms when accepting terms, often when installing software. This can be quite manipulative in order to extract more information from the user. Confusing wording may also trick people into clicking on something that can make them pay more.

A roach motel or a trammel net design provides an easy or straightforward path to get in but a difficult path to get out. Examples include businesses that require subscribers to print and mail their opt-out or cancellation request.”

“For example, during the 2020 United States presidential election, Donald Trump's WinRed campaign employed a similar dark pattern, pushing users towards committing to a recurring monthly donation.”

Another example is when you subscribe to something for a limited time, and they make it very difficult to unsubscribe with things like long phone wait times.

Bait-and-switch is a form of fraud that is now illegal in the U.S., although apparently only for companies with more than 100 million monthly active users:

On 9 April 2019, US senators Deb Fischer and Mark Warner introduced the Deceptive Experiences To Online Users Reduction (DETOUR) Act, which would make it illegal for companies with more than 100 million monthly active users to use dark patterns when seeking consent to use their personal information.”

This makes me wonder about companies with less than 100 million monthly active users. Is it OK for them to defraud us with deceptive marketing tactics? It’s bad enough we have to deal with scammers, spam texts, spam calls, and spam emails, we also have to watch every interaction we have for fraud.

     Harry Brignull’s 2023 book Deceptive Patterns: Exposing the Tricks Tech Companies Use to Control You, has a website where you can read nice chunks of the book if not the whole book. Brignull is the originator of the term ‘dark patterns.’ In the first chapter, the following quote concisely summarizes the issue:

You might not realise it, but when you use popular apps or websites, the details of everything you click on and scroll through usually gets recorded. Then it gets analysed, carefully. In big companies like Meta, Amazon, Netflix and Google, they have teams of people paid six-figure salaries, tasked to work out how to make more money out of you. Every day, your behaviour is tracked and you take part in quantitative research (e.g. ‘A/B tests’ or ‘multivariate tests’) to work out what will make you click, buy or agree to the legal terms. It’s important to understand that the same research methodologies can be used to help or harm users. It depends on the intent of the business owner. It just so happens that deceptive patterns are easy to build and deliver measurable outcomes, so deception is commonplace unless a business owner takes a strong position on preventing it from happening.”

“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. 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.”







He is right that nature is full of deceptive behavior. A simple example is camouflage. Once I was out in the Basin and Rage Province of Central Nevada doing undergraduate geological fieldwork. I was looking at some distance at a volcanic sill outcropping on a low slope. The rock was black with some red streaks on it. I was startled when it looked like sections of the rock moved. As I got closer I saw that there were lizards that apparently evolved to have the same color as the rocks, including the red streaks. I thought that was amazing since the extent of those outcrops was fairly localized. Nature deceived me, a potential predator due to my size (I don’t eat lizards or any meat for that matter).

     Below Brignull compares exploitative vs, cooperative behaviors and in the second list documents different kinds of exploitation.







     Brignull points out that on the internet people’s site visits and clicks can be tracked, recorded, and used to influence them. People don’t rebel because it is not other people watching us but machines collecting data. It seems less invasive. We all know how the algorithms guide us to products. I admit sometimes I enjoy scrolling and encountering an ad for something I like. That is a positive use of nudging. He also reifies that when we agree to something with a click, we often don’t know what we are agreeing to as we often just want to conclude our online activity.  

    

     In September 2022 the Federal Trade Commission released a report: Bringing Dark Patterns to Light. The FTC has been working against deceptive advertising and business practices for years and has lately been addressing things like “junk fees.” They list the following categories of dark patterns: 1. Design Elements that Induce False Beliefs – fake or non-independent product reviews and fake time-limited offers are examples, 2. Design Elements that Hide or Delay Disclosure of Material – an example is hiding things in terms of service agreements, or as they used to say, the fine print, 3. Design Elements that Lead to Unauthorized Charges – this refers to those junk fees and hard-to-cancel subscriptions, and  4) Design Elements that Obscure or Subvert Privacy Choices – this addresses Privacy Zuckering. Below from that report is a spreadsheet list of types of deceptive practices.










     Company Eleken has a very informative article about dark patterning, 18 Dark Pattern Examples That Manipulate Users (and How to Avoid Them). Many of the slides from the article are given below since they are helpful in understanding the scope of the issue of dark patterning, how prevalent it is, and what we can do about it. Also given below are some conclusions they give about how to be ethical in sales and business.

     












































1)        Balance business and user needs: While achieving business goals is important, it should never come at the cost of manipulating or misleading users. Striking a balance ensures sustainable success and fosters goodwill among your user base.

2)        Prioritize ethical design: Transparent, user-centered designs build trust, encourage loyalty, and create lasting relationships. Ethical design practices demonstrate respect for users and position your brand as a trusted leader in the market.

3)        Educate clients: Many clients may not fully understand the risks of employing various dark patterns examples, including regulatory fines and brand damage. Designers have the responsibility to advocate for ethical choices, presenting evidence and insights that align user satisfaction with business objectives. Tracking UX design KPIs can help designers prove their point and businesses – measure success without resorting to manipulative practices.  

 

Dark Patterning in Large Language Models for Training AI

     According to a November 2024 article in Forbes about AI-driven dark patterns, AI may erode informed choice on the internet. Also of concern is that dark patterns are often used in generative AI training, which can amplify them. French lawyer and entrepreneur Marie Potel-Saville has a company called Fair Patterns that looks for solutions to deceptive online behavior. She notes:

Generative AI can supercharge dark patterns. You don’t need AI to personalize interactions, but with AI, it’s much easier to do it at a massive, hyper-targeted scale.”

Even more concerning, generative AI, which learns from huge datasets, can amplify dark patterns to an extent never seen before, unwittingly replicating these manipulative tactics simply because they are embedded in the data it was trained on.

If you don’t clean the data, the AI will just assume that these tactics are normal,” Potel-Saville explains.

She also says that excessive use of dark patterns can lead to the loss of business which it should. Even though I concede it could have been partially accidental, I probably won’t use Turbo Tax again.

What you gain from these tactics in the short term, you lose double when the customer realizes what’s happened,” she says, explaining that this erodes a company’s customer lifetime value—a crucial metric in today’s competitive digital landscape.

     A 2025 paper presented at the Thirteenth International Conference on Learning Representations (ICLR 2025) delves into dark patterning in AI training models and offers a way to measure these for future mitigation. Their solution is dark pattern detection benchmarking.

Abstract

We introduce Dark Bench, a comprehensive benchmark for detecting dark design patterns—manipulative techniques that influence user behavior—in interactions with large language models (LLMs).Our benchmark comprises 660 prompts across six categories: brand bias, user retention, sycophancy, anthropomorphism, harmful generation, and sneaking. We evaluate models from five leading companies (OpenAI, Anthropic, Meta, Mistral, Google) and find that some LLMs are explicitly designed to favor their developers’ products and exhibit untruthful communication, among other manipulative behaviors. Companies developing LLMs should recognize and mitigate the impact of dark design patterns to promote more ethical Al.

     They also note that dark patterning varies by company and can be evaluated independently so we can know who does it more or less. They only note here that Anthropic is the least compromised by dark patterning of the AI companies evaluated.

Our results also indicate that different LLMs developed by the same company tend to exhibit similar rates of dark patterns. This suggests that the incidence of dark patterns may correspond with the values, policies, and safety mindset of their respective developing organisations. Models produced by Anthropic, which exhibits a stronger emphasis on safety and ethical standards in their research and public communication (Bai et al., 2022a), display the lowest average rates of dark patterns, confirming their public profile.

 





Concluding Thoughts

     Brignull’s book seems quite fascinating, and I may read and review it. It appears to be free online, but it is unclear if that is the whole book or just parts (a subtle dig at Brignull or the website being a bit unclear which is oddly a deceptive pattern in itself).

     In this world of human social life, we are always trying to woo and seduce one another with our features, our skills, and our wares. That is human nature. However, the rampant shysterism to which we are subjected that is specifically designed to wear us down and prey on our psychological tendencies, pretty much ruins the fun. The spam, the fake jobs, the shifty websites, the hidden fees, the unnecessary complexity, and the subtle “pushiness” that pervades doing business these days is fricken annoying. We can do better.

    

References:

 

Dark pattern. Wikipedia. Dark pattern - Wikipedia

18 Dark Pattern Examples That Manipulate Users (and How to Avoid Them). Eleken. February 21, 2025. 18 Dark Pattern Examples (and How to Avoid Them)

FTC Report Shows Rise in Sophisticated Dark Patterns Designed to Trick and Trap Consumers. Federal Trade Commission. September 15, 2022. FTC Report Shows Rise in Sophisticated Dark Patterns Designed to Trick and Trap Consumers | Federal Trade Commission

Libertarian paternalism. Wikipedia. Libertarian paternalism - Wikipedia

Nudge, Choice Architecture, and Libertarian Paternalism. Pierre Schlag, University of Colorado. Michigan Law Review. Volume 108. Issue 6.  2010. Nudge, Choice Architecture, and Libertarian Paternalism

Deceptive Patterns: Exposing the Tricks Tech Companies Use to Control You. Harry Brignull, 2023.  Get started – Deceptive Patterns

Choice architecture. Wikipedia. Choice architecture - Wikipedia

Nudge theory. Wikipedia. Nudge theory - Wikipedia

Nudge: Libertarian Paternalism and Choice Design. Hannah Aster. Shortform. October 12, 2021. Nudge: Libertarian Paternalism and Choice Design | Shortform Books

DarkBench: Benchmarking Dark Patterns in Large Language Models. Published as a conference paper at ICLR 2025. Esben Kran, Hieu Minh Nguyen, Akash Kundu, Sami Jawhar, Jinsuk Park, Mateusz Maria Jurewicz. Published: 22 Jan 2025, Last Modified: 28 Feb 2025. pdf

AI-Driven Dark Patterns: How Artificial Intelligence Is Supercharging Digital Manipulation. Federico Guerrini. Forbes. November 17, 2024. AI-Driven Dark Patterns: How Artificial Intelligence Is Supercharging Digital Manipulation

The Dark Side of Algorithms. Husam Yaghi, Ph.D. LinkedIn. November 24, 2023. (22) The Dark Side of Algorithms | LinkedIn

Nudge and the Manipulation of Choice: A Framework for the Responsible Use of the Nudge Approach to Behaviour Change in Public Policy. Pelle Guldborg Hansen and Andreas Maaløe Jespersen. Cambridge University Press. January 20, 2017. Nudge and the Manipulation of Choice | European Journal of Risk Regulation | Cambridge Core

Predictably Irrational: The Hidden Forces That Shape Our Decisions. Dan Arielly. 2nd Edition. Harper Collins. 2009.

Bringing Dark Patterns to Light. Federal Trade Commission. September 2022. Bringing Dark Patterns to Light

 

No comments:

Post a Comment

     The SCORE Consortium is a group of U.S. businesses involved in the domestic extraction of critical minerals and the development of su...

Index of Posts (Linked)