Dark patterns: Can criminal law remedy the shortcomings of antitrust law?

Abstract

Antitrust laws' remedies to address Big Tech's anticompetitive business models have often been questionable. Since the fundamental design choices of business models pose significant challenges, due to dark patterns, Austrian criminal law may compensate for antitrust laws' enforcement limits. It is argued that Amazon's Prime and BuyBox, Google's Search and its AI Bard, Activision Blizzard's lootboxes, and Engagement Optimised Matchmaking (EOMM) algorithms contain dark patterns that may constitute (commercial) fraud within the meaning of §§ 146, 148 StGB. This is because these business models are designed to deceive consumers to unlawfully enrich Big Tech and video game publishers.

Antitrust law’s regulatory dilemma

Antitrust law has undergone a significant transformation in recent years. While it initially focused on regulating apparent anticompetitive practices, it has broadened its scope by scrutinising design choices in Big Tech business models. ¹ However, due to questionable remedies ² and protracted litigation, ³ antitrust interventions can foster legal uncertainty, making the regulation of product design increasingly arduous. ⁴ This is exacerbated by the inability to learn from past mistakes in designing remedies. ⁵ Moreover, the Commission's decision in Microsoft/Activision Blizzard ⁶ demonstrates its unpreparedness to understand complex digital product designs and innovation agendas ⁷ highlighting another significant failure of the Commission to learn from past mistakes. ⁸ While regulatory efforts are being made to curb Big Tech's anticompetitive behaviour, ⁹ they do not adequately address digital business models ¹⁰ and may in fact be a restatement of existing standards ¹¹ that fails to account for digital business models that do not reach the artificial filtering system of the Digital Markets Act (DMA). ¹² Additionally, the question of whether dark patterns are anticompetitive is yet to be resolved. ¹³

Because dark patterns affect all demographics uniformly, ¹⁴ regulatory efforts could never be limited to one area of law or regulatory approach. ¹⁵ In the United States, for example, there are suggestions to regulate dark patterns that focus on data collection in a cooperative effort between privacy and antitrust law. ¹⁶ In Europe, the Digital Services Act (DSA) prohibits the use of deceptive or manipulative product designs, i.e., dark patterns. ¹⁷ Failure to comply can result in fines of up to 6% of an undertaking’s global turnover. ¹⁸ However, like the DMA, the DSA operates an artificial filtering system ¹⁹ that limits its effectiveness. This is particularly problematic because ‘97% of the most popular websites and apps used by EU consumers deployed at least one dark pattern.’ ²⁰ Moreover, we know from antitrust law that Big Tech undertakings often treat fines as prizes and do not take them seriously, which further hinders effective regulation. Thus, digital regulation and antitrust law may not be sufficient to regulate dark patterns. ²¹

Criminal law can complement the regulatory efforts of dark patterns if they constitute fraud. Austrian law provides for the criminal prosecution of organisations under the Verbandsverantwortlichkeitsgesetz (VbVG) or Corporate Criminal Liability Act 2006. ²² In addition to penalties for individuals of up to six months imprisonment for fraud and up to five years imprisonment for commercial fraud, ²³ the VbVG allows a legal entity to be held criminally liable if it was either intentionally or negligently responsible for crimes committed by decision-makers or employees.

Following an explanation of dark patterns, the article evaluates three case studies of undertakings and their use of potentially fraudulent dark patterns, namely Amazon's Prime and BuyBox, Google's Search and its Artificial Intelligence (AI) Bard, and Activision Blizzard's (now Microsoft's) lootboxes, and Engagement Optimised Matchmaking (EOMM) algorithms in Section II. Section III. builds on this assessment to determine the applicability of §§ 146, 148 Strafgesetzbuch (StGB). Section IV. provides a brief assessment of whether the Austrian findings may constitute a blueprint for fraud enforcement throughout Europe.

Dark patterns

Dark patterns materialise not by accident, but are a deliberate effort by undertakings to manipulate or hook consumers to bring about behavior that benefit the undertakings. ²⁴ These practices have not only been applied in website and apps but are also being used extensively to dynamically influence video game experiences. ²⁵ While ‘[d]eep learning techniques can be implemented to measure and tailor a game play experience to an individual player's skill level, enjoyment, and mood’, ²⁶ the assumption that video games are designed for entertainment that benefits users, is flawed. In fact, games are riddled with dark patterns, ²⁷ that are designed to either retain players or influence and predict their spending behavior, ²⁸ much like their Big Tech counterparts. ²⁹ The question is, what exactly are dark patterns? ³⁰

Explaining dark patterns

This article adopts the working definition by the Organisation for Economic Co-operation and Development (OECD) which states that:

‘Dark commercial patterns are business practices employing elements of digital choice architecture, in particular in online user interfaces, that subvert or impair consumers autonomy, decision-making or choice. They often deceive, coerce or manipulate consumers and are likely to cause direct or indirect consumer detriment in various ways, though it may be difficult or impossible to measure such detriment in many instances.’ ³¹

The goals of dark patterns can generally be described as:

‘getting consumers to purchase, purchase more of, or continue to purchase, a good or service that they would otherwise not purchase or purchase in lesser quantity; to spend more money on a purchase or time on a service than desired; or to give up more personal data than desired – with the ultimate purpose of increasing business revenue.’ ³²

To achieve these goals, the OECD recognises several categories of dark patterns. Some of these are relevant to the case studies that follow. The most common dark patterns are so-called interface interferences, which can include false hierarchies, preselections, disguised advertisements, ³³ and confirmshaming. ³⁴ Interface interferences generally refer to situations where the user interface is manipulated to privilege certain actions over others, thereby confusing users with limited discoverability to induce actions favorable to the undertaking. ³⁵

False hierarchies visually or interactively prioritise one or more options over others, especially when the items should be parallel rather than hierarchical, ³⁶ thereby conceiving the user to select the option that is in the undertaking's best interest. These are most commonly used in apps and e-commerce websites, ³⁷ as well as in browsers and search engines. ³⁸ Preselections are somewhat related. Instead of a hierarchy that is designed to conceive users into choosing an option that is not in their best interest, the option that is in the undertaking's best interest is already preselected. ³⁹ This may be a purchase or, more prevalently, data collection. ⁴⁰ Disguised advertisements entice users to click on something, such as a download function or a hyperlink to another website that turns out to be an advertisement. ⁴¹ Finally, confirmshaming involves trying ‘to shame the user into making a particular choice’. ⁴² Gray et. al. refer to this as toying with the user's emotions. They make the example of ordering delivery food where the decline option reads ‘No thanks, I'll have a microwave dinner tonight’, ⁴³ thereby emotionally inducing the user to choose the option that benefits the undertaking.

Another common dark pattern is nagging. Nagging is a repeated disruption of the user's expected behavior, where the user is forced to perform additional tasks that are not directly related to their current goal. This can take the form of pop-ups, audio messages, or other actions that disrupt or divert the user's focus. ⁴⁴ The purpose being that the user engages in something beneficial to the undertaking. ⁴⁵

A further instance of a dark pattern is obstruction, which may include making it unnecessarily hard to cancel settings or services, as well as using intermediate currency to obstruct the true costs of digital purchases. ⁴⁶ Gray et. al. ‘define obstruction as impeding a task flow, making an interaction more difficult than it inherently needs to be with the intent to dissuade an action.’ ⁴⁷ Impeding cancellations is a dark pattern that is often used by subscription services to trap users into paying for a service they do not want. For example, a user may be able to sign up for a free trial of a service with just a few clicks, but then have to call customer service to cancel, only to be pressured to stay. ⁴⁸ Intermediate currency involves users spending real money to acquire virtual currency, aiming to alter their perception of value and spending behavior. This dark pattern is often observed in gaming, where in-game purchases are made in a previously acquired virtual currency to create a level of detachment between real-world financial implications and the in-game transactions, prompting altered spending behaviors within the virtual environment. ⁴⁹

Another manifestation of dark patterns is social proofs. These involve activity messages of other users and testimonials. Mathur et. al., explain that the social proof principle rests on the knowledge that individuals observe others’ actions to guide their own behavior. This tendency is exploited by the social proof dark pattern to speed up a person's decision-making or purchases, taking advantage of the bandwagon effect. ⁵⁰ Similarly, unverifiable testimonials seek to deceive consumers based on false or misleading information, for example product reviews. ⁵¹

Urgency is the last group of dark patterns that is relevant for the present purposes. These constitute artificial manifestations of urgency by informing consumers of low stock and high demand, as well as countdowns or limited-time notifications. Urgency dark patterns use imposed deadlines to speed up user decisions and purchases, exploiting users' scarcity bias to make discounts more appealing and signal potential savings loss. When combined with social proof dark patterns, it may cause a “fear of missing out” effect. ⁵²

In discussing the harms caused by dark patterns, the OECD recognises three categories of consumer detriment: privacy harm, psychological detriment and time loss, and financial loss. ⁵³ While fraud concerns causing ‘a financial or other material loss’, ⁵⁴ it is most straightforward to focus on financial loss. Financial loss also constitutes the most frequently recognised adverse impact of dark patterns, ⁵⁵ through ‘direct or indirect acquisition of revenue.’ ⁵⁶ Direct financial loss is caused by conceiving users to spend more than planned. ⁵⁷ Whereas indirect financial loss is caused by:

‘preselection (e.g. a more expensive variant is preselected), urgency-related dark patterns (e.g. the consumer is pressured into buying a product they may not have needed), or confirmshaming (e.g. the consumer is shamed into maintaining a subscription they may not need). For dark patterns such as hidden or hard to cancel subscriptions, the unintended financial expenditure may occur on an ongoing basis and could amount to significantly larger losses than those incurred from one-off purchases.’ ⁵⁸

Most important for the purposes of this article is that, whether direct or indirect, the consumer's financial loss is an immediate consequence of an undertaking conceiving the user with one or several dark patterns. The following section evaluates business models that employ these dark patterns.

Case studies

The purpose of this section is not to discuss whether the undertakings in question use dark patterns, but how they do so, to lay the groundwork for the criminal law application. This will include a short discussion of the relevant dark patterns and their consequences, particularly an illustration of how the deception of users results in a financial loss. However, this section emphasises not only the financial loss to consumers but also the prior aptitude and effectiveness in deceiving consumers about their online choices, as required for the offence of fraud.

Amazon's prime and buybox

The analysis of Amazon's use of dark patterns takes place in two stages. The first level considers dark patterns that Amazon has already implemented and been identified, highlighting the undertaking's frequent use of these practices. Amazon used confirmshaming to get users to sign up for its Prime membership by using ‘No thanks, I don't want Unlimited One-Day Delivery’ ⁵⁹ instead of a decline button. In addition, once subscribed, Amazon made it hard for consumers to cancel their subscriptions, ⁶⁰ instead manipulating them to continue their memberships. ⁶¹ In particular, Amazon manipulated consumers into signing up for its auto-renewing Prime membership when users tried to buy something from Amazon without a subscription, as challenged by the US Federal Trade Commission. ⁶² It made it easy for users to subscribe to the Prime membership through the Amazon Checkout, Amazon Prime Video, and Amazon Music, while requiring users to go through multiple pages riddled with dark patterns such as confirmshaming, obstruction, and toying with emotion when attempting to cancel the subscription. ⁶³ The financial loss that resulted from users being conceived manifested itself in additional unwanted subscriptions or the cost of extended subscriptions.

On the second level, Amazon's use of dark patterns is not limited to its Prime membership. Users are inundated with recommendations on what products to buy when using the Amazon marketplace. These recommendations, however, are not necessarily beneficial to the consumer but are ‘aimed at steering [users] to those options on which [Amazon] earns greater profits.’ ⁶⁴ Specifically, ‘[a]bout three-quarters of the time, Amazon placed its own products and those of companies that pay for its services in that position even when there were substantially cheaper offers available from others.’ ⁶⁵ By allowing users to seamlessly purchase these products through its BuyBox, ⁶⁶ Amazon uses false hierarchies to conceive consumers about pricing and product quality. According to Angwin and Mattu, this led users to overspend by $1400 USD on 250 tested products, with an average price difference of $7.88 USD. ⁶⁷ This process is exacerbated when using Amazon's Alexa to make purchases. When purchasing a product using the digital voice assistant, Amazon typically offers its products as the default. ⁶⁸ This also uses false hierarchies that conceive users in terms of price and quality, causing an immediate direct financial loss, when cheaper products are available, or an immediate indirect financial loss when better quality products are available for similar prices.

Google's search and bard

Much like Amazon, Google is known to use false hierarchies to further its own interests. In Google Search (Shopping), Google manipulated shopping-related search results by favoring its comparison shopping service over that of competitors. The Commission found that this reduced ‘the ability of consumers to access the most relevant comparison shopping services.’ ⁶⁹ Google failed to inform consumers that this ranking was manipulated, i.e., constituted a false hierarchy. ⁷⁰ Much like the discussed dark pattern examples, to put it in the words of Commissioner Vestager, Google ‘denied European consumers a genuine choice of services’. ⁷¹ The General Court concurred with the Commission. As part of the case, it was noted that:

‘Google’s product search results were not determined solely by their relevance for the consumer, as there were commercial considerations underlying the processing of those results. That is at odds with the legitimate expectation on the part of consumers that Google would be neutral in the processing of results. BEUC claims that Google manipulates the search results by making results from competing comparison shopping services invisible.’ ⁷²

A close relative of the false hierarchy dark pattern is preselection. In the case of Google, this was the preselection of Google Search, the Chrome browser, and various other apps in the app interface, as challenged by the European Commission. ⁷³ The General Court concurred with the Commission, finding that by installing its apps and search as defaults, Google exploited users status quo bias that they were unlikely to switch ‘which could not be offset by competitors’. ⁷⁴ Similar to the preselection dark pattern, the option that is in the undertaking’s interest was preselected or preinstalled. Related to both is the Google Search (AdSense) case where Google required website publishers ‘to reserve the most prominent [and profitable] space on their search results pages for a minimum number of search ads from Google’, ⁷⁵ thus influencing consumers through the placement of advertisements.

Each of these cases demonstrates Google's frequent use of practices designed to conceive users for the undertaking's benefit. While the above practices have been identified and prohibited as abuses of dominance under Art. 102 TFEU, given the prominence of dark patterns in search browsers, ⁷⁶ Google's willingness to use them, and the fact that not all of them will be caught by antitrust laws, it is highly likely that users will continue to be conceived by Google's services through dark patterns. Much like Amazon, users suffer financial loss when Google manipulates the search results to promote products or services that are beneficial to itself and may be of a higher price or lower quality.

This is true even after antitrust law or the DMA have attempted to regulate Google's behavior, as evidenced by recent complaints from competing price comparison sites. They suggest, despite DMA obligations, Google is deceiving consumers by misrepresenting hierarchies and preferential treatment of its own or Google-favored search results. ⁷⁷ This is not an isolated problem, but extends to other Google features, such as search boxes, ⁷⁸ highlighting the need for complementary regulation, that can have real teeth.

These concerns are exacerbated by the fact that the use of AI can further obscure the detection of dark patterns, warranting further regulation. ⁷⁹ Consider, for example, when a user enters the search query “What is the best xyz product?” or “Please recommend xyz product to buy for abc purpose.” We know from voice assistants that undertakings will favor themselves when given the opportunity. ⁸⁰ While an AI response to these queries will depend to some extent on the user's query phrasing, it will certainly be shaped by the training data provided, ⁸¹ as well as other parameters that may have been introduced during training or set to purposefully influence the response to produce results more favorable to Google's services. ⁸² While some may argue for a distinction between AI and dark patterns, as manipulations may occur without an undertaking's intent, ⁸³ this leeway has been gambled away based on past behavior, but will certainly play a key role in criminal law proceedings. The financial loss is again the recommendation of lower quality products, or more expensive products, due to Google promoting search results that are beneficial to itself rather than consumers or at the very least neutral.

Activision Blizzard's lootboxes and EOMM algorithms

The use of deceptive designs or dark patterns in video games is common, especially in mobile games. ⁸⁴ This is due to the industry's shift from the one-time sale of games, also known as commodification, to rent-seeking over time, also known as assetization, ⁸⁵ which created a need for publishers to continuously monetise players. However, this also created the risk that players would be continuously targeted, ⁸⁶ paying more for video games through microtransactions than they had previously paid through game purchases. ⁸⁷

This could be assessed in isolation, for example by considering how games are designed to steer users towards in-game purchases, for instance by offering lootboxes only for a limited time employ the dark patterns of scarcity or by notifying/nagging the user to make purchases. ⁸⁸ In the same vein, many games include match-making systems to ensure a competitive and entertaining game. Inter alia, Riot Games' League of Legends' appears to base its Match Making Rating (MMR) primarily on skill, ⁸⁹ suggesting that EOMM algorithms or comparable systems need not (and should not) be designed to deceive players, ⁹⁰ but rather optimise their user experience. EOMM algorithms become problematic when their sole focus is to increase play time or ‘optimize engagement’, which publishers seek to profit from. ⁹¹ In their most problematic form, EOMM algorithms integrate (deceptive) microtransaction strategies. ⁹² Such patents have become increasingly common in the industry. ⁹³

EOMM was originally based on an assessment of gamer’s motivational factors to drive engagement and explicitly disconnected from game revenue. ⁹⁴ An Activision Blizzard patent proposed the arrangement of ‘matches to influence game-related purchases’ ⁹⁵ by either proposing items to the player that he or she previously encountered, ⁹⁶ to ‘target particular players to make game-related purchases’, ⁹⁷ or more pertinently:

‘when a player makes a game-related purchase, the microtransaction engine may encourage future purchases by matching the player […] in a gameplay session that will utilize the game-related purchase. Doing so may enhance a level of enjoyment by the player for the game related purchase, which may encourage future purchases. For example, if the player purchased a particular weapon, the microtransaction engine may match the player in a gameplay session in which the particular weapon is highly effective, giving the player an impression that the particular weapon was a good purchase. This may encourage the player to make future purchases to achieve similar gameplay results.’ ⁹⁸

Other alternatives include matching players that recently made an in-game purchase with slightly weaker players to simulate that the purchase improved their skill level. ⁹⁹

Publishers use a combination of multiple dark patterns to most effectively steer users to make purchases. First, many lootboxes or items purchased through microtransactions are subject to an artificial scarcity (limited items of a certain rarity) or urgency (limited-time notifications or countdowns) deliberately introduced by publishers to accelerate players' purchasing decisions. This is often accompanied by social proof dark patterns. ¹⁰⁰ Because video games function much like social media, providing social interaction, players are constantly exposed to other players' experiences with in-game items (testimonials). Activision Blizzard's patent takes this a step further by matching players who have certain (promoted) items with players they target for further purchases, ¹⁰¹ thereby using social proof dark patterns and creating an environment where players are constantly in fear of missing out. ¹⁰²

Furthermore, when players engage in a microtransaction, they are placed in an environment that makes them feel that their purchase is much more valuable than the item's true value, ¹⁰³ thereby inducing additional purchases later. ¹⁰⁴ This is arguably a form of false hierarchy, where an item is presented as more important than it is. Moreover, these interactions are paired with interface interferences such as confirmshaming, where users receive notifications that they are letting down their real or virtual teammates by not acquiring certain items, or other in-game designs that are effectively disguised advertisements. ¹⁰⁵ All of this takes place in a obstructed environment, in the sense that intermediate (in-game) currencies deceive players about the true cost of their in-game purchases and spending patterns, ¹⁰⁶ in particular in-game gambling. ¹⁰⁷

It is not uncommon for players to purchase items that they do not need to progress in a game, or that are (merely) required to progress faster, and to create artificial hurdles that can only be overcome through purchases. ¹⁰⁸ In the case of lootboxes, this often results in the purchase of useless items instead of useful ones. The deception works either by encouraging the user to make an initial purchase, for example by exposing them to players who already own certain items, or by solidifying their purchasing behaviour and encouraging additional purchases, for example by artificially creating usefulness in a useless item. Because the game environment is designed to dynamically adapt to the player's preferences and purchasing behaviour, ¹⁰⁹ many elements of contemporary games fall into the dark pattern of interface interference that manipulates the player. ¹¹⁰ The financial loss results from the acquisition of useless items, items of inferior quality or usefulness, and the induced continuation of unnecessary purchases.

The following section situates these case studies within the context of Austrian criminal law.

§§ 146, 148 StGB - Fraudulent dark patterns?

Austrian criminal law standardises the basic offence of fraud in § 146 StGB. § 148 StGB includes the qualification of commercial fraud. The offence of fraud states that:

‘[a]ny person who by deceiving another about material facts causes the other person to do, acquiesce, or omit to do an act which causes financial or other material loss to the other person or to a third person and who has the intention to gain an unlawful material benefit for himself, herself, or a third person is liable to imprisonment for up to six months or a fine not exceeding 360 penalty units.’ ¹¹¹

This translates into the objective elements of the offence that include (1) the deception about material facts and (2) an act, acquiescence or omission that causes financial loss. The subjective elements of the offence comprise (1) the intent to deceive (2) and cause loss, as well as (3) the intent to enrich him- or herself or a third party. ¹¹² ‘For a fraud to be completed, it is not necessary that the victim actually recognises the error and feels “aggrieved”.’ ¹¹³

Objective elements of the offence

The objective elements of the offence, also known as the external aspects, rather than the internal state of mind of the perpetrator, concern the actus reus, i.e., the deception about material facts by the perpetrator, and the corpus delicti, in our case the induced act, acquiescence or omission of the user that causes him or her financial loss. Finally, the objective elements include the conditio sine qua non or causality, in the sense that the deception must lead to a relevant error of motive on the part of the user that is wholly or partially causal for the financial loss. ¹¹⁴

Deception about material facts

Deception is a misrepresentation of facts ¹¹⁵ that is intended to mislead another person or to reinforce a false belief. ¹¹⁶ To be relevant in the sense of § 146 StGB, it must have influenced the self-damaging behavior of a deceived person, ¹¹⁷ otherwise it is an irrelevant motive error that does not amount to fraud. ¹¹⁸ ‘The manner in which the deception takes place is in principle irrelevant, provided that the causal connection between deception, error and damaging disposition of assets is given.’ ¹¹⁹ Thus, most important is ‘the adequacy of [the perpetrator’s] actions calculated to deceive in order to achieve the desired (damaging) result.’ ¹²⁰ A sophisticated deception is not required, it suffices if it causes or reinforces ‘a mistake of fact in another person […]. Easy recognizability of the incorrectness of used specifications does not exclude deception.’ ¹²¹ Negligent reliance and lack of attention on the part of the deceived person, ¹²² easily recognisable falsity, ¹²³ or doubt do also not excuse a relevant deception and the action resulting therefrom. ¹²⁴

Deception ‘means any behavior, even if only implied, that has a misleading effect on the perception of another person. The deception must relate to facts, including objectifiable legal circumstances.’ ¹²⁵ This may include ‘the explicit assertion of false facts, […] the distortion of true facts or in conclusive acts intended (and appropriate) to mislead the other person.’ ¹²⁶ It ‘does not require a qualified untruth; any untrue assertion by which the perpetrator is able to arouse the conviction of its correctness in the other person is exemplary.’ ¹²⁷

‘Whether someone is actually deceived and thereby induced to behave in a certain way is not a question of law, but a question of fact.’ ¹²⁸ In order to deceive another, the perpetrator need not explicitly state false facts; it is sufficient if his or her overall behavior ‘is to be regarded as a tacit declaration of a fact, or by omitting the required clarification of certain facts.’ ¹²⁹ In this context, it suffices that the perpetrator ‘consciously and intentionally creates external circumstances which, in the totality of the situation brought about thereby, are likely to give rise to an erroneous opinion in other persons, and that he [or she] thereby aims at the arousal of such an error’. ¹³⁰ While ‘concealment does not already lie in the omission of the reference to the factual inaccuracy of a (also implied) statement’, ¹³¹ a party ‘who deliberately leaves the other party in the dark about the content of a declaration and wishes to derive advantages for his legal position from the latter's error is not supported by reliance on the objective value of the declaration.’ ¹³²

As outlined above, Amazon's and Google's environments are designed to induce users to engage in actions that are beneficial to the undertaking, e.g., self-preferencing and falsified rankings. Similarly, Activision Blizzard's game environments, are designed to dynamically adapt to user behavior to induce spending and falsify product value. These situations constitute relevant deceptions because ‘even the omission of a clarification required by the practice of fair dealing is to be judged as (deceptive) misrepresentation of a factual situation that induces [a party] to act in a way that damages him [ or her], if certain actions tacitly but conclusively guarantee the fact rightly expected by the [other party]’. ¹³³

For example, ‘manipulated matches following a prior manipulation agreement’ ¹³⁴ amount to a deception about material facts because there is an ‘implied declaration that the subject matter […] has not been deliberately manipulated for personal gain’, ¹³⁵ as is the case whenever Amazon and Google self-preference to the detriment of users, as well as when EOMM's predetermine gameplay experiences. The bar is low here, for example, a shuffle game is fraudulent as soon as the cards are shuffled incorrectly, ¹³⁶ or ‘anyone who, when placing a race bet, conceals the fact that he has reduced the betting risk in his favor by bribing race riders is guilty of fraud.’ ¹³⁷

Deception is most often hidden. Users are not told that rankings are being manipulated, or that they are being induced to purchase a product that financially benefits the undertaking, rather than being directed to the lowest priced or qualitatively superior product. Regarding in-game purchases and Activision Blizzard's patent, the relevant deception may occur by placing the user in games where the item purchased is more useful than it actually is, where the game thereby ‘expressly or impliedly – falsely attributes certain value-increasing characteristics to the product’. ¹³⁸ This may lead users to hope that their purchase has been significantly beneficial.

While ‘hopes only constitute facts within the meaning of § 146 StGB insofar as they contain a promise guaranteeing the future expectation or the forecast is based on concrete facts’, ¹³⁹ the hope that the purchased item is useful becomes a false fact as soon as the user is placed in a game environment and is deceived about its true value and induced to subsequently purchase further items. Placement in a manipulated game environment also overcomes the hurdle that ‘false claims about the value of an advertised item without a guarantee of specific value-determining properties are not aimed at causing a mistake of fact’, ¹⁴⁰ because the game, by placing the user in an environment where the items are more useful, assures the user of specific value-determining properties of the purchased item. This could also be argued if a user expects to purchase the most suitable product for a particular purpose and Amazon or Google steers the user to the most profitable sale from their perspective.

The fact that Amazon, Google, and Activision Blizzard use multiple dark patterns to deceive consumers is irrelevant to the criminal evaluation because the relevant deceptive act is the one that is most likely to cause the victim to transfer his or her property, which may mean that there are multiple deceptive acts. ¹⁴¹ What is ultimately of importance is that ‘fraud as a self-inflicted damage offence presupposes that the deceived person carries out the property-damaging act’ ¹⁴² him or herself. In this context, it should be noted that the deceptions discussed may fulfill the elements of the offence of deception (§ 108 StGB), which is ‘conceived as a subsidiary catch-all provision’, ¹⁴³ although this is beyond the scope of this article.

Act, acquiescence, or omission causing financial loss

The relevant acts for the purposes of this article are the purchases made by the consumer or in the case of recurring purchases the omission of cancelling such. In determining whether these lead to a financial loss, and to determine the extent of the financial loss, ‘the actual course of events and the success in its concrete form are decisive, while it does not matter whether there would have been a loss of assets for other reasons if the act had not been committed.’ ¹⁴⁴ Thus, it is not a material factor whether users would have engaged in the purchases absent the deception. As soon as the deception materialises, the purchase was fraudulently induced. Causality exists ‘if the error was at least co-determining for the financial disposition of the deceived person. This is a question to be clarified at the factual level. Causality that is assumed to be proven cannot be problematized by stating that without the error, if the true facts had been known, other motives that had not become effective would have led to the same disposition.’ ¹⁴⁵ ‘However, such a causal connection only exists if the deceived person was still in this error at the time of the damaging financial transaction, but not if he made this transaction with knowledge of the true facts.’ ¹⁴⁶ Based on the case studies discussed above, it is quite probable that the undertaking's conduct, at the very least is co-determining for the financial disposition by at least deceptively exploiting consumer behavioral biases, ¹⁴⁷ fulfilling causality.

A financial loss exists ‘if the financial situation of the victim is less favorable after the crime than before’. ¹⁴⁸ The financial situation or ‘assets are to be understood as the totality of all economically significant and arithmetically ascertainable values.’ ¹⁴⁹ The assessment of damages is a mixed, legal, and factual question that begins with a factual determination of the economic impact of the deceptive act, followed by a determination of whether assets have been damaged and to what extent. ¹⁵⁰ A financial loss need not be permanent, ¹⁵¹ but it must be a direct result of the deception and not an indirect consequential loss. ¹⁵² In case of an exchange ‘the amount of damages results from the difference between performance and consideration, usually the difference between the purchase price and the proceeds achievable for the injured party, i.e. the differential damage.’ ¹⁵³ For the Amazon and Google case studies, the differential damage, as set out in B. 1., is the amount that users overspend due to the deception, the prolonged memberships, or any other losses occurring due to falsified rankings.

Determining damages for lootboxes and microtransactions may be less straightforward. While damages must be based on the market value and include a reasonable share of overhead costs ‘and the usual profit margin [… a]n unreasonably higher purchase price, even though it is binding under civil law, is irrelevant under criminal law.’ ¹⁵⁴ Due to the low marginal cost of items in video games, and the fact that some in-game items that cannot be traded outside of the game have no economic value, ¹⁵⁵ the market value of items can be very low, increasing the amount of damage.

Moreover, for lootboxes one may note that ‘an impairment of assets always occurs when a chance of winning a gambling transaction is thwarted’, ¹⁵⁶ as may be the case when an EOMM algorithm affects lootbox drop chances. If that happens, the EOMM induces a ‘fraudulent sale of worthless goods [that] results in financial loss in the amount of the sales price paid.’ ¹⁵⁷ Moreover, ‘damage in the full amount of the erroneous performance also occurs if the consideration, taking into account victim-related factors, is worthless from an economic point of view’, ¹⁵⁸ as outlined in the paragraph directly above. Hence, if there is no secondary market for the user to offload his or her items, they become worthless from an economic point of view, resulting in damages of the full amount. This also happens ‘if the buyer receives goods for the price he paid that have nothing in common with the goods he […] desired according to his justified expectations, […] and, due to a lack of (reasonable) utilization possibilities, do not increase his assets in any other way […] then there is in fact no consideration and the deceived party suffers damage to the full extent of the price he defrauded.’ ¹⁵⁹ Relating to the justified expectations, ‘only purely arbitrary considerations must be disregarded.’ ¹⁶⁰ For example:

‘The question of deception about the quality of a wine […] cannot be based solely on whether it meets the sensory quality expectations of consumers; rather, criteria other than taste can also play a decisive role in the economic value of a product, such as the expectation of the absence of serious defects in its consistency. Wine adulterated by the addition of diethylene glycol is generally economically worthless, irrespective of its taste quality, due to the lack of consumer interest in the product.’ ¹⁶¹

Based on the case law discussed, it is highly probable that the conduct described in the case studies meets all of the objective elements of fraud. The undertakings intent must also cover these elements.

Subjective elements of the offence

The subjective elements of fraud require intent to deceive, intent to cause (financial) loss and intent to enrich. Austrian criminal law distinguishes three levels of intent in § 5 StGB, which are tiered subject to the intensity of the intent. The highest tier is purpose or dolus directus specialis, found in § 5 (2) StGB, which is dominated by the will component of the offence in the sense that the perpetrator aims and intends to realise all elements of the offence. ¹⁶² The second tier, knowledge or dolus principalis, found in § 5 (3) StGB, requires that the perpetrator does not aim to realise all elements of the offence, but in the realisation of another, even legal goal, he or she considers the realisation of the offence certain. ¹⁶³ The lowest tier of intent, conditional intent or dolus eventualis, found in § 5 (1) s. 2 StGB, requires that the perpetrator neither intends to realise the offence, nor considers it certain, but seriously considers it and proceeds anyway. ¹⁶⁴ The offence of fraud merely requires the perpetrator to act with dolus eventualis for all three elements of the offence, ¹⁶⁵ at the time the deception occurs. ¹⁶⁶

Dolus eventualis involves a knowledge component, namely, that the perpetrator considered that the realisation of the offence was possible. This alone is not sufficient for a finding of dolus eventualis but could also constitute deliberately negligent conduct. ¹⁶⁷ The differentiator is the will component. Namely, ‘if the perpetrator nevertheless decides to commit the offence because he is willing to accept the course of events that constitutes the offence […] then he is acting with conditional intent’. ¹⁶⁸ Not recognising or wrongly assessing the risk, as well as recklessly trusting in the non-occurrence of success amounts to deliberate negligence. ¹⁶⁹

For dolus eventualis to extend to causation, it is not enough that the perpetrator ought to have known or was reckless or careless. ¹⁷⁰ In this sense, a conscious acceptance by the perpetrator just barely ‘covers both the knowledge component and the will component of conditional intent.’ ¹⁷¹ However, ‘the mere recognisability of a factual occurrence that corresponds to a legal offence is not sufficient to assume an act with - even conditional - intent.’ ¹⁷² Similarly, mere ‘indifference in the sense of an inner apathy’, ¹⁷³ or false hope and thoughtlessness do not suffice for intent. ¹⁷⁴ But for intent to extent to ‘the essential characteristics of the offence […] it is not necessary for the offender to make a legally accurate assessment of the normative terms that require special interpretation. Rather, it is sufficient that he recognises the social meaning of this term and in this way becomes aware of the specific lack of value of the violation of legal interests, at least in layman's terms.’ ¹⁷⁵

‘Conditional intent is to be assumed if the perpetrator seriously considers the realization of a factual situation corresponding to the legal description of the offence and actually within the realm of possibility to be possible and accepts it, even if out of deliberate indifference’. ¹⁷⁶ This follows from the notion that ‘the will to realize a fact necessarily presupposes that the perpetrator imagines this fact’, ¹⁷⁷ or regards its realisation as obvious. ¹⁷⁸ However, ‘it is not necessary for the perpetrator to inwardly “affirm” or “approve” the realization of the offence in order to assume conditional intent.’ ¹⁷⁹ But such ‘a tacit acceptance of what is (seriously) considered possible is a plus in terms of [willpower/]intent compared to mere acceptance of the outcome’. ¹⁸⁰ Moreover, if the perpetrator acts knowingly, this makes the consideration of conditional intent superfluous and fulfills the subjective elements. ¹⁸¹ In relation thereto, if the perpetrator aims to realise a certain act, such as the deception of consumers, one could even assume the existence of dolus directus specialis. ¹⁸² For the purpose of the following assessment, it is assumed that there are relevant persons in the undertakings concerned who make the decision regarding the implementation of the dark patterns discussed. However, since the author does not know whether it is a programmer, manager, or psychologist involved in the design of these dark patterns who makes the final decision, intent is discussed in the abstract.

Intent to deceive

The intent to deceive is relatively straightforward. Amazon, Google, and Activision Blizzard falsify rankings or situations to exploit consumers behavioral biases and deceive them about the true value or quality of a product. Here, ‘the decisive factor is […] the perpetrator's awareness of causing the deceived person to be mistaken about the facts through his conduct’. ¹⁸³ It is not implausible that the relevant persons in these undertakings at the very least recognise the social meaning of their actions, ¹⁸⁴ or are deliberately indifferent to the realisation of the deception. ¹⁸⁵ In the case of Activision Blizzard's conduct, which was codified in patents prior to their execution, it is likely very plausible that actions are taken with knowledge of the deception. ¹⁸⁶

Intent to cause (financial) loss

With respect to the intent to cause a financial loss, ‘the perpetrator must (insofar) be aware that by creating (or reinforcing) the error, he is causing the deceived person […] to dispose of property and thereby cause a direct loss of property’. ¹⁸⁷ For Amazon and Google, this is more straightforward, as it can be inferred from their conduct that they do not intend to let the user buy any product he or she wants, but rather the product that is most advantageous to them, or in the case of Amazon, the subscription period. Therefore, it is not implausible that they are at least deliberately indifferent to causing financial loss to the user by steering the user to products that are either more expensive or of inferior quality.

In the case of Activision Blizzard, ‘the fact that the [in-game] item is not individually useful to the victim must be included in the perpetrator's intent.’ ¹⁸⁸ This is relatively easy because the programmers determine the usefulness, or lack thereof, for each item. For example, when designing a lootbox, programmers explicitly determine the likelihood of useless or worthless items. Similarly, when designing a game environment, the programmers determine the usefulness of each item. While ‘the intention to demand a certain amount of money from the deceived person [for an item] does not in itself allow the conclusion to be drawn that the perpetrator considers the success of his demand to be obvious’, ¹⁸⁹ the level of in-game steering and deception applied, as exemplified by Activision Blizzard's patent, does allow the interference that they consider success to be likely or are at least deliberately indifferent to it. This is especially true given that the game environment is adjusted based on user behavior and data to achieve said success, as discussed above.

Intent to enrich

‘The permissible economic pursuit of profit by an entrepreneur cannot simply be equated with the intention to unlawfully enrich oneself.’ ¹⁹⁰ Intent to enrich is present if one ‘through the conduct of the deceived person, wants to unlawfully and at least temporarily increase his economic assets […] by the value of the assets lured out.’ ¹⁹¹ Unlawful refers to an ‘increase in assets without a reason approved by the legal system’. ¹⁹² Enrichment is ‘only relevant […] on the internal side of the offence. It is not essential that an objective enrichment occurs’. ¹⁹³ The intention to enrich must be ‘present at the time of the deceptive act’. ¹⁹⁴

Enrichment is relatively straightforward in the sense that the purchase price or extended subscription, is either all or part of the enrichment, depending on the consideration provided. ¹⁹⁵ Whether the enrichment is unlawful depends on whether ‘the perpetrator has no claim to it or believes he [or she] has no claim to it’. ¹⁹⁶ The fact that the deceived party is given a compensable consideration, such as a worthless or more expensive but qualitatively inferior product, does not exclude the intention to unlawfully enrich, but ‘requires a recognizable intention to offset at the time of the offence.’ ¹⁹⁷ While this is not a simple inference, one may ask why the undertakings in question would choose to manipulate and deceive consumers, if they believe they have a legitimate claim to the consumers' assets.

Qualification – Commercial fraud § 148 StGB

The offence of commercial fraud states that:

‘[a]ny person who commits a fraud commercially is liable for imprisonment for up to three years; any person who commits an aggravated fraud under § 147 para. 1 to 2 commercially is liable to imprisonment for six months to five years.’ ¹⁹⁸

Based on § 70 (1) StGB:

‘a person commits an offence commercially, if the person commits the offence for the purpose of obtaining a sustained, more than negligible income for the longer term through the repeated commission of the offence, and […] 1. the person employs specific skills or means in the commission of the offence which suggest that the offence will be committed repeatedly’. ¹⁹⁹

A more than negligible or ‘not merely insignificant continuous income is one that exceeds the amount of EUR 400 per month on an annual average.’ ²⁰⁰ For this to be recurring and continuous it does ‘not mean unlimited in time, but only the extension over a longer period of time’, ²⁰¹ e.g., three months. In this context, ‘commercial activity is always to be assumed if the perpetrator's intention is to obtain a [...] regular source of income by repeating the [same] criminal offence several times, i.e. to make a living from it, at least in part.’ ²⁰² For example, by steering several users to disadvantageous purchases or subscriptions, or users to disadvantageous item purchases.

On the subjective side, this requires purpose or dolus directus specialis, (§ 5 (2) StGB) for a ‘a regular or long-term (and not merely one-off) source of income in the sense of a recurring inflow of funds by repeating the offence several times.’ ²⁰³ If the offence is committed on several occasions, ‘it is not necessary to demonstrate commercial intent for each offence.’ ²⁰⁴ However, ‘remuneration and the intention to repeat the activity alone are not sufficient for commercial activity.’ ²⁰⁵ The relevant intention must be accompanied by one of the conditions listed in § 70 (1) StGB. ²⁰⁶

In the cases at hand, it is most likely that § 70 (1) s. 1 StGB, the use of special skills or means, which are ‘alternative forms of commission’, ²⁰⁷ is relevant. These indicate ‘a recurring offence if they testify to the offender's professionalism. They are “special” if their mastery or possession is unusual in the situation and can be explained by the practiced or well-considered approach of the perpetrator’, ²⁰⁸ and ‘suggest recurring commission if they are so elaborate that their use for only one offence appears unreasonable.’ ²⁰⁹ Similarly, intention in the context of commercial activity is primarily a question of fact, taking into account the perpetrator's conduct before, during and after the act. ²¹⁰ For example, ‘an extremely professional approach [is] sufficient grounds for assuming commercial activity’. ²¹¹

For all the undertakings in question, this is likely to be fulfilled by the design and programming of either ranking algorithms and EOMM algorithms, as well as the digital platforms or environments created to steer users (and the efforts made to hide these – dark – patterns from users), a practice that constitutes both a professional approach relevant to the intention, as well as the condition in § 70 (1) s.1 StGB. Thus, it is quite plausible that the discussed conduct of Amazon, Google, and Activision Blizzard meets the requirements for fraud and commercial fraud in the context of §§ 146, 148 StGB in connection with § 70 StGB, if the subjective elements of the offences are met as discussed.

Moreover, interesting for the argument of complementary regulation with teeth is the fact that ‘not only the direct perpetrator is responsible for the qualification of commercial commission of the offence, but also every participant in the offence who participates in a commercial fraud committed by another person with the intention of making a continuous profit for his own benefit, which can also apply to the contributory offender’. ²¹²

For the sake of completeness, it ought to be noted that the conduct in question may also qualify as aggravated fraud under § 147 (3) StGB, which states that ‘[a]ny person who brings about damages exceeding 300 000 Euro through the offence is liable to imprisonment for one to 10 years’, ²¹³ however, as with deception (§ 108 StGB) this is beyond the scope of this article.

Austrian mayfly or European replicability?

The potential for using criminal law to tackle the issue of dark patterns extends beyond Austria and could be applied throughout the European Union. While there is no harmonised European standard for (commercial) fraud, ²¹⁴ there are increasing efforts to harmonise or Europeanise criminal law enforcement. ²¹⁵ A broad comparison of the Austrian fraud provisions with fraud provisions across Europe suggests that the Austrian perspective outlined above could serve as a blueprint for dark patterns to constitute the offence of fraud. This finding is based on the relative closeness of European fraud provisions and the fact that fulfilling the offence of fraud may be considered more stringent in Germanic countries than in other European countries, leading to the premise that if the practices in question are considered fraudulent in Austria, this is likely to be the case in other European jurisdictions as well.

By way of illustration, in their comparative assessment of fraud in the Germanic and Nordic countries, Husa and Tapani found that ‘the Nordic legal family is closer to Continental law than to common law’, ²¹⁶ and in particular, that the structure of the law of fraud in the Germanic and Nordic countries is ‘pretty much identical.’ ²¹⁷ This concerns the three general elements of fraud, ²¹⁸ for which Husa and Tapani found that the jurisdictions under discussion consider two conceptions to be the most important. Namely, the ‘victims’ state of mind […] i.e. the deception must be an operative cause allowing a defendant to obtain the property’ ²¹⁹ and a protected legal interest, i.e. the property in question.

The most important, what they call the micro-difference, is rather the nature of legal thinking, such as more extensive codification in Germany, ²²⁰ as well the fact that Germany has a somewhat more restrictive interpretation of fraud than Switzerland, Austria and the Nordic countries. ²²¹ Nevertheless, the overall picture that emerges from Husa and Tapani’s work is that if an act is considered criminal fraud in Austria, it is likely to be so in other Germanic and Nordic countries. ²²²

This echoes to some extent the findings of Ramage in the context of commercial fraud, who compares Germany and France with the United Kingdom and finds, in a simplified manner, that the UK regime, due to its broader definitions is more encompassing than Germany, with France being the most limited jurisdiction of the three. ²²³ In another comparative discussion, Poniatowski and Wala find that the offence of fraud in the Baltic and Eastern European countries is based on similar foundational elements, mainly deception and the intent for unlawful gain but that Germany and Switzerland have more detailed provisions on fraud. ²²⁴ In other words, if the fairly similar Germanic countries recognise that an act constitutes fraud, the Baltic and Eastern European countries are likely to do the same.

While this section does not attempt to provide an exhaustive comparison of all European jurisdictions, in the larger European scheme, the Germanic countries, and thus the Austrian §§ 146 and 148 StGB, represent a more stringent regime than most of their European counterparts. It may therefore be fair to say that if dark patterns are caught by the Austrian fraud provisions, they are likely to be caught by their European counterparts as well, and may at the very least inform the European debate in these jurisdictions.

Conclusion

Antitrust law and digital regulation have arguably made toothless attempts to regulate dark patterns. On the one hand, this may be in part because we are only beginning to understand the full extent of the impact of dark patterns on users and, more importantly, on their purchasing behavior. On the other hand, it may be the case that (digital) undertakings have become accustomed to being regulated with fines and see them as a necessary cost of doing business and acquiring and locking customers in. Thus, if these regulatory attempts do not have the desired success, complementary criminal law regulatory measures may provide the teeth that digital regulation has so far lacked.

This article has used Amazon's Prime and BuyBox, Google's Search and its AI Bard, and Activision Blizzard's lootboxes and EOMM algorithms as examples of how they may constitute or use dark patterns to deceive consumers. In the context of Austrian criminal law, these dark patterns likely satisfy the objective elements of the offence. False rankings, deceiving environments and manipulated situations are used to deceive consumers about material facts of the purchases they are induced to make; such as the quality of the products or whether they are overspending or purchasing effectively worthless (in-game) items. Proving the subjective side of the offence is a more difficult task. However, since the offence of fraud (§ 146 StGB) under Austrian criminal law ‘only’ requires dolus eventualis, and this is satisfied by the perpetrator's deliberate indifference combined with the social understanding of the harmful nature of his or her conduct, it is not implausible that this conditional intent is present. Similarly, since the intent (dolus directus specialis) required for the qualification of commercial fraud (§ 148 StGB) ‘only’ has to extend to the recurring commission of the offence and to the fact that it represents a continuous source of income, ²²⁵ in conjunction with the (deceptive algorithmic) professionalism of the undertakings in question, which speaks to the special skills and means used, as well as to the professional nature of the (fraudulent) commercial activity, it is again not implausible that the elements of § 148 StGB are fulfilled. Moreover, it is quite possible that the Austrian blueprint may serve useful for other European jurisdictions that have comparable fraud provisions.

While it should not be our goal to use criminal law to regulate undertakings, neither should undertakings' compliance with digital regulation be stretched to the point where criminal law enforcement becomes necessary for other regulatory measures to be taken seriously.