A Kantian response to the Gamer’s Dilemma

The Gamer’s Dilemma consists of three intuitively plausible but conflicting assertions: (i) Virtual murder is morally permissible. (ii) Virtual child molestation is morally forbidden. (iii) There is no relevant moral difference between virtual murder and virtual child molestation in computer games. Numerous attempts to resolve (or dissolve) the Gamer’s Dilemma line the field of computer game ethics. Mostly, the phenomenon is approached using expressivist argumentation: Reprehensible virtual actions express something immoral in their performance but are not immoral by themselves. Consequentialists, on the other hand, claim that the immorality of virtual actions arises from their harmful consequences. I argue that both approaches have serious difficulties meeting the moral challenge posed by the Gamer’s Dilemma. They tend to confuse the morality of in-game actions either with the morality of their real-world counterparts or with the morality of games as objects. Following this critical analysis, I will develop a Kantian argument and defend it against two objections. So far, deontological responses to the Gamer’s Dilemma have been sought in vain. Yet, with Kant, its moral challenge can be met by looking at the gamer’s reasons. From this perspective, the Gamer’s Dilemma is based on a false assumption: the moral status of gaming acts does not derive from a normative equation with their real-world counterparts but only from their justifications.


Introduction
The Gamer's Dilemma was introduced into the philosophical debate by Morgan Luck (2009) and consists of three intuitively plausible but conflicting assertions (Luck, 2018, 157): i. Virtual murder 1 is morally permissible. ii. Virtual child molestation is morally forbidden.
iii. There is no relevant moral difference between virtual murder and virtual child molestation in computer games.
The third premise results from the observation that common arguments regarding the morality of gaming typically either excuse or condemn both in-game murder and in-game child molestation. The position of the 'Ludic amoralist' who insists that 'It's just a game!' (Ostritsch, 2017;Patridge, 2011) excuses all virtual crimes. In contrast, arguments that place virtual crimes in close moral relation to their realworld counterparts ultimately condemn all of them. As a consequence, gamers have to choose: "Either they acknowledge that acts of virtual murder and virtual paedophilia are morally prohibited, or they acknowledge that both are morally permissible" (Luck, 2009, 35).
Numerous attempts to resolve or dissolve the Gamer's Dilemma line the field of computer game ethics, but a decidedly deontological response is sought in vain. I will develop the latter in this paper and show its advantages over other approaches. In the first part, I will reformulate the philosophical challenge of the Gamer's Dilemma and argue that Footnote 1 (continued) consequentialist and expressivist approaches have serious difficulties meeting it. In the second part, I will develop a deontological response that is based on Kantian ethics and meets the philosophical challenge of the Gamer's Dilemma. In the last part, I will address two objections that arise in view of the Kantian response.

The moral challenge of the Gamer's Dilemma 2
Before a response to the Gamer's Dilemma can be elaborated, the philosophical challenge it poses must be completely clear. The first two premises draw a moral distinction between two virtual crimes that is intuitively accepted by the majority of gamers and non-gamers. The third premise denies a moral-theoretical foundation for this distinction by claiming that there is no moral difference between virtual murder and virtual child molestation. Although this is not a moral statement, it refers to a moral differentiation that has practical implications (if there is no moral difference between virtual murder and virtual child molestation, gamers face the Gamer's Dilemma). The philosophical challenge of the Gamer's Dilemma is therefore inherently a moral challenge. It asks about the morality of certain virtual actions in computer games and, closely related, about the moral differences among these virtual actions. 3 In what follows, I will discuss two lines of argumentation that currently dominate the debate: the consequentialist and the expressivist approach. The systematic compilation neither claims to be exhaustive nor to represent the full potential of individual approaches but rather focuses on the general difficulties they tend to face in adequately meeting the moral challenge of the Gamer's Dilemma. 4

The consequentialist approach
The consequentialist states that consequences of virtual child molestation are more harmful than consequences of virtual murder. In its more convincing version, the argument emphasizes the role of habituation effects that could result in the brutalization of gamers (McCormick, 2001;Waddington, 2007), rather than claiming that virtual crimes encourage the commission of real crimes and turn players into murderers or abusers. This line of thought has recently been reinforced by C. Thi Nguyen's considerations on the phenomenon of "value capture" (2000,200). According to Nguyen, the general danger of games is posed by offering pleasantly simplified modes of agency that players might adopt for reality (Nguyen, 2020, 202-203): In value capture […] the simplified value takes over as the primary guide in my practical reasoning.
[…] The worry here is not that I can be incentivized in counterproductive directions, but that my values are transformed by the seductive clarity of simplified values.
[…] The pleasures of games can give us a motivation to simplify our values in potentially problematic ways.
This danger could be considered higher for virtual child molestation than for virtual murder because, in comparison, it represents a clearer (reprehensible) value logic, for example by its autotelic structure (Kjeldgaard-Christiansen, 2020), by discriminating against a certain social groupwomen (Bartel, 2012) or children (Patridge, 2013) -, or by endorsing a specific "morally problematic worldview" (Ostritsch, 2017, 117). All this does not apply to typical cases of virtual murder (of course, exceptions can be constructed).
Although the consequentialist approach is discussed in research (Luck, 2009;Montefiore & Formosa, 2022), it receives comparatively little attention and is usually dismissed quickly. Mostly, this is justified by the lack of empirical data. Studies comparing the effects of virtual murder with the effects of virtual child molestation do not exist. Moreover, significant long-term effects on the gamer's character from specific in-game actions are unlikely (Ferguson, 2 I thank two anonymous reviewers for their valuable comments, from which this chapter in particular has benefited substantially. 3 Despite a variety of recent attempts to narrow the dilemma down (Luck 2022;Montefiore & Formosa 2022) or to split it up (Ali 2022;Kjeldgaard-Christiansen 2020), I turn to the Gamer's Dilemma in its original form, which has lost none of its provocative power. Narrowing it down to particularly explosive cases is not necessary, since a full response to the Gamer's Dilemma should not only explain clearcut cases but rather resolve the moral conundrum for all conceivable cases. After all, clear-cut cases pose no dilemma. At the same time, it is important to note that the Gamer's Dilemma by itself already provides a very narrow context: It is about virtual actions within computer games that do not affect other people. Thus, the focus is neither on non-gaming virtual spaces nor on multiplayer games. Accordingly, while different normative aspects may play a role in it, the Gamer's Dilemma is one dilemma that can be met with one response. Recently, Ali (2022) has questioned this, turning one dilemma into three, depending on whether we are talking about virtual or gaming actions, but he himself admits in footnote 10: "It may be that there is a more general resolution that addresses both types of acts, for instance, focusing on acts that are 'not real'" (Ali 2022). This is what I am aiming for. Essentially, the Gamer's Dilemma is about murders and child molestations that are, in a sense, not real murders and child molestations (see also Davnall 2021, 225). 4 Accordingly, it is not about an exegesis of the debate so far, but about a problem-oriented presentation of general difficulties associated with the mentioned lines of argumentation and their basic assumptions. Of the specific texts and authors referred to, some cope well, others less well, with the highlighted challenges. Unfortunately, the limited scope of this paper does not allow for extensive discussion of the specific strengths and weaknesses of individual approaches. 2015; Reinecke & Klein, 2015). Of course, pathological gaming (e.g., due to gaming addiction) may cause significant long-term damage to those affected, and some virtual actions may increase aggressiveness in the short term. But the Gamer's Dilemma is not about harms of pathological behavior or short-term mood swings but about specific and weighty moral dangers of certain in-game actions. These are neither empirically proven nor very likely. With these findings, no difference can be found between virtual murder and virtual child molestation. Actually, those findings seem to imply that both virtual crimes are equally morally unproblematic because any in-game crime does not appear to result in weighty harmful consequences. This is the position of the 'Ludic Amoralist' (Ostritsch, 2017), who rejects premise (ii) of the Gamer's Dilemma and thus seems to provide a valid answer to its moral challenge. But this answer is incomplete. It cannot explain our different intuitions underlying the Gamer's Dilemma, but rather denies that there is any difference. However, empirical research suggests that there are different intuitions (Formosa et al., 2023), which leads to a reversal of the burden of proof: Apparently, the Gamer's Dilemma poses a major challenge to the Ludic Amoralist, not the other way around.
There is another, more serious difficulty for consequentialist argumentation: The assumption of the brutalization of gamers presupposes that habituation effects extend beyond gaming situations. As stated by Sicart (2009), Schulzke (2010, and in part by Nguyen (2020, 189-215), gamers might acquire certain character traits while performing ingame actions, which then could become practical effective in comparable real-world situations. However, this argument overestimates the normative analogy between in-game and game-external situations. Carrying out virtual murder or virtual child molestation requires completely different competences, knowledge, desires, and circumstances than carrying out their real-world counterparts. The conditions for agency are different in gaming than in reality: neither the same values nor the same actions are at stake (Ulbricht, 2022, 62-68). A consequentialist argument would have to go a long way explaining why (and which) in-game values and situations can be applied to game-external contexts (and how they differ in cases of virtual murder and virtual child molestation).

The expressivist approach
The philosophical debate about the Gamer's Dilemma is dominated by another approach: expressivist argumentation. The basic line is simple: Some virtual crimes express something reprehensible, others do not. Variants of this thought differ in what the 'expressed thing' exactly is. Roughly, four arguments can be distinguished (mixed forms also occur): The reprehensible thing that is expressed when committing a (reprehensible) virtual crime consists in… a. …the reprehensible character of the player. b. …the reprehensible game. c. …its reprehensible real-world counterpart. d. …a reprehensible statement.
With regard to the Gamer's Dilemma, the expressivist states that virtual child molestation expresses something reprehensible more often (Ali, 2015;Ramirez, 2020) or exclusively (Bartel, 2012;Patridge, 2013) compared to virtual murder. Thus, in contrast to virtual murder, (a) a vicious desire of the player shows up (more often) when performing virtual acts of child molestation, (b) games that include child molestation are (more often) reprehensible in contrast to games that include (solely) murder, (c) virtual child molestation (usually) adopts significantly more (im) moral properties of its real-world counterpart than virtual murder, or (d) virtual child molestation (usually) sends a reprehensible message, unlike virtual murder.
Expressivist theories are initially convincing. By locating morality outside of play, expressivists satisfy the intuition that play is intrinsically amoral. Nevertheless, they face significant argumentative hurdles, which are not always taken seriously enough. One fundamental difficulty for any expressivist approach is to go beyond a descriptive explanation of moral intuitions and provide normative justification (in a strong, moral sense) for why certain virtual crimes are reprehensible. As said, the expressivist strategy is to show that certain in-game acts express reprehensible things: Play is amoral but may stand as a symbol for another thing that is immoral. However, this seems to imply that only the thing expressed is reprehensible-not play itself as mere expression. Expressivists thus face the challenge of establishing a robust moral dependency between expressive actions and expressed reprehensibility that goes beyond intuitional connection.
In addition, each version of the expressivist theory struggles with its own more specific challenges. Regarding the focus on the gamer's character (a), it is unclear why gamers should express attitudes towards ordinary crimes while performing in-game crimes. After all, gamers are primarily dealing with virtual actions, not with their real-world counterparts. Why should my desire to murder virtually reveal anything about my disposition toward real murder? Some authors state that the reprehensible attitude of the player is (only) expressed when playing reprehensible games (Ostritsch, 2017;Patridge, 2011). However, this argument tends to confuse the morality of gaming with the morality of games: player or game may be immoral here, but not the virtual crime as such (which is the core of the Gamer's Dilemma).
If one focuses on games as moral objects (b), one faces a similar difficulty in showing that a normative connection exists between the morality of games or in-game contexts (Ali, 2015;Nader, 2020) and the morality of virtual actions in those games. A direct moral connection seems hard to justify since we can easily imagine morally acceptable actions in immoral games (say, for research purposes) and, vice versa, morally problematic actions in morally acceptable games (say, humiliating other players). In determining the morality of virtual actions, the gamer's motivations seem to play a much more decisive role than the morality of the game, in contrast to what Rami Ali states (2015, 269-270): [I]f virtual acts depend wholly on the gamer's context, then any in-game act will turn out impermissible or permissible depending on the gamer's intention in the performance. The morality of virtual acts will turn on whether the gamer engages with these acts in a morally perverse manner or not, and not on the type of act performed (whether virtual murder or virtual pedophilia). In this sense, depending wholly on the gamer's context trivializes the dilemma.
Prima facie, it is the other way around: Not relying on the 'gamer's context' (i.e., her motivations) means not speaking about the morality of actions and thus not addressing the moral challenge of the Gamer's Dilemma. Of course, one could then inquire if immoral gaming is (in)appropriate for games and their in-game contexts-but this is a different question that is not posed by the Gamer's Dilemma.
The third approach (c) is often mixed with other lines of argumentation. Descriptions of moral dangers or shortcomings of virtual crimes regularly focus not on virtual actions as such but on their real-world counterparts. However, it is unclear why and how moral properties of ordinary actions should be transferred to their virtual equivalents. One proposal is given by Ramirez (2020, 145): A player commits an act of virtual murder in those cases where she directs her character to kill another and in which her decision affects her psychologically, physiologically, and behaviorally in the same way that a real decision to commit murder would.
[…] Insofar as virtual murders are to be defined in terms of their connection to actual murder, PBVM [Psychological and Behavioral Virtual Murder, S.U.] aims to capture the features of a virtual murder that include these psychological and behavioral elements of murder (real or virtual) about which we might worry.
There are two problems with this argument. First, Ramirez' definition is too exclusive. It is empirically questionable wether such cases of virtual murder actually occur. Typically, virtual murder is 'physiologically and behaviorally' quite different from real murder; just look at the different body movements (you normally do not murder by pushing buttons on gaming controllers). Comparable psychological states among gamers and murderers seem even rarer. Second, it is not clear why virtual murder should be treated morally similar to real murder if it is a normatively incomplete copy. Prima facie, the essential moral issue of murder is the intentional taking of a life without consent, not the psychology or movement of the murderer. Why should there be a relevant moral analogy between virtual and real murder if the former lacks this central moral property?
In many cases, moral analogy is implicitly assumed. For example, Luck states that "child molestation is grave enough that, by engaging with it[s] representation in a carefree or light-hearted manner, we treat it too lightly-whilst the same is not true of murder" (Luck, 2022(Luck, , 1306. If we assume that there is indeed a difference of 'graveness' between murder and child molestation (which could be doubted), this might have moral implications for our handling of such crimes but not for our handling of virtual crimes. Luck's considerations may concern game developers or games that potentially aim at representing (real) crimes, but not gamers, who typically are not dealing with murder and child molestation but with virtual murder and virtual child molestation.
While it is likely that virtual actions share some aspects with their real-world counterparts (otherwise, it would be a mystery why we use designations of real actions to describe virtual actions), a separate argument must be made for moral dependency, which cannot be presumed when talking about the morality of gaming. It has to become clear why a moral derivation relation should exist despite the high number of normatively relevant differences (in body movement, context, intention, people affected, etc.).
The last approach (d) ascribes a representational role to virtual crimes, such that by performing them, factual statements are made that may be reprehensible-for example, because these statements discriminate against women (Bartel, 2012) or children (Patridge, 2013). However, it is unclear how virtual crimes should make such statements. Making a statement is an action. Who performs this action? The virtual action itself can hardly be the agent. Instead, players, game developers, or games come into question. If one assumes that players send immoral messages by performing certain virtual crimes, one encounters the challenges related to the first approach (a) discussed earlier. Furthermore, the gameplay as such does not seem to be a sufficient indicator to derive statements from the player. Considering the second option, the connection between performed action and game developers is very unclear: How can one person (the developer) make a statement through the actions of another person (the player)? How do we even determine the developers' intention? The first source for this is the game itself. So, is it simply a statement of the game that is expressed when playing it? With this reasoning, we run into the difficulties of approach (b): It is hard to justify why a player's action should necessarily express a game's statement.
For the sake of the argument, let us assume that virtual actions can make statements for themselves. What could such a statement be about? Prima facie, its subject is the performed action. But why should an affirmative statement about a virtual crime be reprehensible? The answer cannot be "Because virtual crimes are reprehensible!" because then what is to be shown is presupposed. If the answer is "Because an affirmative statement about real crimes is expressed when performing virtual crimes!", then we end up with the challenges of approach (c).

Two adequacy conditions to meet the moral challenge of the Gamer's Dilemma
The difficulties presented in completely and adequately meeting the moral challenge of the Gamer's Dilemma can be summarized by two adequacy conditions. First, the response should provide a moral-theoretical argument and not remain in the realm of Descriptive Ethics. The Gamer's Dilemma poses a moral-theoretical puzzle that has to be solved by moral-theoretical reasoning. A complete response both explains the source of our moral intuitions and justifies why these intuitions are (not) morally reasonable. 5 Second, an adequate response to the Gamer's Dilemma should do justice to the subject matter: virtual crimes. Accordingly, one has to deal with actions (play) rather than objects (games), and with virtual crimes rather than real crimes. While it is true that game and gaming, as well as virtual crimes and real crimes, may be normatively related, this tells us nothing about the moral status of in-game actions as such. Moreover, as we have seen, neither play and game nor virtual crimes and real crimes seem to share a very close moral bond. First, one can perform morally acceptable acts in reprehensible games and reprehensible acts in morally acceptable games. Second, virtual crimes and real crimes differ so much in terms of context and execution that the morality of one can hardly tell us anything about the morality of the other. What makes real murder immoral is not what makes in-game murder (potentially) immoral-no one's life is taken violently and willingly. In sum, the adequacy conditions for a complete and adequate response to the Gamer's Dilemma call for the following: 1. To deliver a complete analysis, that is… a. …to provide a descriptive explanation of our intuitions and… b. …to provide a moral-theoretical reflection of their basis.
2. To do justice to the subject matter, that is… a. …not to confuse in-game actions with games as objects and… b. …not to confuse in-game actions with ordinary actions.

A Kantian response
Consequentialist and expressivist considerations are not flawed as such. Understood descriptively, they provide promising explanations for our intuitions and thus easily fulfill adequacy condition (1a). They also point to some relevant normative characteristics of the Gamer's Dilemma, and I do not categorically rule out the possibility that consequentialists or expressivists could, with careful reasoning, satisfy all adequacy conditions. Nevertheless, grasping the moral core of what might be reprehensible about virtual crimes as such remains a fundamental challenge for both approaches. In light of this, it is surprising that a genuine deontological approach has hardly been pursued so far. 6 Such an approach is attractive because it does not need to make empirical assumptions (e.g., about brutalization) while still focusing decidedly on the moral status of actions. Thus, adequacy conditions (1b) and (2a) are satisfied from the outset. However, conditions (1a) and (2b) remain challenging: to explain our different intuitions regarding virtual crimes and to do justice to the specificity of play.

Are virtual crimes morally forbidden?
Imagine the following: Anton plays a computer game. He enters a store (in-game), murders the shopkeeper, and then abuses his child. 5 This condition does not apply to approaches that decidedly commit to a descriptive explanation, such as the excellent paper by Jens Kjeldgaard-Christiansen (2020), which provides illuminating insights into the moral psychological background of the Gamer's Dilemma, or the proposal by Thomas Montefiore and Paul Formosa (2022), who suggest "that the robustness of the intuitive difference may indeed have a normative basis, but that basis may lie in aesthetic and conventional norms". These papers are enlightening in their own sense but are not suited to solve the moral-theoretical puzzle posed by the Gamer's Dilemma. I also include Garry Young's approach (2016) in this line of argument. He describes socially constructed norms that ground our moral intuitions regarding the Gamer's Dilemma. Therefore, Young's remarks remain in the realm of Descriptive Ethics: Young's 'Constructive Ecumenical Expressivism' may explain social conventions and moral intuitions, but it cannot justify moral norms in a strong sense (Ostritsch & Ulbricht 2021). There are important normative differences between social and moral norms (Bicchieri 2006 Two questions arise: Are Anton's actions morally reprehensible? And are Anton's virtual crimes-virtual murder and virtual child molestation-morally equivalent? To exclude simple cases (i.e., to take the moral challenge of the Gamer's Dilemma seriously), let us assume that Anton plays alone, that he knows perfectly well that he (only) plays, and that he does not use the game instrumentally for gameexternal purposes (e.g., to prepare a real crime): Are his actions (still) morally reprehensible? 7 According to Kant, an action is immoral only when the agent's maxim (as a first-personally complete and appropriate justification of action) contradicts the categorical imperative. The categorical imperative reads in its classic form: "[A]ct only in accordance with that maxim through which you can at the same time will that it become a universal law" (Kant, 1998, 4:421/31). Regarding Anton, we must ask whether the maxim to virtually murder and abuse children can be willed as universal law without contradiction. Taking adequacy condition (2b) into account, the answer seems clear: Yes, murdering and abusing children in-game can typically be willed as universal law. After all, Anton does not violate any rights while playing; no moral duty seems to apply, as many theorists have already stated (Huizinga, 1980, 6;Ostritsch, 2017). Let us take a closer look at this argument by first considering a criminal's maxim: A. Criminal's Maxim: "I want to murder the shopkeeper and abuse his child!".
Let this be an (incomplete) 8 description of a maxim that results in real-world crime. This action is clearly reprehensible with Kant; it is not willable that its maxim becomes universal law. Now what distinguishes (A) from typical ingame actions?
B. Gamer's Maxim: "I want to quasi-murder the shopkeeper and quasi-abuse his child!".
How should we understand description (B) and its quasioperators? Anton's play is appropriately described not as murder and child abuse but as in-game murder and in-game child abuse. One is therefore mistaken if one claims that Anton murders and abuses while playing. He does not do that, and he is aware of it. In terms of action theory, he has the practical knowledge that he does not carry out the acts in the real world but only in the game world. 9 He knows that he must perform a different "primitive action" (Davidson, 2002, 49) while playing than is needed to perform the real-world counterpart. Using Kendall Walton's framework (Walton, 1978, 10-12), Anton only acts as if doing these things. He is performing quasi-actions, and he is aware of it. 10 Otherwise, it would not be possible to solve the fiction-theoretical puzzle that all in-game actions like Anton's action (B) pose: How is Anton, as a physical person, able to commit crimes in the fictive game world? After all, an ontologically robust "barrier against physical interactions between fictional worlds and the real world" (Walton, 1978, 5) is commonly assumed. The puzzle's solution is that Anton does not really commit crimes, but makes himself believe that he does so. 7 The following arguments are heavily inspired by Ulbricht's considerations (2022, esp. chapters 2.2.2 and 3.2). 8 A Kantian maxim integrates the full normative dimension of an action that the agent attributes to it. In the following examples (A, B, and C), I concentrate on the normative core of the agent's justification and will not fully represent her maxim with all of its possible side aspects. This reduced but focused perspective is sufficient for my argumentative purposes. 9 Elizabeth Anscombe established the notion of 'practical knowledge' to point out that agents always know what they are doing and how they are doing it (Anscombe 1963, §45). An essential part of the practical knowledge of a gamer is that he is in fact playing a game (Ulbricht 2022, 20-42). This is not taken seriously enough by Flattery's Kantian account. He writes about a gamer named Reed who "represented his act and circumstances in a way that is indistinguishable from how he might represent his act and circumstances in an analogous but nonvirtual situation. Thus, it's plausible that Reed's maxim, while in his state of immersion, was something like this: when I'm in the presence of someone who's defenseless and cornered, shoot them for the thrill of it" (Flattery 2021, 760). In my view, if we imagine Reed as a gamer, we must imagine him as fully aware of his situation: that he is actually not 'in the presence of someone who's defenseless and cornered'. Obviously, Reed adequately perceives his surroundings, which is why he presses buttons on a controller and does not do something else. Understanding the gaming situation differently does not happen incidentally but is only possible with an imaginative transformative act (Ulbricht 2022, 79-95): a specific form of make-believe that Flattery does not address. 10 I do not claim that every verbal formulation of a gamer's intention necessarily contains a quasi-operator, as in (B). On the contrary, when I am asked what I am doing while playing, I usually do not answer "I quasi-explore the cave", but simply "I explore the cave". And this seems perfectly fine. So, we need to differentiate. In the performance of quasi-actions, or 'fictional actions' (Ulbricht 2022, 28-36), typically no quasi-operator is immediately present, because while playing in this sense, we do not use a normatively adequate terminology to describe our actions (which would include a quasioperator), but rather a fictional terminology: We immerse ourselves in the fictive world. At the same time, we are fully aware of the normative dimension of our doing, which calls for the 'quasi-operator' in the maxim of our actions (we do not forget that all things are done only in play-which is why we make ourselves believe that they are not). This double-layered perspective of gamers has been described many times. By Johan Huizinga as "the consciousness, however latent, of 'only pretending'" (Huizinga 1980, 22) and more recently by Nguyen, who distinguishes the "outer layer" from the "inner layer" of gaming agency (Nguyen 2020, 56), which are both present while play: "At least to some degree, we must have the psychological capacity to maintain these layers simultaneously-to run the outer layer in the background" (Ngyuen 2020, 58).

Page 7 of 11 39
And this, in turn, can be willed as universally carried out way of acting in a Kantian sense. It is impossible to make oneself an exception if one merely plays. Therefore, gaming acts are typically not reprehensible. But this does not mean morally forbidden gaming acts do not exist at all. Consider the following maxim: C. Gamer's Maxim: "I want to kill the shopkeeper and abuse his child!".
What distinguishes (C) from (B)? In (C), we have gaming action but no quasi-operators. Thus, the maxim is subjectively indistinguishable from the maxim (A) of ordinary action. Since the Kantian view aims at maxims alone, it follows that acts (A) and (C) are equally reprehensible. But how exactly is gaming to be understood under (C)? Are such actions even possible? Consider this alternative display: C*. Gamer's Maxim: "I want to quasi-kill the shopkeeper and quasi-abuse his child!".
The representation of action (C*) resembles (B) in every respect except that the quasi-operators are crossed out. And actually, this is exactly what Anton does when he performs virtual crimes with maxim (C): He mentally deletes quasioperators (which would be normatively adequate for describing play, as Anton knows). He summons all his imaginative powers to (mis)understand the gaming context as possibility of performing real-world crimes. Thus, Anton does not accept the fiction of play but imaginatively transforms his ingame crime into a real crime. He tries (without any chance of success) to really carry out murder and molestation while playing. This makes Anton's action not only paradoxical but irrational. This is crucial: While every play inherently has the tendency to be paradoxical (when pretending, one does things that one does not really do), Anton's action in (C) is, moreover, irrational: He does not really commit crimes, is aware of this fact (Anton does not forget that he is playing), and yet tries to commit them. Why is he doing so? Because he imagines that he is actually committing crimes while playing: Subjectively, he is de facto realizing his reprehensible maxim, even though from an objective observer's perspective, he is not. Anton is aware of both perspectives. Using Nguyen's terminology, this phenomenon can be explained as follows: Anton "submerge[s]" himself "in the temporary agency of the game" (Nguyen, 2020, 10), while simultaneously caring about non-temporary ends that extend beyond the game. By placing his gaming under real ends, Anton himself subjects it to the moral law. He acts under the idea that his virtual actions are no longer mere play but moral seriousness. His virtual actions, then, are actually morally forbidden because they realize reprehensible maxims that contradict the categorical imperative. 11 This answers this section's first question: Anton's virtual actions are morally forbidden only if he performs them with a maxim that contradicts the categorical imperative. It should be noted, however, that such a way of gaming is exceptional. Usually, players accept the fiction of the game and do not try to transform it. To interpret one's own play in such an alienating way as in (C) is rather exotic.

Is there any moral difference between virtual crimes?
By answering the first question, the answer to the second one, and thus the Kantian response to the Gamer's Dilemma, is quickly formulated. From a Kantian perspective, the categorical moral distinction between virtual murder and virtual child molestation cannot be sustained. Adequately understood, both acts take place in a fictional setting and are therefore amoral: The maxims contain quasi-operators that make contradiction with the categorical imperative impossible. However, as soon as the in-game actions are (mis)understood as realizations of reprehensible maxims, as in case (C), virtual murder and virtual child molestation are equally morally forbidden because both contradict the categorical imperative. As a result, there is no Gamer's Dilemma with Kant. The truth of premises (i) and (ii) must be rejected. Virtual murder can (also) be morally forbidden, and virtual child molestation (also) be morally permitted, because an action's moral classification depends solely on the agent's maxims, not on its depiction. For the moral distinction between virtual acts, the (non-)presence of quasi-operators in the maxim is crucial, not the superficial categorization into 'murder' or 'child molestation'. Intended quasi-murder and quasi-molestation are permitted; intended murder and molestation are forbidden.

Isn't virtual child molestation worse than virtual murder?
A first objection to the Kantian response may be stated like this: Virtual child molestation is far worse than virtual murder!. This assumption is based on a robust moral intuition-in fact, the very intuition that gives rise to the Gamer's Dilemma. Apparently, the Kantian response does not address this intuition: It treats virtual murder and virtual child molestation as morally analogous. While the Kantian response delivers a consistent moral analysis of virtual crimes, it seemingly has difficulties grasping the intuitional ground of the Gamer's Dilemma and thus satisfying adequacy condition (1a).
There are numerous promising approaches within the framework of Descriptive Ethics that are compatible with the moral analysis given above and capable of explaining our intuitions regarding the Gamer's Dilemma (Kjeldgaard-Christiansen, 2020; Montefiore & Formosa, 2022). Rather than summarizing these approaches here, I will try to provide a Kantian explanation, which should add to the already existing contributions. I begin with the following thesis: Due to the intermedial omnipresence of killing acts, habituation effects to corresponding depictions (murders included) have occurred. This reduces the moral salience of corresponding fictive acts. In the case of (child) molestation, as a taboo, the opposite is true. 12 This is a sociological thesis that needs empirical substantiation. However, I think it is hardly controversial and is supported by our anecdotal experiences in different types of media. Be it in novels, movies, or games, we regularly witness acts of killing and murder, often included for entertainment purposes and even performed by the heroes of the story. In contrast, depictions of (child) molestation are extremely rare and intuitively very sensitive: They stick with us for a long time and are perceived as unthinkable crimes for any (anti-)heroic character (Vaage, 2015, esp. chapter 5). This contrast affects the 'moral salience' of the fictive actions. By this, I mean two things. First, our judgement changes: Unlike (child) molestation, murder seldom strikes us as morally problematic in fiction. Second, our perceptions differ, especially in gaming: Unlike acts of molestation, acts of killing are regularly not even perceived as such while playing but as entertaining obstacles, exercises in skill, or competitive challenges. Experienced gamers do not constantly think, "I am about to kill people!", but "I am about to finish the level!" or "I am about to win the game!". The first thesis is followed by a second one: The moral salience of an action type influences our maxims when playing this action type. To specify this, let us take a closer look at Kant's concept of maxims (1998, IV:421/31): A maxim is the subjective principle of acting, and […] contains the practical rule determined by reason conformably with the conditions of the subject (often his ignorance or also his inclinations), and is therefore the principle in accordance with which the subject acts [.] In a nutshell, the maxim is the synopsis of how the agent understands her action. Accordingly, the second thesis refers to how the player grasps her action factually and not to how she should grasp it. This is the reason why empirical circumstances such as habituation effects may play a significant role in the formation of maxims.
From the combination of both theses, the following conclusions can be drawn: Killing, as an intermedial omnipresent action type, tends to disappear from maxims of virtual killing, but molestation, as an intermedial taboo, tends to stick within maxims of virtual molestation. Some examples will illustrate this. Common maxims of virtual killing might entail the following justifications: "I want to finish the level!" or "I want to score points!" or "I want to defeat my brother!" or "I want to save the kingdom!". We usually accept these descriptions as reasonable justifications for in-game killing and murder. The first description is a perfectly adequate reason for killing numerous creatures while playing Super Mario Bros. (Nintendo Research, 1985). The second and third descriptions are plausible justifications for virtual killings within competitive games, such as Counter Strike (Valve, 2000), Super Smash Bros. (HAL Laboratory, 1999), or Age of Empires (Ensemble Studios, 1997). The fourth description justifies a variety of murderous actions in role-playing games like The Legend of Zelda: Breath of the Wild (Nintendo Entertainment, 2017) or The Witcher 3: Wild Hunt (CD Project RED, 2015). Now imagine a gamer who uses the above justifications not for virtual killing or murder but for virtual (child) molestation. Immediately, we would no longer accept the justifications as reasonable descriptions of her in-game acts. We would rather accuse the gamer of overlooking something obvious when she describes her play in this way. One important reason for this is that we have not become as accustomed to fictive (child) molestation as we have to fictive killings. In fact, the vast majority of gamers will find it very difficult to perform virtual (child) molestation without explicitly intending what they are doing. Molestation in games is so exotic that it is hard to imagine its description disappearing from the maxim. The gamer's maxim seemingly has to integrate molestation as a goal. This, in turn, immediately raises the moral question of how exactly she understands her action and whether her maxim 12 I write of 'killing' and 'murder'. These crimes are not the same. Killing may be justified (e.g., in self-defense), but murder is not. Intermedial conventions of representation differ as well; we witness fictive killings more often than fictive murders. Nevertheless, both practices are more common in media (esp. in computer games) and intuitively more justified than (child) molestation (Formosa et al., 2023). In addition, the boundaries between murder and killing are often blurred in games, so that we virtually murder far more often than we might think, e.g., when a game is set in war, when it is a stealth game, or when murdering enemies is primarily an exercise in skill.
contradicts the categorical imperative. Specifically, whether she negates the quasi-operator of her maxim or not.
To summarize, there is a Kantian answer to the first objection. While there is indeed no categorical moral difference between virtual murder and virtual child molestation from the deontological point of view, contingent intermedial conventions of representation have led to cases of reprehensible virtual child molestation being factually more likely than cases of reprehensible virtual killing. The intermedial omnipresence of killing acts (murders included) fosters a habituation so profound that the description of the fictive act may disappear entirely from the maxim of virtual murder. As a result, corresponding actions are mostly clearly amoral: There is nothing reprehensible about playing through levels, scoring points, or saving (fictive) kingdoms. Unlike acts of killing, acts of molestation, and child molestation in particular, constitute an intermedial taboo. There is increased sensitization regarding corresponding depictions, which makes it almost psychologically impossible not to refer to the fictive act in maxims of virtual (child) molestation. We are morally highly sensitive to such actions and cannot ignore them. This has moral consequences. With Kant, virtual crimes may only be immoral if the maxim entails the crime's description. Only then does negating the quasi-operator lead to moral offense. So, the risk of immorality is de facto greater in cases of virtual (child) molestation than in cases of virtual murder, which may provide a normative foundation for our different moral intuitions regarding the Gamer's Dilemma. 13

Aren't real crimes worse than virtual crimes?
There is a second objection: Carrying out real crimes is much worse than carrying out virtual crimes!. Most people would agree: Even if some virtual crimes are morally reprehensible, they are not nearly as evil as their real-world counterparts. At least with regard to the Gamer's Dilemma, this seems obvious: Unlike the real world, in game worlds, no one is actually killed, no one is harmed, and so on. The Kantian approach, however, has difficulties justifying this distinction. Depending on the maxim, actions are either forbidden or permitted (or commanded), because Kant knows no gradations of the (im)moral: Every crime is equally objectionable, whether virtual or not. Even outside of gaming contexts, Kant's failure to give different weights to reprehensible acts leads to counterintuitive conclusions: Lying is as immoral as murdering, and failed murder is as morally forbidden as successful murder.
As a first approach to the objection, let us stay with the last example and draw an analogy to our case. Important insights can be gained by comparing reprehensible virtual murder (e.g., Anton's virtual murder under the description (C) with a maxim that does not contain a quasi-operator and therefore contradicts the categorical imperative) with (also reprehensible) ordinary attempted murder. Both actions share some morally relevant properties: the agent wants to commit murder, the corresponding maxim contradicts the categorical imperative, and no one dies. Considering these similarities, it is actually not completely implausible to place the virtual crime on a comparable moral level as the real crime. Both can be understood as attempts to actually commit murder. However, one normatively decisive aspect has been omitted so far: Attempted murder does not fail willingly and has a chance of success, whereas reprehensible virtual murder does not. The latter is an attempt that cannot succeed, and the gamer is aware of this: She knows that her murder will fail. This results in a central irrationality in her action: It contradicts her intention. But how is this a contradiction to the categorical imperative? Actually, two contradictions take place, which are normatively connected by moral transitivity: I. Rational contradiction: Action x is to be realized by another, mismatched action y. II. Moral contradiction: Action x contradicts the categorical imperative and is ipso facto immoral. III. Moral transitivity: Action y, as intended (though mismatched) realization of action x, is immoral.
To illustrate this, let us again consider Anton's virtual crime under description (C): I. Rational contradiction: Anton wants to commit real murder and child molestation through virtual actions. II. Moral contradiction: Real murder and child molestation contradict the categorical imperative and are ipso facto immoral. III. Moral transitivity: Thus, Anton's virtual actions are immoral, because they are intended to realize crimes that contradict the categorical imperative.
Anton's decision to carry out an action that is doomed to fail represents his first central contradiction. Only because of this irrationality does his gaming maxim contradict the categorical imperative. Such a derivational relation is not found in ordinary crimes. But does this constitute a moral difference? It does not. In both cases, the agent's maxim ultimately contradicts the categorical imperative. Thus, both actions are to be judged morally the same: They are forbidden. 13 The considerations in this section are purely descriptive. I make no statement about whether it is desirable for habituation effects to be present in the case of virtual murder but not in the case of virtual (child) molestation. Presumably, it is not. However, while the specific characteristics of reprehensible virtual crimes do not ground a categorical moral distinction in the deontological sense, they do provide clues to further Kantian differentiations that may justify our diverging intuitions. For example, in his Groundwork of the Metaphysics of Morals, Kant divides moral duties "into duties to ourselves and to other human beings and into perfect and imperfect duties" (Kant, 1998, IV: 421/31). This marks a relevant distinction for our case: Attempted game-external murder violates a perfect duty to other human beings, whereas forbidden virtual murder violates a perfect duty to ourselves because only the agent is affected by murder that necessarily (and knowingly) fails. 14 More important differences can be found when taking a look at Kant's Metaphysics of Morals, where he distinguishes between fault and crime and between just and unjust acts (Kant, 1991, VI: 224/50): An unintentional transgression which can still be imputed to the agent is called a mere fault (culpa). An intentional transgression (i.e., one accompanied by consciousness of its being a transgression) is called a crime (dolus). What is right in accordance with external laws is called just (iustum); what is not, unjust (iniustum).
Shortly after, Kant makes a further distinction between legality and morality: "The conformity of an action with the law of duty is its legality (legalitas); the conformity of the maxim of an action with a law is the morality (moralitas) of the action" (Kant, 1991, IV: 225/51).
Following this classification, attempted game-external murder is a deliberate violation of external legislation, in which not only the maxim but also the act are contrary to duty. It is therefore an illegal and unjust crime. By contrast, reprehensible virtual murder has other normative characteristics. It is usually an unintentional violation of the moral law, in which only the maxim contradicts the duty. Forbidden virtual murder is thus an immoral but just fault. It is 'just' because it is lawful according to external legislationno applicable law forbids gaming. But why is it usually unintentional and thus only a fault? Admittedly, this point is debatable. But consider the following: The immorality of virtual murder essentially depends on rational contradiction, which ultimately consists in self-deception. Anton imagines that he could fulfill his game-external maxim while playing.
The key word is 'imagines': Anton does not forget that he is only playing. He knows that he is not really murdering anyone (otherwise he would not have to imagine it). He knows that he is the only one affected by his action. From this, I conclude that he is potentially unaware of nonetheless committing a moral offense. He may think his imaginative play protects him from moral guilt. However, the opposite is true: His fiction-transformative imagination leads to the moral condemnation of his actions. This is why forbidden virtual acts may usually count as faults and not crimes.
To sum up, reprehensible virtual murder, unlike gameexternal murder, affects only the agent and is neither illegal nor unjust. Furthermore, it is usually only a fault and not a crime. It is forbidden just like real murder, but the latter primarily concerns legality while virtual murder primarily concerns morality. The same conclusions can be drawn with regard to other moral offenses, such as (virtual) child molestation. Therefore, the second objection can be countered as follows: With Kant, reprehensible in-game actions have no different deontological status than their game-external counterparts. Both are morally forbidden for the same reason: They contradict the categorical imperative. Nonetheless, Kant allows us to identify important normative differences between the acts that may ground our intuitions. However, for deeper evaluative weightings, one must turn to other, value-based ethics.
Funding Open Access funding enabled and organized by Projekt DEAL. No funds, grants, or other support was received.

Competing interests
The author has no relevant financial or non-financial interests to disclose.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.