Layered Responsibility: Designers, Decision-Makers and the Moral Burden of Social Media Manipulation

Abstract

This paper asks who bears moral responsibility when social media platforms facilitate political manipulation. It argues that moral responsibility in this domain is gradual and layered, and that it rests primarily on the designers and decision-makers who build and govern these platforms, rather than on corporations understood as abstract entities or on users understood as autonomous individual agents. The argument proceeds in three steps. First, it reconstructs the classical conditions of moral responsibility (knowledge, rational deliberation, freedom to act otherwise, continuity of the self and embeddedness in relations of accountability) and shows that responsibility admits of degrees. Second, it examines and rejects two alternative attributions: to the corporation, which lacks the capacities of a moral agent, and to users, whose capacity for autonomous deliberation is systematically undermined by the very architectures they inhabit. Third, it shows that designers and decision-makers satisfy the conditions of responsibility in a robust form, owing to an epistemic privilege grounded in iterative design methodologies that allow them to anticipate and measure the behavioural effects of their choices. Drawing on Kant, Arendt, Young and the contemporary literature on organisational responsibility, the paper concludes that the omission to act on foreseeable harm constitutes a moral failure that cannot be dissolved into the anonymity of the system or transferred to those whom the system is designed to influence.

Keywords: moral responsibility; degrees of responsibility; corporate agency; design ethics; omission; structural responsibility; social media

Social media platforms are deliberately designed products whose algorithmic architectures determine what information reaches whom, at what moment and with what emotional intensity. When these architectures are exploited for political manipulation, the question of moral responsibility becomes pressing: who answers, morally, when such systems facilitate the erosion of democratic deliberation?

Public debate typically offers two answers. Responsibility is attributed either to corporations, as collective entities, or to individual users, considered autonomous agents who ought to have known better. This paper argues that neither answer is adequate. Corporations do not, in a strict sense, meet the conditions of moral agency, and users act within environments deliberately designed to limit critical reflection and decisional autonomy. Designers and decision-makers, by contrast, combine technical knowledge of the systems with the capacity to anticipate the consequences of design decisions. This position confers on them both causal power and moral responsibility.

The thesis defended here is that the principal moral responsibility for digitally mediated political manipulation is gradual and rests largely with the designers and decision-makers within the companies that develop the social platforms. This conclusion follows from an analysis of the position designers occupy in the development of digital products: they possess knowledge of user behaviour, they anticipate the effects of design decisions, and they have the practical capacity to act otherwise. These three conditions, knowledge, anticipation and decision-making power, are precisely those that ground the attribution of responsibility.

The argument proceeds as follows. Section 2 defines moral responsibility and distinguishes it from causal and legal responsibility. Section 3 argues that responsibility admits of degrees. Sections 4 and 5 examine and reject the two alternative attributions, to corporations and to users. Section 6 establishes the responsibility of designers and decision-makers through the concepts of foreseeability and moral omission. Section 7 answers the objection from organisational constraint and develops the model of layered responsibility.

To determine who is morally responsible, one must first define what moral responsibility is, who can be morally responsible, and whether responsibility is absolute or admits of degrees. Moral responsibility refers to the attribution of ethical accountability to an agent who satisfies certain philosophical and practical conditions. It must be distinguished both from legal liability and from mere causal responsibility.

Moral responsibility cannot be reduced to the simple fact of producing consequences in the world, since this would amount to an erroneous superimposition of causal responsibility onto ethical responsibility. Whereas causal responsibility is limited to identifying the force or factor that triggered an event, the moral dimension requires a subjective bridge between the agent and their effects, defined by the capacity for knowledge and control. Producing an outcome is therefore a necessary but insufficient condition for the attribution of responsibility. In the Aristotelian framework, an agent cannot be blamed or praised for consequences that arise through force (external constraint) or through ignorance (the absence of the capacity to anticipate the outcome), situations in which the principle of the action remains external to the subject (Constantinescu, 2013). An action becomes morally relevant only when it results from a deliberate choice (prohairesis), through which the subject not only causes a physical effect but also anticipates and evaluates its significance (Constantinescu, 2013).

Deontological perspectives, Kant's in particular, reinforce the point from another direction: morality resides in the quality of the will and in good intention, not in the physical effects of the deed. An agent who makes every ethical effort may be exonerated if a negative outcome is produced by an unfortunate accident beyond their control, whereas an act with good consequences but performed from purely selfish or accidental motives may lack moral content altogether (Mureșan, 2009). Responsibility thus concerns agency: the capacity of the subject to set ends and consciously choose the means, which transforms the production of consequences from a mechanical process into an owned act that justifies the reactive attitudes of the community, such as indignation or gratitude (Constantinescu, 2013).

Five conditions recur across the tradition.

Knowledge and awareness. Starting from the Socratic paradox recorded by Plato and explained by Irwin (1995), according to which no one does wrong willingly but only from ignorance of the good, it follows that responsibility presupposes understanding what we do and the moral significance of our actions. Augustine (1961) later developed this intuition, showing that although knowledge of good and evil is indispensable, it is not sufficient: an agent may know that an action is wrong and still desire it and commit it. The episode of the stolen pears in the Confessions illustrates that moral responsibility requires the conjunction of awareness with an act of the will.

Reason and deliberation. Kant (1785/1997) grounded moral responsibility in autonomy. For Kant, a responsible agent is one who can act according to the moral law as understood by reason, and who can recognise themselves as bound by duties they give to themselves. The capacity to deliberate and to govern oneself through reason draws the line between actions that can be praised or blamed and those that remain outside the domain of responsibility.

Freedom to act otherwise. We do not hold someone responsible for what they had no power to avoid. Aristotle's distinction between voluntary and involuntary actions established that constraint and ignorance eliminate or diminish responsibility (Aristotle, 1925). Augustine (1961) insisted that praise and blame would be meaningless without free will, and modern philosophy condenses the principle in the formula "ought implies can", used by Kant (1785/1997) to underline that responsibility presupposes the capacity to act otherwise. Even compatibilist theories, such as those of Hume (1748) and, later, Strawson (1962), retain the idea that a responsible agent must act from their own will, free of constraint or compulsion, even if their choices are causally determined.

Identity and continuity. Responsibility presupposes that the person who acted is the same person who answers for the action. Locke (1690) argued that personal identity, grounded in the continuity of consciousness, is the basis of responsibility: to punish Socrates awake for what Socrates did asleep would be "no more of right, than to punish one twin for what his brother-twin did" (Locke, 1690, II.xxvii.19).

Relations of accountability. Finally, responsibility is not exercised in isolation but within a network of relations. Strawson (1962), in "Freedom and Resentment", argues that responsibility is bound up with the reactive attitudes, resentment, gratitude and indignation, through which we hold one another to account. Iris Marion Young (2006) showed that responsibility can be both retrospective and prospective, requiring agents to recognise their place in wider structures that affect others and to act with that interdependence in view.

From these perspectives a coherent definition can be formulated. A moral agent is a being who possesses knowledge of good and evil, who can deliberate rationally about their actions, who enjoys sufficient freedom of choice, who maintains personal identity over time, and who stands in relations of accountability to others. As Constantinescu puts it, the sphere of moral responsibility ascription "includes actions performed voluntarily (that is, the action was caused by the agent, who knows the particular circumstances of their actions, and the action is performed without constraint) and, at the same time, based on a deliberate decision" (Constantinescu, 2013, p. 25). To describe someone as morally responsible is to affirm that they are the author of their own actions in a sense deep enough to justify praise, blame or obligation.

If moral responsibility presupposes knowledge, deliberation, freedom and continuity of consciousness, it becomes relevant how these conditions are satisfied in different situations. Agents do not always act from equal positions of control, information or power, which raises the question of whether responsibility can be attributed uniformly to everyone involved in a shared action or outcome. Coates and Swenson answer that it cannot: "agents can be more or less responsible or blameworthy in virtue of the degree to which (i) they are sensitive to the range of reasons at issue, or (ii) the degree to which their sensitivity fits into an intelligible pattern" (Coates & Swenson, 2013, p. 631).

In Kantian ethics, moral responsibility has a categorical character. Categorical imperatives are formulated as universal obligations that apply to all rational agents without exception (Kant, 1785/1997). From this perspective, if an agent satisfies the conditions of moral agency, rationality, freedom and the capacity for deliberation, they are fully responsible for their actions. An action either conforms to the moral law or contradicts it, and responsibility admits no gradations.

This conception encounters difficulties when applied to situations in which moral obligations conflict. Ross (1930) shows that moral duties, which he calls prima facie duties, do not function as absolute rules but as moral reasons whose weight varies with context. The duty of fidelity may conflict with the duty to prevent harm, and resolving such a conflict requires evaluating concrete circumstances. Within this framework, the attribution of responsibility cannot ignore the fact that agents often act under moral tension and with incomplete information.

The same problem arises in analyses of collective responsibility. Smiley (2022) observes that the literature frequently supports differentiating moral responsibility among members of a group according to factors such as the role played, the degree of causal contribution and the level of control over the action. Leaders or instigators of a collective action are often considered more responsible than marginal or reluctant participants, even when all contribute to the same harmful outcome. Coates (2022) develops this idea, showing that responsibility admits of degrees along several relevant dimensions: the agent's degree of control over the action, their sensitivity and responsiveness to reasons, and the deliberative capacities available to them at the moment of acting.

Moral responsibility is therefore more adequately understood as scalar. This interpretation allows us to differentiate between agents who act from distinct positions of power, knowledge and control without abandoning the idea of moral obligation. It is this scalar conception that the remainder of the paper applies to the three candidate bearers of responsibility: corporations, users, and the designers and decision-makers who build the platforms.

A frequent assumption in ethical debates about technology is that large companies, particularly in the Big Tech sector, can and should be held morally responsible for the harms caused by their products or design decisions. This position has been firmly contested by philosophers such as Manuel Velasquez (2003) and Jerry Sheppard (1994), who argue that corporations, despite their legal status as persons, are not moral agents in any philosophically relevant sense.

In law, corporations are treated as legal persons: they can own property, enter into contracts, sue and be sued, and, in some jurisdictions, even claim certain constitutional rights. This legal personality is, however, a functional fiction. It exists to facilitate economic activity and the regime of legal liability, not to affirm the existence of moral agency. Velasquez (2003) is explicit on this point: legal personality is a convenience of the legal system and must not be confused with the presence of moral capacities such as intentionality, conscience or moral awareness. Sheppard (1994) formulates a similar argument, observing that although the law treats corporations as persons for regulatory purposes, this does not mean they can feel guilt, care for others or reflect morally on their own actions. These capacities belong to natural persons, not to institutional constructions.

Velasquez (2003) adds that action within a corporation is always reducible to the decisions of individual persons: designers, engineers, managers and executives acting within well-defined roles. There is no metaphysically distinct agent called "the corporation" that deliberates, chooses or assumes responsibility. To treat the corporation as a moral person is, in his view, a category error that obscures the essential question of who is in fact responsible.

Even when a company appears to act morally, through donations to social causes or by taking a political position, such actions are often strategic, initiated by concrete individuals and reversible with a change of leadership. Apple's refusal to unlock a user's iPhone for the FBI, Patagonia's long-standing environmental activism and Twitter's 2020 decision to ban political advertising were all led by individual executives, Tim Cook, Yvon Chouinard and Jack Dorsey, and were frequently aligned with brand strategy or long-term reputation management. They do not demonstrate that the organisation as such possesses a moral character of its own.

As Sheppard (1994) emphasises, even organisational culture is created, imposed and interpreted by people, especially boards and executive leadership, who often operate under fiduciary constraints that prioritise shareholder interests over moral principles. The result is that, although corporations may be causally responsible for harmful effects, they are not adequate subjects of moral responsibility. They lack the autonomy, emotional life and reflexive capacity that define moral agents. Any analysis of digital manipulation must therefore avoid the abstraction of "blaming the company" and focus on the human actors who possess intention, foresight and the power to act differently.

This does not render the organisation morally irrelevant. Constantinescu and Kaptein (2021) show that, although organisations are not moral agents in the philosophical sense, they create the conditions that facilitate or inhibit individual ethical behaviour. They propose a two-level model: at the individual level, responsibility depends on satisfying the Aristotelian conditions of causation, knowledge, freedom and deliberation; at the organisational level, responsibility derives from the way organisational practices support or undermine those conditions. Applied to digital platforms, this framework suggests that although Meta or TikTok are not moral subjects in the strict sense, they bear a degree of responsibility for creating an environment in which manipulative design decisions become possible and are even encouraged by the prevailing incentive structures. The corporation is thus a site of responsibility, but not its primary bearer.

If the corporation is not the primary bearer of responsibility, one might locate it at the other end of the chain, in the users who share, believe and act upon manipulated content. In contrast with designers and those holding ultimate decision-making power, who act with foresight and authority, end users of digital platforms cannot be considered fully morally responsible for the political manipulation facilitated by social media systems. Although they may participate in the circulation of content or act on manipulated information, their conduct does not consistently satisfy the philosophical conditions required for full moral responsibility.

From a Kantian perspective, moral responsibility springs from autonomy and rational will (Kant, 1785/1997). A morally responsible person does not merely react to stimuli or act out of habit, but is capable of acting according to principles they could will as universal laws. For Kant, freedom does not mean doing whatever one wishes, but acting from respect for the moral law as formulated by reason, which presupposes full awareness of the nature and consequences of one's actions. In digital environments characterised by infinite scroll, opaque recommendation systems and emotionally charged micro-interactions, users frequently lack precisely this awareness. The architecture of the platforms is designed to bypass reflective judgement and to encourage automatic behaviour. The more fluid and "invisible" the interface, the more opaque the system becomes, and the fewer occasions the user has to consider the moral implications of their own engagement.

Heidegger (1977) observes in The Question Concerning Technology that technologies which become stably integrated into everyday practice tend to withdraw from conscious attention. Their functioning becomes habitual and their presence is no longer perceived as a distinct object of experience. Under these conditions, technology no longer appears as a medium or form of mediation, but as the implicit background of action and perception. When the technological infrastructure and its design elements are not explicitly perceived, users have fewer opportunities to reflect on how their behaviour is structured, oriented or facilitated by the digital environment. Similarly, Beard and Longstaff (2018) note that technological systems tend to embed particular modes of reasoning and operation, which become standards of functioning as the technology is used and come to structure users' interactions and decisions without being made explicit in ordinary experience. In Kant's terms, then, most users act under the guidance of external forces rather than autonomous moral deliberation.

Mill (2005), working within a utilitarian framework, shows that moral judgement depends on the conditions under which choices are made. In On Liberty, he argues that authentic moral development requires not only the freedom to act but also the capacity to choose wisely, on the basis of experience, education and deliberation. If users are captive in manipulative design loops, the conditions necessary for rational reflection are gravely degraded. Mill would argue that the mere existence of an option is not sufficient; individuals must also be supported or empowered to exercise it meaningfully, and it is precisely this that the digital architectures of social media platforms undermine. As Vermaas et al. put it, "the concept of freedom presupposes a form of sovereignty with respect to technology that human beings simply no longer possess" (Vermaas et al., 2011, p. 97).

A reasonable objection may be raised at this point: what of those users who do manage to resist manipulation, who think critically, verify sources or deliberately limit their exposure to platforms designed for addiction? Does their conduct not show that users can, in the end, be morally responsible? The objection is legitimate. Not all users are entirely powerless; some make real efforts to inform themselves, to question what they consume and to regulate their digital behaviour. If responsibility is a matter of degree, as Section 3 argued, users may bear some responsibility for how they relate to digital platforms, particularly when they possess the capacity and awareness necessary to act differently.

This partial responsibility must, however, be understood in context. Individuals who succeed in resisting are often privileged in one way or another: they may have access to higher education, a stable emotional environment, strong support networks or prior training in critical thinking. These conditions are far from universal. In Romania, a 2024 Ministry of Education study on adult competences found that a substantial proportion of adults fall into the category of functional illiteracy, lacking the capacity to understand or interpret written information in a meaningful way (Ionescu, 2025). Under such conditions, the ability to verify information, recognise manipulation or engage in reflective judgement is severely compromised. To demand full moral responsibility in these circumstances is a misattribution of responsibility. Although resistance is possible in principle, it often requires an extraordinary level of effort and knowledge, comparable to navigating a casino designed precisely to make one lose. The existence of partial moral agency does not shift the burden of responsibility away from those who design these systems with full knowledge of their effects.

Thus, although end users may bear some moral responsibility, particularly where they possess the tools and awareness needed to act otherwise, this responsibility is limited and secondary. Moral responsibility is not absolute but relational and context-dependent (Talbert, 2025), and the structural, psychological and informational constraints imposed on users by platform design severely limit their capacity for moral autonomy.

Designers and decision-makers occupy an epistemically and practically privileged position with respect to the effects of the platforms they build. This position is constitutive of their professional role. Product teams, composed of product managers, UX and interaction designers, user researchers, program managers and data scientists, are trained to understand user behaviour, to anticipate reactions, to test hypotheses and to optimise systems deliberately. Their training typically includes an ethical component: university programmes in product management and design at institutions such as Stanford, Dartmouth and Harvard Kennedy School, as well as professional certifications such as those of Nielsen Norman Group or Google's Project Management certificate, which draws on the Markkula Center's Framework for Ethical Decision-Making (Kwan, McLean & Raicu, 2021), include modules on ethical decision-making and social responsibility.

More importantly, their working methodology gives them continuous access to the effects of their decisions. Design Thinking, the dominant framework in digital product development, is an iterative process of empathy, problem definition, ideation, prototyping and testing, in which each phase is evaluated against behavioural data (Moran & Brown, 2024). Every iteration produces quantitative and qualitative evidence about how a product is adopted, used and perceived, and thereby about its social, cultural and political impact. The evolution of user behaviour is not opaque to the people who build the system; it is observable and measurable. Beard and Longstaff (2018) recall Richard Buchanan's description of design as "an art of anticipation" (p. 5), and remind us that "just because we can do something doesn't mean we should" (p. 5). From a Kantian perspective, then, designers and decision-makers satisfy the conditions of moral agency in a robust form: their role structurally requires deliberation, justification and informed choice.

If a feature is designed from the outset with the explicit aim of manipulating attention or producing ideological isolation, its problematic character is evident from the standpoint of any major moral theory. The specific ethical difficulty of digital platforms arises, however, in another register: features are introduced with descriptively legitimate aims, such as improving user experience, increasing relevance or reducing friction, but through optimisation and scaling they come to produce harmful systemic effects.

In such cases, an appeal to initial intention is not sufficient to suspend moral evaluation. Even in the absence of malicious intent, the failure to recognise the emergence of harmful effects and to react adequately constitutes a distinct, but comparably serious, form of moral failure. As Hannah Arendt (1964) shows, moral harm does not always result from an explicit will to do evil; it may arise from thoughtlessness, from continuing a professional practice without questioning the meaning and consequences of the systems to which one contributes. What Arendt calls "the banality of evil" designates precisely this normalisation of harm by agents who suspend moral reflection and reduce action to routine, efficiency and institutional conformity.

This thoughtlessness is amplified by siloing, the fragmentation of work into isolated decision units, in which small teams confine themselves to highly specific technical tasks without visibility over, or influence upon, the final product (The Ethics Centre, 2021). Designers come to perceive themselves as a mere part of a vast bureaucratic mechanism, which facilitates the externalisation of ethical responsibility to other departments or to higher levels of management. May (1992) argues that this form of institutional organisation produces an extended moral negligence, consisting in the deliberate failure to investigate the probable social consequences of one's own work under the pretext of lack of time or political competence. Once an agent chooses to ignore foreseeable risks, a fostering mechanism is activated that normalises harm within the organisational culture, transforming individual inaction into a systemic group policy (Constantinescu & Kaptein, 2021).

The conclusion is reinforced by Valentin Mureșan's analysis in Trei teorii etice: Kant, Mill, Hare, where he shows that intention alone is never sufficient to exonerate an agent when the consequences were foreseeable: "if you could have anticipated the consequences of your deed and did not, the fault is yours" (Mureșan, 2009, p. 131; my translation). The argument is particularly relevant to digital technologies, because moral responsibility attaches not only to the direct production of harm but also to the omission to prevent foreseeable harm when the agent had the means to foresee it. It is precisely this condition that is characteristically satisfied in the development of social media products. When design mechanisms begin to generate recurring harmful effects, and teams choose to maintain and optimise them without corrective intervention, this can no longer be interpreted as an accidental consequence of system complexity. It takes the form of a moral omission: the deliberate maintenance of a design whose harmful effects have become internally detectable.

This view is supported by what French (1984) calls the Principle of Responsive Adjustment: an agent who refuses to modify their conduct after being confronted with the harmful effects of their work associates themselves morally with those outcomes, transforming initial ignorance into deliberate complicity. In Aristotelian terms, responsibility includes not only acts of commission but also omissions, insofar as the agent is able to evaluate the significance of their professional silence relative to the end pursued (Constantinescu, 2013). For platform designers, this amounts to a breach of the duty of due care (Naoe, 2008), since their technical position obliges them to foresee problematic uses and to integrate safeguards. When these professionals reduce moral action to routine and to conformity with profit metrics, ignoring that their "product" is ultimately human behaviour, they fall into Arendtian thoughtlessness and normalise systems that function as "hostile scaffolds" of attention (Voinea, Marin & Vică, 2024).

The responsibility of designers and decision-makers does not cancel the agency of those who create and disseminate manipulative content, but it shows that their success is inseparable from the choices embedded in the platforms' architecture. Designers create the conditions for capturing attention and amplifying content, conditions that are subsequently exploited maliciously, which makes the creators of the platforms co-responsible for the scale and effectiveness of the manipulation. Designers cannot invoke neutrality, ignorance or lack of authority. If they understand these effects and continue to act in the same way, they are responsible. If they understand them and fail to inform decision-makers, they are responsible. If they do not investigate the effects, although they have both the professional duty and the capacity to do so, they remain responsible.

The idea that specialised knowledge entails moral responsibility is ancient. In the Gorgias and the Republic, as Schmid (2020) explains, Plato distinguishes the true arts (technē) from flattery and manipulation: the true craftsman not only knows what he is doing but aims at the good of the one upon whom he acts. Irwin (1995) explains how Plato warns that when a practice is decoupled from its moral end, it degenerates into a form of power over others without ethical orientation. This is exactly the danger digital designers face: they exercise technical and psychological power, but often operate without a clear conception of the good.

An important objection to the conclusion that designers and decision-makers are the principal bearers of moral responsibility concerns the role of the organisational context in shaping individual action. Designers do not act alone; they work under the coordination of managers, boards and organisational policies. Constantinescu and Kaptein (2021) show that, in organisational settings, the conditions of individual moral responsibility, causation, freedom, knowledge and deliberation, can be directly influenced by organisational practices. Consequently, moral outcomes produced within organisations cannot be attributed exclusively to individuals, but also to the structural and cultural configurations that stimulate or inhibit ethical behaviour.

The objection is accurate as a description, but it reinforces rather than diminishes the responsibility of designers. If designers are aware of harmful effects and inform decision-makers, and leadership deliberately continues, responsibility shifts to the board or the executives. But if designers observe harm and say nothing, whether out of fear, inertia or indifference, responsibility remains with them. And if they fail to notice effects that they are professionally capable of noticing, that failure is also theirs. As Constantinescu and Kaptein put it, "the more passive individuals are (relative to their capacity to fight or report wrongdoing), the more responsible and blameworthy they are" (Constantinescu & Kaptein, 2021, p. 22). In all these scenarios, designers constitute the first line of moral defence in the creation of digital systems. Executives may take strategic decisions, but they are often not experts in behavioural design or in the mechanisms of psychological manipulation; it falls to designers to signal risks, propose alternatives and refuse participation in systems that erode user autonomy or dignity.

It is true that organisational structures and institutional culture can discourage ethical initiative. This does not suspend the responsibility of designers; it calls for a finer analysis of how moral agency is exercised within contemporary organisations. In large technology companies, design and product work is frequently not conceived as the mere execution of orders from above. Specialists in product design, UX and product management often identify problems, propose solutions, formulate hypotheses, set product direction and justify decisions against empirical data, while senior management provides feedback or approves direction. This structure confers on designers a real margin of moral initiative, even in the absence of an ideal organisational climate. Indeed, the problematic character of an organisational environment makes the obligation to signal, to propose less harmful alternatives or to contest harmful decisions more stringent, not less.

It is, of course, possible that ethical initiatives formulated by designers are deliberately rejected by executive leadership for strategic or political reasons. In such a scenario, the principal moral responsibility justifiably shifts to the higher decision-making level. This shift, however, is not automatic; it presupposes a genuine attempt at moral intervention by those who identify the problem. The absence of any such attempt cannot be justified by simply invoking organisational culture. Nor is that culture an autonomous entity external to agents: it is the product of daily practices, repeated decisions and norms tacitly accepted by members of the organisation. Culture can explain behaviour, but it cannot morally justify it.

What emerges is a model of layered responsibility, which can be developed on three levels: individual, leadership and team, and structural. Responsibility, on this model, is not a finite resource to be divided, but a phenomenon that multiplies through the interaction of these levels. Within Big Tech corporations, it can neither be attributed exclusively to a single individual nor dissolved in the anonymity of the system.

At the individual level, designers and product managers satisfy the fundamental conditions of moral agency: causation, knowledge and deliberation. The designer becomes a "choice architect". Through Design Thinking, they possess, or should possess according to professional ethics, an in-depth knowledge of how a feature will affect user behaviour. May's concept of the "benighting act" is decisive here: an agent is responsible for present harms if, through a voluntary decision in the past, they omitted to acquire the knowledge necessary to avoid them (May, 1992, p. 63). The omission to learn how to identify, and then act upon, foreseeable harm is itself a moral failure.

At the leadership and team level, responsibility takes on a collective dimension through what Peter French (1984) calls the Corporate Internal Decision Structure, which subordinates and synthesises the acts of persons into a corporate decision. Leadership is responsible for creating the institutional conditions that either stimulate or inhibit individual ethical conduct. The virtue of congruence, for instance, requires that the signals managers transmit, such as the metrics by which success is measured, be aligned with the ethical values the organisation declares (Constantinescu & Kaptein, 2021). An organisation can thus be morally evaluated according to whether its practices embody organisational virtues such as clarity, congruence, discussability and sanctionability, and deficiencies in these practices increase the probability of unethical conduct and justify a greater degree of organisational responsibility.

At the structural level, Iris Marion Young's theory of structural responsibility establishes that agents remain responsible even when they act within complex, distributed and institutionalised systems in which no single action appears to cause harm on its own. Young distinguishes the liability model, which seeks a single culprit for a past event, from a model of political responsibility, which is forward-looking and concerns the way all actors in a structure contribute to sustaining processes that produce unjust outcomes (Young, 2011). From this perspective, platform designers are participants in the production of structural processes that facilitate political manipulation, and are obliged to recognise their place in architectures that affect the autonomy of others (Young, 2006). The designer cannot invoke the complexity of the algorithm to dissolve their own agency, since structural responsibility imposes an obligation to intervene actively in correcting systems that undermine the cognitive resilience of society.

This paper has argued that moral responsibility for digitally mediated political manipulation cannot be dissolved into the anonymity of the system or transferred exclusively to users, but must be located at the level of the agents who hold technical and epistemic control over the platforms.

It began by reconstructing the conditions of moral responsibility, knowledge, rational deliberation, freedom to act otherwise, continuity of the self and relations of accountability, and by distinguishing moral from merely causal responsibility. It then argued, with Ross, Smiley and Coates, that responsibility is scalar rather than binary, growing with an agent's degree of control, knowledge and influence.

Two alternative attributions were examined and found insufficient. Corporations, though legally relevant, do not satisfy the conditions of moral agency; they always act through the concrete decisions of designers, managers and leaders, even if their practices shape the conditions under which those decisions are made. Users, though not entirely devoid of agency, act within architectures deliberately designed to bypass reflection, and their capacity to resist depends on resources that are far from universally distributed. Their responsibility is real but limited and secondary.

Designers and decision-makers, by contrast, satisfy the conditions of responsibility in a robust form. Through iterative methodologies of testing and measurement, they possess an epistemic privilege that allows them to anticipate the social consequences of their decisions. Kantian ethics grounds their responsibility in the duties that bind rational agents; Arendt shows that moral failure may arise not from malice but from thoughtlessness; and Young establishes that agents remain responsible within complex, distributed systems. Taken together, these perspectives show that designers cannot evade moral responsibility by invoking the organisational framework, institutional pressure or the collective character of decisions.

Responsibility is therefore layered: designers make ethically relevant decisions daily, teams and leadership set priorities and incentive systems, and business models oriented towards the exploitation of attention turn manipulation into a systemic outcome rather than an accidental deviation. The omission to act in the face of foreseeable harm, under the pretext of professional routine or institutional pressure, constitutes an owned moral failure. Designers remain, in this sense, the first line of defence of human dignity in the age of algorithms. As AI-driven personalisation and automation expand, the gap between what platforms can foresee and what they choose to act upon will only widen, making the recognition of this responsibility an essential condition for protecting democratic integrity.