Negotiated Harms in Moralized Policies: The Case of Duterte’s War on Drugs

Viewing a policy as harmful can lead to its moral condemnation. However, this harmfulness can be constructed and negotiated to lead to different moral positions by building upon available, accessible, and relevant discourses. This study examined how individuals constructed and negotiated harm in moral reasoning about a contentious policy, Philippine President Duterte’s war on drugs, locally known as tokhang. We conducted thematic analysis with attention to discourse to analyze interviews with 12 Filipino young adults, using the Theory of Dyadic Morality as a starting point to make sense of constructions of harm. Reasoning about tokhang showed different constructions of intentional agents and vulnerable victims serving as the basis for moral positions. Moral condemnation of the war on drugs emphasized the vulnerability of its victims and the intentionality of the government and police as agents. On the other hand, moral justification of the policy constructed drug war victims as agentic and guilty of crimes, the police as potentially vulnerable victims acting according to protocol to defend themselves, and rogue agents acting independently of the policy. Ambiguous positions were also made possible when the causality of harm is unclear. These constructions and negotiations were built upon broader discourses deployed in the sociopolitical context of urban young adults, with individual contexts and characteristics contributing to variations in the accessibility and relevance of certain discourses and resulting moral positions. applied productively to the study of other policies, where discourses on elements of harm can be identified, crafted, and deployed to potentially shape moral positions and provide opportunities for discussion and negotiation on the basis of The present study is part of a project with broader goals of understanding young adults' moral reasoning using thematic analysis with attention to discourse (Braun & Clarke, 2006, 2021; Gill, 2000). It examines moral reasoning from a social constructionist perspective (Burr, 1995), allowing moral reasoning to be examined as constructed within social, historical, and cultural contexts through social interactions, and having implications for people’s actions in their social worlds.

2015). However, the harmfulness of a policy can be constructed and negotiated to lead to different moral positions by individuals building upon available, accessible, and relevant discourses. One important policy that has been heavily discussed and negotiated, and that can be examined in terms of harm is Philippine President Duterte's war on drugs. The implementation of this policy has been linked to thousands of deaths and other human rights violations (United Nations Office of the High Commissioner for Human Rights [UN OHCHR], 2020), but has also maintained popular support (Hapal, 2019;Lamchek, 2017) partly owing to messages about benefits to community security (Uyheng & Montiel, 2021;Warburg & Jensen, 2020). Given these conflicts, along with the centrality of harm as a moral concern, the present study aims to analyze how individuals construct and negotiate harm to arrive at their moral positions on the war on drugs. This can potentially create space for discussion and understanding to minimize societal harm, both from the policy and polarizing, morally-charged public discourse.
We use a discursive approach to analyze moral reasoning about the war on drugs. Discursive approaches consist of a wide range of theoretical and disciplinal traditions, all of which converge in the "rejection of the realist notion that language is simply a neutral means of reflecting or describing the world, and a conviction in the central importance of discourse in constructing social life" (Gill, 2000, p. 172). Shared features include: (1) Interest in how individuals use language, focusing on the content and organization of discourse rather than assuming that what is said corresponds to internal thoughts and beliefs; (2) The view of language as constructive, with discourses shaping possibilities for thought, action, and identity or subject positions, and constructed, where accounts are built by choosing from discursive possibilities; (3) Emphasis on discourse as a form of action serving a particular function within social interactions, involving simultaneous analysis of discourse and interpretive context; and (4) Consideration of discourse as involving rhetorical organization, where discourse establishes one of several competing versions of the world (Gill, 2000). In the present study, a discursive approach is used to analyze individuals' moral reasoning at the level of language used in the social context of an interview, and examine how they construct and negotiate their position on the war on drugs on the basis of harm by drawing from existing discourses about the policy.
We also use the Theory of Dyadic Morality (TDM; Schein & Gray, 2016 as a starting point to make sense of constructions and negotiations of harm. TDM argues that an issue can be constructed as morally relevant when an intentional agent is perceived to cause harm to a vulnerable patient or victim. Though harm is considered to be relevant to everyone, disagreements arise when individuals construct different degrees of harm in the situation, based largely on their judgments of the agency and vulnerability of agents and victims, and the causality of harm between them. Despite the suggestion that TDM may be useful in political discourse (Schein & Gray, 2018), the theory has yet to be used in discursive research, a gap which this study intends to address.

Duterte's War on Drugs and Project Tokhang
The war on drugs has been central among President Duterte's policies. Though wide in its scope, its most controversial feature and thus the focus of this study is "Project Tokhang", officially described in the Philippine National Police (PNP, 2016) memorandum circular as "the conduct of house to house visitations to persuade suspected illegal drug personalities to stop their illegal drug activities" (p. 3). The circular justifies this on the basis of "the worsening drug problem that has victimized mostly the underprivileged and impoverished sector of the society" as "the government seems to have overlooked the worsening drug problem at the grassroots level" due to its focus on pursuing high level drug traffickers (p. 2). The source of harm then is identified to be "the worsening drug problem". The document describes the operations in benign terms, but hints at the use of force in euphemistic terms such as "the neutralization of illegal drug personalities" (p. 3). Messages from Duterte are more explicit in encouraging violence. For instance, in a speech with police as his audience, he said "Do not bullsh*t with me but do your duty, I will die for you. Do your duty and if in the process you kill 1,000 persons because you were doing your duty, I will protect you." (Francisco, 2016). More broadly, his speeches position the killings as moral and part of his rights and duties as a leader upholding his campaign promises. He also constructs drug users as dehumanized and deserving targets, for instance referring to them as slaves with brains shrunken by drug use and beyond rehabilitation (Camacho & Montiel, 2021) and uses war metaphors to construct drug users and dealers as enemies threatening society to legitimize their exclusion or even extermination (Brasilino, 2019).
The UN OHCHR (2020) has critiqued the language in the official police circular and the verbal remarks from the president for encouraging the use of lethal force. They, along with other human rights, legal, and health groups, countered the administration's discourses by emphasizing respect for human rights and the need to treat the drug problem as a public health issue (Geronimo, 2017;UN OHCRC, 2016. Human rights groups (e.g. PhilRights, 2021) have also reported that the poor are most often the targets of tokhang, with the families of those killed in the drug war adversely affected. These critiques, in turn, have been echoed by the Catholic Church and academic institutions (Curato, 2017), with the Church adding arguments about the sanctity of life in their statements against tokhang (Valles, 2018).
Discourses from both the administration and its critics have been widely covered by mainstream media. However, news reports on incidents of drug-related killings have contributed to strengthening state discourses by privileging state sources and using "neutral" event-driven reporting that give the government the power to control the narrative (Soriano, David, & Atun, 2021). Among the most prominent discourses in these news reports is that of nanlaban 1 , which portrays killings as police self-defense because of "suspects" fighting back (Lamchek, 2017). News reports also heavily covered killings by masked vigilantes, labeling those killed by vigilantes as victims and others killed in police encounters as suspects. In drawing primarily from police sources, these reports increase the accessibility of state discourses that drive fear of illegal drug use and crime and dehumanize drug users victimized by tokhang (Soriano et al., 2021). Beyond mainstream media, social media has also become fertile ground for amplifying, modifying, and negotiating these varied institutional discourses. However, with the combination of social media algorithms, hired trolls sowing discord, and emotional content of news related to the drug war (Ong & Cabañes, 2018), some messages can be intensified and become more easily accessible depending on one's background, social context, and psychological characteristics.
With different sources and contexts for interactions, individuals can access, appropriate, and negotiate discourses about tokhang in various ways to create their own moral positions. Ethnography in urban poor areas (Warburg & Jensen, 2020) and analysis of Facebook comments (Camacho & Montiel, 2021;Hapal, 2019;Uyheng & Montiel, 2021) show the popularity of these institutional messages, most prominently state discourses on rightful punishment of drug personalities for the sake of community security. The killings may even be positioned as a moral means to achieve retributive and divine justice to legitimize tokhang, with this legitimation strengthened by constructing critics such as the political opposition and mainstream media as corrupt (Camacho & Montiel, 2021). Though discourses about social justice, the safety of family and neighbors, and religion can be relevant and used to oppose tokhang, these tend to be deemed subordinate to community security (Warburg & Jensen, 2020). Discourses of human rights were less consistently used, likely owing to constructions of drug personalities as non-human (Hapal, 2019;Warburg & Jensen, 2020). However, this limited use of human rights discourses can also be a function of sociolinguistic differentialsanalysis of Facebook comments revealed that regardless of position on the policy, community security discourses were often expressed in Filipino, whereas discourses about democratic integrity were in English (Uyheng & Montiel, 2021). In a country where language and social class are strongly intertwined, these observed differences in language use demonstrate the impact of one's social position and contexts of social interactions on the accessibility and relevance of particular discourses.
Along with the accessibility of certain discourses in relation to one's social context, some differences in support for the war on drugs can also be explained by individual-level cognitions (e.g., personal support for the president and perceived association between drugs and crime in Labor & Gastardo-Conaco, 2017; binding moral foundations, right-wing authoritarianism, and threat perception in Nerona, 2017) and affect (e.g., hatred and compassion in Labor & Gastardo-Conaco, 2017). Such internally-focused social psychological approaches can be enriched by qualitative discur sive approaches, as publicly-available discourses appropriated by individuals can influence their positions concerning this issue. Additionally, the study can complement online discursive studies discussed in this section. Compared to a Facebook comments section, the context of a face-to-face, semi-structured interview is more private and allows for negotiation and clarification between participant and researcher. Knowing individual participants' backgrounds also provides insights into potential sources of the discourses they appropriate and negotiate. These are especially important 1) From the rootword laban, meaning fight, nanlaban describes the suspect as having fought or resisted arrest.
given the assumptions in discursive research that language is produced and serves particular functions within different social interactions, and constructed from individuals' social, cultural, and historical context.

Moral Reasoning Constructed on the Basis of Harm
The range of available discourses on tokhang provides a variety of possibilities for constructing and negotiating its harmfulness. The Theory of Dyadic Morality, which argues that harm perception is the basis for moral judgments (Schein & Gray, 2016, can provide a useful starting point for understanding how harm is constructed. Fundamen tal to recognizing harm, and thus, immorality is the synthesis of elements of harm: an intentional agent, a vulnerable patient or victim, and the causality of harm. TDM also argues that acts are more likely to be judged as immoral when harm is perceived, and harm is more likely to be perceived once an act is judged as immoral. Initial discourses about policies that people are exposed to may thus have an impact on harm perceptions and moral judgments, and social discussion can contribute to a feedback cycle between moralization and harmification of certain acts (Schein & Gray, 2018).
Coming from the social cognitive tradition, research on TDM mainly uses quantitative experimental and correlation al methodologies that measure intuitive moral judgments (Schein & Gray, 2018). Despite the positivist approaches associated with the theory, it lends itself well to a qualitative discursive approach given its central argument that judgments of harm and immorality are constructed from the synthesis of elements of harm. Whereas the theory presumes that this construction occurs as an internal cognitive process, it may also be posited to take place at the level of social interaction through language. The use of a qualitative discursive methodology with interview data affords more flexible exploration of how various discourses are accessed and negotiated to produce different moral positions.
Given the premise that harm perceptions depend on constructions of "an intentional agent causing damage to a vulnerable patient" (Schein & Gray, 2018, p. 1), judging an action or a policy to be immoral would rely on identifying a vulnerable victim lacking agency, an agentic perpetrator with ill intentions, and clear causality between the agent's actions and harm experienced by the victim. The degree of vulnerability attributed to individuals and groups can be reduced by victim blaming and dehumanization (Schein & Gray, 2018), in turn reducing perceptions of harm and immorality of actions that cause damage to victims. Victim blaming increases agency and reduces vulnerability ascribed to victims as they are deemed dangerous and threatening, and thus deserving of harm (Putra, Holtz, & Rufaedah, 2018;Soares, Barbosa, & Matos, 2018). Dehumanization of victims makes them appear less able to feel suffering (i.e., less vulnerable), more easily justifying violence against them (Castano & Giner-Sorolla, 2006;Soares et al., 2018). Political leaders can perpetuate dehumanization and victim blaming to justify violence towards particular targets, for instance by constructing and reproducing negative presentations of victims to justify their oppression (Augoustinos & Every, 2007).
Similarly, it is not enough for agents to inflict damage on another for an act to be seen as harmful and immoral; they must be constructed as agentic and intentional. When the agent is seen as having no choice on the matter, relatively weaker than the target, or serving the greater good through "virtuous violence" (Schein & Gray, 2018, p. 27), the act is less likely to be perceived as immoral. In matters of social issues and policies especially, this can be influenced by political frames (Montiel & Shah, 2008), the use sanitizing language such as jargon and comparison with other agents inflicting greater harm (Soares et al., 2018), and threat perception (Thompson & Lee, 2004).

The Present Study
The study aims to understand moral reasoning about Duterte's war on drugs, specifically tokhang by asking the following research questions: (1) How do individuals construct their moral positions on tokhang? and (2) How do they negotiate harm to legitimize or contest tokhang? The study uses the Theory of Dyadic Morality (Schein & Gray, 2016 as a starting point to analyze how discourses on tokhang can be constructed around elements of harm and contribute to moral positions on this policy. However, it also extends TDM through the use of a qualitative methodology with attention to discourse, allowing for analysis of these elements of harm as combined, constructed, and negotiated by people to reason about their moral positions. This study, while focused on the more micro-level interaction of an interview and the individual psychological experience of moral reasoning, allows an analysis that makes connections to circulating discourses, thus situating moral reasoning in a larger constellation of social forces. Such an approach may be applied productively to the study of other policies, where discourses on elements of harm can be identified, crafted, and deployed to potentially shape moral positions and provide opportunities for discussion and negotiation on the basis of harm.

Method
The present study is part of a project with broader goals of understanding young adults' moral reasoning using thematic analysis with attention to discourse (Braun & Clarke, 2006, 2021Gill, 2000). It examines moral reasoning from a social constructionist perspective (Burr, 1995), allowing moral reasoning to be examined as constructed within social, historical, and cultural contexts through social interactions, and having implications for people's actions in their social worlds.

Participants and Recruitment
Non-probability sampling was used to recruit 12 urban Filipino young adults aged 21-29. The sample was limited to this age range given the goals of the broader study to examine moral reasoning in young adulthood; nevertheless, focusing on young adults may be informative as this life stage opens more opportunities for sociopolitical participation and influence compared to the younger years. Most participants considered tokhang to be morally wrong than not wrong, but as will be seen in the results, different exceptions and nuances arise in relation to these positions. All except one had completed tertiary education and almost half had some postgraduate education (see Table 1 in the Supplementary Materials for other participant details). The participants were recruited through referrals by sharing a research flyer among the first author's (DPO) networks via Facebook and personal communications.
To ensure ethical conduct of the research, detailed information about the study was given to potential participants upon initial contact. They were assured of confidentiality and anonymity of the data, informed about the risks and benefits of participation, and reminded of their rights as participants. After their initial agreement, these details were discussed again on the day of the interview before they officially gave their informed consent.

Data Collection
Data were collected by DPO using face-to-face, semi-structured interviews, with two interviews per participant (ranging in length from 36 to 111 minutes, average of 70 minutes/interview) from February to April 2019. Interviews were conducted in English, Filipino, or a mix of both, and held at locations convenient for the participants. Data for this paper were derived mainly from the first interview session where tokhang was one of three moralized issues discussed. The interview guide was adapted from Jensen's (2015) Three Ethics Reasoning Assessment (see Supplementary Materials), but the interview format allowed for flexible probing of responses. Participants were asked whether each issue is morally wrong or not wrong and why, about exceptions to its being wrong or not wrong, and their encounters with different arguments about the issue. Afterward, participants responded to counterarguments to their position to give more opportunities for negotiation of moral positions. The second interview discussed questions beyond the scope of the present study, but also included clarifications to their responses in the previous interview. After the interviews, participants were given refreshments and gift certificates as tokens.

Data Analysis
Interviews were analyzed using thematic analysis with attention to discourse (Braun & Clarke, 2006, 2021 with AT LAS.ti (Version 8.3.0; Scientific Software Development GmbH, 2018), following the steps outlined by Braun and Clarke (2006): (1) Familiarization with the data, (2) Coding, (3) Searching for themes, (4) Reviewing themes, (5) Defining and naming themes, and (6) Producing the report. This analysis drew insights from discursive methodologies, an approach that has previously been integrated with thematic analysis (Braun & Clarke, 2021). In the interest of transparency and rigor (Yardley, 2008), each step is detailed in the succeeding discussion.
Familiarization with the data was done through transcription and repeated reading. DPO transcribed sensitive sections of the interviews and reviewed all recordings and transcripts produced by trained student assistants 2 . DPO also wrote initial analytic notes during the transcription process and discussed these with the second author (MGO) to better clarify the directions of the analysis.
All responses relevant to tokhang were coded, and extracts could have multiple codes when containing multiple units of meaning. For instance, the extract "It's really noticeable that tokhang is pro-bourgeoise so it's the poor who are the targets. Drug lords have protectors, they're not made answerable and the victims who are always killed are the poor… it's an oppression…against the poor." was coded as "disadvantaged victims" and "not the only guilty ones". This coding process was done by DPO and codes were constantly reviewed, revised, and discussed with MGO as themes were created, keeping with the assumption that analysis is interpretative and reflexive (Braun & Clarke, 2021).
Initial themes were created from these codes, informed by but not limited to the elements of harm in TDM. Themes were reviewed for their coherence and revised recursively. In line with a discursive approach, the report included illustrative data extracts, linking the analysis with broader sociopolitical sources and discussing the implications of using certain discourses over others (Gill, 2000). Extracts have been translated and edited for clarity and conciseness, and participants' names have been changed to ensure anonymity.
Thematic analysis from a social constructionist and discursive approach assumes that reality is a social and cultural product built upon language and discourse (Braun & Clarke, 2006, 2021Burr, 1995). Thus, individuals' positions were analyzed as constructed from discourses available in the sociocultural context. Additionally, responses were not taken to represent participants' thoughts or beliefs but investigated as a form of action within a social setting, that is, reasoning about their position in the context of an interview. Such positions are not taken to be static but constantly negotiated and reconstructed as they interact with different discourses about the issue. The intention then is not to produce generalizations representative of Filipinos' positions on tokhang, but instead attend to the contents, organization, and functions of discourses used to reason about the (im)morality of this policy.
To gain trustworthiness (Merrick, 1999) as well as transparency and rigor (Yardley, 2008), we ensured that each analytic claim made was supported by data, and identified when points were specific only to one, some, or most participants. Data and codes were systematically organized in Atlas.ti (Version 8.3.0; Scientific Software Development GmbH, 2018) allowing for greater transparency, and analysis was recursive to contribute to research rigor (Yardley, 2008). Additionally, we acknowledge our positionality on the war on drugs: we strongly believe that it is harmful, immoral, and ineffective. Aware of this position, participants and their responses were treated with respect and curiosity, especially those with opposing positions, maintaining sensitivity to context in data collection, analysis, and writing (Yardley, 2008).

Results and Discussion
Regardless of their position, participants were seen to use the same logic in their moral reasoning on tokhang by identifying victims and agents, negotiating their vulnerability and agency, and establishing the causality of harm (Schein & Gray, 2016. We first discuss constructions of tokhang victims, particularly "drug personalities" and broader members of society, and follow this with an exposition of how participants negotiated their constructions of the police as the primary agents implementing tokhang. The last section tackles reasoning about intentions, means, and outcomes of the program, and distinctions between systemic and individual agents of harm. The relationships among these constructions are illustrated in Figure 1. 2) All transcribers signed a confidentiality agreement to protect the privacy of the participants.

Drug Users and Pushers: Victims or Victimizers of Society?
Moral reasoning on tokhang largely revolved around constructions of the policy's targets, presenting opposing perspec tives on their vulnerability and innocence on the one hand, and agency and guilt on the other (Gray & Wegner, 2009;Schein & Gray, 2016. Support for the policy was built partly on blaming tokhang victims by constructing them as criminals committing heinous crimes, highlighting their agency and guilt. This is evident in the following exchange with Edward 3 (M, 24): Edward justified tokhang by constructing some victims as criminals deemed to have caused irreversible damage to others, and thus deserving of death. However, he also expressed ambivalence, even pity, towards drug dependents, hinting at negotiation of their agency and guilt to judge their deservingness of harm. Still, what is clear here is blame towards victims of tokhang and shifting of their roles from victim to perpetrator to justify harm towards them (Gray & Wegner, 2009;Schein & Gray, 2016, echoing official discourses from the police and President Duterte regarding the criminality of drug users (Hapal, 2019;Lamchek, 2017;PNP, 2016). In highlighting the blameworthiness of those killed by police, the latter's actions and the policy are legitimized. Such patterns of victim blaming have also been seen in work on racism (Augoustinos & Every, 2007) and Muslim minorities targeted by violence (Putra et al., 2018).
One means for countering this victim blaming is the construction of tokhang victims as vulnerable victims of society with limited agency. Kat (F, 29) argued, for instance: I think since we offer the people no choice, we're not given basic services, no housing, no education, they don't have access, it's like we breed the criminals, right? It's like this system makes them criminals, and then we kill them for it.
Kat enumerated the different ways tokhang victims are disadvantaged and magnified their lack of agency by referring to them in passive terms and portraying them as having no choice ("the system makes them criminals, and then we kill them for it"). Echoing broader discourses about the difficult circumstances leading to addiction (e.g., Cristobal, 2016), Kat underscores tokhang victims' vulnerability and concludes that the harm they experience is morally wrong. Through these discourses, external and/or uncontrollable factors are brought in as explanations, facilitating the drug dependents' victim status (Schein & Gray, 2016. Thus, even before they were victimized by tokhang, they were already suffering from their circumstances, compounding their victimhood. This construction of vulnerable victims and tokhang's harmfulness are further strengthened by arguments that it systematically and disproportionately targets the poor. Patrick (M, 21) claimed, It's really noticeable that tokhang is pro-bourgeoise so it's the poor who are the targets. Drug lords have protectors, they're not made answerable and the victims who are always killed are the poor… it's an oppression…against the poor.
3) Names have been changed to protect participants' anonymity. By referring to tokhang as "an oppression against the poor", Patrick established the injustice of the policy, where the vulnerable are victimized and the real culprits, drug lords, are immune from punishment because of their relative power. Critics of the policy from different sectors have similarly branded this as anti-poor, and such discourses have become widely available at the height of the news coverage about its implementation (e.g., Aldama, 2018). Patrick's contrasting tokhang victims with drug lords also breaks down the logic of the deservingness of criminals killed in tokhang because other more guilty criminals are not punished.
The injustice of tokhang systematically targeting the poor can be countered by employing the language of reason, as did Edward (M, 24) who said, By ratio, and statistics and probability, there are really more addicts in [the lowest socioeconomic] class E, that's why it becomes anti-poor. So there will really be more addicts because there's a larger ratio of poor people.
Edward's statement downplayed the systemic victimization experienced by the poor and instead constructed it as but a natural consequence of the policy's implementation given the logic of "ratio, statistics, and probability". In this way, tokhang is constructed to be fair and impartial and the policy remains morally acceptable.
Overall, opposing constructions of tokhang targets lead to different moral positions, with arguments involving negotiation of victims' degree of vulnerability and agency (Schein & Gray, 2016. When victims are deemed more vulnerable, harm towards them is judged to be more immoral. When they are constructed as more agentic and less vulnerable, the situation can be judged as less immoral, and the victim said to deserve the harm experienced (Gray & Wegner, 2009). However, as the next section demonstrates, others outside of the immediate situation can also be considered victims and consideration of these others can affect the moral acceptability of a policy.

Innocent and Vulnerable Victims Multiplied: Families, Communities, and Society as Victims
Moral condemnation of tokhang was further argued by constructing more innocent and vulnerable victims beyond the supposed targets of police operations, magnifying its harmfulness. One argument focuses on the innocence of those killed in tokhang. Melanie (F, 28) shared, "based on what we see in the news, some of these policemen seem to be misinformed. There've been cases of mistaken identities. So that's what makes me angry about it. " She expressed her anger towards the killing of innocent victims and referred to news reports of innocent victims, including children (e.g., the Kian delos Santos case, see Hapal, 2019), being killed in cold blood. However, she also constructed the police to have been misinformed, potentially reducing their blameworthiness as their actions can be taken as mistaken rather than intentional.
Victims can also extend beyond those in the immediate encounter. Most prominent would be the families left behind, as Marian (F, 21) stated: …it's a whole, societal problem…in different communities. It's not just the person involved who's affected, they're whole families, so they're just taking away the life of one person suddenly, without really explaining or the family understanding why.
Marian argues that the harm caused by tokhang extends beyond the person killed -from families, to communities, and even society. Stories and images of the families left behind by victims of tokhang have featured prominently in the news landscape, with institutions such as the Catholic Church (Valles, 2018) using this as part of their critiques of tokhang, giving the public access to discourses about the extent of victimhood.
This harm to society brought about by tokhang was made even more apparent in Liana's (F, 29) reasoning: The reality that people are being killed, you can't separate it from the system of impunity…anybody can be a victim of it. So on the surface it may feel more peaceful? And more safe? But, if you make an enemy or someone gets pissed at you, then you could be next, or it could be your son or daughter or whatever. Yeah, it's not nice and it robs you of your ability to speak up against the government.
Liana directly addressed the discourse of community security by highlighting the threat of tokhang for ordinary citizens, regardless of innocence or guilt. She also referred to the culture of impunity along with the policy's impact on the right to free speech ("it robs you of your ability to speak up against the government"), echoing warnings from legal and human rights groups about the vulnerability even of ordinary citizens to tokhang (UN OHCHR, 2020). Appropriation of discourses that identify more victims of the policy thus magnifies its harmfulness and immorality (Schein & Gray, 2016. Strikingly, this attention to innocent victims and the extent of victimhood did not feature in the arguments of those in favor of tokhang, presenting an opportunity for amplifying this set of discourses to increase its accessibility. After all, it expands victimhood beyond the policy's (often vilified) targets and attends to larger communities affected by the individuals and events involved. However, aside from focusing on victims, constructions of agents and their intentions are also important in making sense of moral positions on tokhang.

From Agent to Victim: The Police Are Only Doing Their Jobs and Protecting Themselves
Central to moral reasoning about tokhang are constructions of police as agents who implement the policy. Though identified as the agents in most instances, the police were not always constructed as intentional and agentic in their actions towards tokhang victims. Prominent among several participants, regardless of their stance, is the nanlaban discourse, referring to the directive to shoot to kill if "suspects" fight back. Indeed, many of them cited this as an instance when tokhang was not morally wrong. For instance, Sarah (F, 24) said, "let's say you defended yourself, it's either your life or the life of the offender. Because it's a matter of life or death. " What is striking in this statement from Sarah, who considered tokhang to be morally wrong, is the use of the term "offender" to refer to tokhang victims and the accessibility of self-defense as an exception to tokhang being morally wrong. This reflects the nanlaban defense created and communicated by the administration and amplified by the media. With self-defense being less controversial than killing in cold blood, this creates a gray area for legitimizing the use of deadly force (Lamchek, 2017;Soriano et al., 2021). This also portrays the police as being more vulnerable and less agentic, deeming them less culpable and tokhang less immoral (Schein & Gray, 2016. Additionally, it allows for roles to be switched, as police become potential victims and "suspects" become the perpetrators, providing a mitigating feature such as self-defense to make an obviously harmful action permissible (Schein & Gray, 2018). Justifications of self-defense have been seen in other work on police violence (Soares et al., 2018;Thompson & Lee, 2004), and can extend to other policies involving the use of force on people constructed as dangerous threats.
Killing "suspects" in self-defense can be further legitimized by constructing it as part of protocol and invoking uncertainty about causality. This is exemplified in the following extract from Michael (M, 29), Based on the news, before during the first launching many people died because of the claims that they fought back and there were shootouts. And other people say that these were extrajudicial 4 . So I don't know what really happened… But for me, I think they've laid low already, there aren't as many [killings]. Because now they have a protocol on how they'll manage if there's a shooting…now they've established that tokhang means they talk to them, invite them [to surrender], right?...If you're not doing anything, why would you resist, right?
In the first part of his statement, Michael presented two different sides ("based on the news" and what "other people say") and introduced uncertainty ("I don't know what really happened") to make his position defensible. This suggests the effectiveness of the administration's strategies to muddle public judgments about the killings (Lamchek, 2017). This lack of clarity can mitigate judgments of harmfulness and immorality as the agent is no longer as clearly responsible for harm inflicted on victims (Schein & Gray, 2016. Additionally, Michael introduced the possibility that some aspects of tokhang could have been wrong initially, but referred to protocols and the publicized intent of the program to encourage voluntary surrender (PNP, 2016) to construct it as morally acceptable. Thus, the nanlaban discourse creates a 4) Many of the tokhang-related deaths have been labeled as extrajudicial killings (EJKs) carried out in vigilante-style executions (Hapal, 2019;UN OHCHR, 2016). loophole for harms caused by police, as they are constructed as acting as expected of their role. Similar strategies have been observed in police justifications of violence, where jargon is used to sanitize the harmful action (Soares et al., 2018). What is striking in the case of tokhang is how this terminology has seeped into public discourse and provides discursive handles to justify the killings. The end of his statement also echoes an often-used argument supporting tokhang: "you have nothing to fear if you are not doing anything wrong" (Hapal, 2019, p. 197). Again, this blames the victim and absolves the agent. In spite of the strength of the nanlaban discourse, it can be rejected as Carlos (M, 23) did: It's a hard line for me that it's not okay for the police to kill even if their lives are at risk because their role as law enforcers should have strict requirements for skills. So they should catch suspects for example without killing them and they should be prepared for any scenario…there's a limited budget for equipment, for training, but I think that's where you see government priorities. Because if they don't value life, if many people die…then that's fine with them.
Carlos expressed an ideal for the police and government to meet and attributed their inability to meet these standards to low prioritization because of devaluation of human lives. This provides an alternative discourse to nanlaban by highlighting the government's responsibility to train police effectively to prevent even killing in self-defense. This also creates a culpable agent beyond the police: a government that devalues human life, alluding to the president's dehumanizing remarks that encourage and justify killing (Camacho & Montiel, 2021;Francisco, 2016). In the succeeding section, the implications of blaming a system and its leaders in contrast to individuals directly involved will be discussed further.

A Good Program With a Few Bad Apples vs. A Rotten and Ineffective System
In assessing the causality of harm, constructions of the intentions, means, and outcomes of tokhang contributed to participants' moral judgments. Distinctions in constructions of tokhang at the level of policy and implementation allow for nuances in assigning credit and culpability. Specifically, some participants negotiate their approval of the policy and criticisms against its violent outcomes by constructing tokhang as a well-intentioned and effective program with a flawed, objectionable implementation. For instance, according to Patrick,

Tokhang is a good project actually. But what makes it very, very, very wrong is the extrajudicial killing. I would agree if tokhang would stay as knocking on people's doors to check who are the drug users.
In his response, Patrick distinguished between the official purpose of tokhang of going from house-to-house to encour age voluntary surrender and rehabilitation (PNP, 2016), and the killings perpetuated in the implementation of the policy. As a seminarian and church worker, Patrick had opportunities to witness rehabilitation programs implemented by religious organizations, contributing to his positive appraisal of some aspects of the policy. In another part of his interview, he also conceded that tokhang contributed to peace and order through the rehabilitation aspects of the program, appropriating discourses of community security as a benefit of tokhang (Soriano et al., 2021;Uyheng & Montiel, 2021). Several other participants similarly brought up this claimed outcome of the policy, whether or not they considered tokhang to be morally acceptable. Thus, even if the killings are condemned as morally wrong, the policy itself becomes difficult to contest as a whole because of constructions of good intentions and outcomes built on widely available discourses.
A few participants, however, critiqued the program for being ineffective at actually bringing about peace and order and argued that there are better ways to fight the drug problem. According to Carlos, If we treat drug addiction simply as a problem that needs to be eradicated without thinking about why it exists in the first place and the people suffering from drug addiction are treated as less of humans, then nothing will happen. Many people will die and nothing will happen.
As opposed to Patrick's focus on rehabilitation, Carlos emphasized prevention where addressing root causes of drug addiction would be more effective in eradicating drugs. He also critiqued the dehumanization of drug dependents (similar to that found in the president's talk, see Camacho & Montiel, 2021), saying that those suffering from drug addiction "will die and nothing will happen" if they "are treated as less of human". In other parts of the interview, he constructed drug addiction as a health issue, emphasizing that the criminal approach used in tokhang will not effectively address the problem. In other words, Carlos constructed tokhang as harmful and producing none of its supposed benefits. These discourses about public health and prevention are likely to have been accessible to Carlos given his background in development and human rights work and liberal education.
Constructions of the intentions underlying tokhang and the means used for implementing it can also come together in critiquing the policy itself. Melanie said, for instance, I think it's morally wrong, because based on the news that we have, most of these people don't even go through due process, regardless of whether they use drugs or they push drugs. At the end of the day I think everyone has the right to be heard. Especially if you don't know them, these people are basically strangers. And looking at the larger context, some people argue that tokhang is done for political gain and all that. So for me, that's why there's so many levels of wrong when it comes to tokhang.
Melanie's argument opened with condemning tokhang on the grounds of its violation of the right to due process and constructed this as inalienable ("regardless of whether they use drugs or they push drugs", "everyone has the right to be heard"), drawing from institutional discourses about human rights violations linked to tokhang (Curato, 2017;UN OHCRC, 2016Valles, 2018). The second part of her response compounds the policy's wrongness by linking it to self-serving intentions. This carries resonances from reports revealing corruption in the drug war, including police blackmailing and framing drug "suspects" and selective punishment of small-time drug users and dealers, informants about the drug trade, and even political opponents (Lamchek, 2017). Thus, harms to victims become even more immoral when combined with these intentions of people in power. What is especially striking in the data is that participants who critiqued the policy at the systemic level were those with postgraduate education in liberal universities. Thus, it may also be the case that this social milieu occupied by the participants along with their identities make such discourses more accessible and relevant to them.
In contrast to those who implicate larger forces, tokhang supporters can blame individual rogue agents who use the policy to their advantage. Edward for instance claimed, After some time since tokhang started, they said that there was proof showing that the evidence was just planted, which is wrong because…if you were going to arrest someone, it should be as neutral as possible. You don't arrest someone because you have personal gain there, or someone just told you. You arrest someone for the good of the community, not for the good of yourself.
In this account, Edward referred to corrupt police in individual terms ("you") but did not condemn tokhang. There was no mention of larger factors enabling this to happen, suggesting that it was more a product of individual rogue policemen. Indeed, in high-profile cases of police killings, the administration has occasionally punished individual policemen and painted them as isolated incidents (Hapal, 2019;Lamchek, 2017), allowing the policy and its enabling systems to remain morally defensible.
Departing from discourses blaming the government and the police are accounts attributing the killings to unidenti fied vigilantes, further reducing condemnation of tokhang as a policy. Such mention of vigilante killings came solely from Richard (M, 25), who initially said that tokhang was morally wrong but clarified his stance as such: "The only ones who carry out tokhang are those masked men on motorcycles. When it's the police who act, often only those who fight back (nanlaban) are killed. " As mentioned, Richard initially said that tokhang was morally wrong, but as the interview progressed, it became clear that his conception of tokhang referred to the vigilante killings rather than the policy itself. It is also important to note that he was the only participant living in an urban poor area where tokhang operations were often carried out and who did not complete tertiary education. Thus, his construction of tokhang may have been built on experiences and informal conversations within his community and news reports which heavily covered vigilante killings and framed police killings as encounters with those who resisted arrest (Lamchek, 2017;Soriano et al., 2021). Through this distinction, the police and administration are thus absolved of blame.
Put together, discourses blaming individuals such as unidentified agents and rogue policemen obscure the systems that perpetuate violence, as the focus is on the individuals on the ground rather than leaders and systemic forces as agents of harm (Brown, 2007).

Conclusions
The study examined constructions of harm in moral reasoning about Duterte's war on drugs. We conducted thematic analysis with attention to discourse to analyze interviews with young adults, using the Theory of Dyadic Morality as a starting point for making sense of constructions of harm. Participants' reasoning about tokhang showed different constructions of intentional agents and vulnerable patients serving as the basis for their moral positions. However, the notion of a victim and perpetrator is not rigid nor bound to individuals, and these can be negotiated in social interac tions. These constructions and negotiations were built upon broader discourses deployed in the sociopolitical context of urban young adults during Duterte's administration, with competing discourses widely available in mainstream and social media from the official policy, the president himself, and both his supporters and critics. However, individual contexts and characteristics also contributed to variations in the accessibility and relevance of certain discourses and the resulting positions constructed and negotiated.
Moral condemnation of tokhang emphasized its victims, typically drug "suspects" killed in police operations and vigilante-style shootings as vulnerable and lacking in agency. This was further magnified by including other victims such as the families of victims and Filipino society as a whole, along with constructions of a self-serving intentional agent, typically the police and the government, systematically violating human rights. With this construction following the template of a vulnerable patient, intentional agent, and clear causality of harm (Schein & Gray, 2016, tokhang was judged as morally wrong. Notably, most of those who expressed absolute moral condemnation of tokhang were educated in liberal universities vocally opposed to tokhang. This context and background thus can increase the accessibility and relevance of liberally-oriented discourses that emphasize human rights and attend to systemic social forces. On the other hand, moral justification of the policy emphasized the criminal involvement of tokhang victims, shifting their role from victims to agents of harm to make them appear deserving of the harm they receive (Gray & Wegner, 2009). At the same time, constructing the police as merely acting in self-defense makes them appear less blameworthy by reducing their agency, and creates an opportunity for them to be portrayed as victims. Identifying agents of tokhang detached from the administration (e.g., rogue corrupt policemen, unidentified vigilantes) also allowed for condemnation of the killings while continuing to accept the policy. Tokhang was further justified by constructing it as having benevolent intentions and outcomes of community security, making it appear to be "virtuous violence" (Schein & Gray, 2018, p. 27) where harm towards deserving victims is considered helpful or even necessary. All of these serve not just as discursive handles for those who support the policy, but also create space for ambivalence among those who may have misgivings about it. The nanlaban discourse along with the publicized intent of the program to rehabilitate drug dependents were especially potent in contributing to this ambivalence. However, as evident in the analysis and other work on tokhang, this can also be contested through arguments about the ineffectiveness of the policy in achieving its supposed outcomes and unveiling corruption behind its implementation (Lamchek, 2017).
In contrast to those who most strongly opposed tokhang, the shared characteristics of those in support of the policy were not as well-defined. However, what was clear in their responses was how their constructions of tokhang essentially echoed the president's and police's messages. The results suggest that public support for abusive, immoral policies may be secured by ensuring the primacy and dominance of discourses that undermine the victim's vulnerability, turn the agents into victims, communicate positive intentions and outcomes, and veil the causality of harm. As found by others, these discourses may have more room to flourish in contexts with flawed institutions that make effective democratic solutions seem untenable and dampen the relevance of human rights discourses (Thompson, 2016;Uyheng & Montiel, 2021).
Overall, these results demonstrate how moral reasoning about a policy can be constructed around negotiated harm based on existing and available discourses around the actors and events involved. In using TDM as a starting point, the analysis provides specific action points for communicating clear moral discourses for opposing harmful authoritarian policies: highlighting their targets' vulnerability, identifying other vulnerable victims beyond those directly involved, unmasking the intentions and agency of the perpetrator, and reducing ambiguity in the causality of harm. The research also highlights the utility of interviews and thematic analysis with attention to discourse to examine how moral reasoning is negotiated, nuanced, and shifting. It also shows how, even in an interview setting, moral positions are built upon existing discourses. Participants often prefaced their responses with references to external sources ("based on the news", "they said", "some people argue"), suggesting how they drew from various arguments available and relevant to them. However, they exercised agency in producing and negotiating their positions by drawing from, combining, and reconstructing arguments provided by institutional sources from different sides of the issue (Camacho & Montiel, 2021). This capacity to produce different positions, however, also depends on the accessibility and relevance of particular discourses to participants, likely as a function of their respective characteristics and contexts. When discourses are more accessible and individuals have more opportunities articulating their position, they are able to adopt positions more consistently and easily. The goal then is to find ways for discourses more consistent with a just society to multiply and gain more traction and to promote access to respectful deliberation of such discourses beyond individuals' echo chambers (Curato & Parry, 2018).

Limitations and Recommendations
As part of a broader developmental psychology project on Filipino young adults' moral reasoning, this study has its limitations. Aside from a narrow age range, all participants except one have tertiary or postgraduate education and are working in corporate or professional occupations. Given sociolinguistic differences in discourses about tokhang (Uyheng & Montiel, 2021), it is likely that participants' backgrounds contributed to limitations in the range of discourses accessible and relevant to them. Additionally, most expressed moral condemnation of tokhang, limiting access to the discourses and negotiations of those strongly supportive of the policy. To address these limitations in positions, the interviews were conducted in a way that allowed for a range of discourses to be explored and for exceptions and ambiguities to surface.
These limitations may be taken as threats to generalizability. However, generalizability, at least in a statistical sense, is not a goal of qualitative discursive research. Rather, discursive research aims for flexible generalizability by identifying and analyzing rhetorical strategies applicable to other contexts (Goodman, 2008). Although the research focuses on the specific topic of Duterte's war on drugs with urban, mostly highly educated Filipino young adults, the study works towards flexible generalizability by using TDM as a starting point for identifying important elements of discourses used to construct moral positions. Future work may examine morally divisive issues in other sociocultural contexts and with more diverse samples to provide further support for and nuancing of the claims made in this study (Goodman, 2008).
Analyzing reasoning about tokhang through the lenses of TDM may also tell us something about the limits of human moral judgments: it is much easier to identify and blame the agent causing harm in the immediate situation than the broader systemic forces and more distal, powerful agents pulling the strings (Brown, 2007). The participants who did articulate this were those who most strongly opposed the policy and had postgraduate education. For the most part, individual blame still predominated. As such, interdisciplinary approaches may provide greater clarity about and interventions for different units of analysis beyond the individual.