Governing the Issue: Strategic Delayed Moralization of AI in Law Enforcement

Step 1: Historical Reconstruction of the Issue Life Cycle and Key Events

We selected sources in the period ranging from 2014 to 2025. We chose this period of time to start our analysis in correspondence with the appearance of the first public scrutiny to ShotSpotter. In our analysis, we began by creating a detailed chronology of events and identifying key stages in the issue’s life cycle: emergence, growth, maturity, and resolution (Mahon & Waddock, 1992). We identified transitions between stages with specific events that resulted in changes in public discourse such as increased attention around the issue and the organization, pressure from stakeholders and convergence of debates around related concerns (Bigelow et al., 1993).

Step 2: Analysis of Strategic Ambiguity and Public Discourse Within Issue Stages

We conducted a detailed analysis of ShotSpotter’s communications and actions during each stage of the issue life cycle. Our approach combined qualitative content analysis (Krippendorff, 2018) with grounded theory techniques (Corbin & Strauss, 2014) to identify and categorize different types of strategic ambiguity. First, two of the authors performed an open coding of all ShotSpotter’s public communications, separately, to identify ambiguity strategies (first-order categories) and types (second-order categories) in ShotSpotter communications. The two authors then met to resolve discrepancies. For what concerns the public discourse sources, we followed Leifeld (2017), categorizing the sources according to (a) number of concerns raised; (b) nature of concerns (factual/technical vs moral); (c) degree of integration of different concerns (low, medium, or high); (d) actors involved; and (e) goal of actors involved in the debate. We analyzed these using a temporal approach, to understand how public discourse changed in different issue stages (Table 1).

OPEN IN VIEWER

Table 1.

Public Discourse in Law Enforcement

Issue stage	Number of concerns	Nature of concerns	Degree of concern integration	Actors involved	Goal of public discourse
Issue emergence	Low	Factual/technical	Low	Local media outlets, independent activists, or auditors	To question technical attributes of AI in law enforcement (ShotSpotter)
Issue growth	Medium	Both factual/technical and moral	Medium	Local media outlets, local unions, and activist organizations	To identify the moral implications of AI in law enforcement
Issue maturity	High	Moral (factual/technical concerns are moralized too)	High	National media outlets, unions, and civil liberties organizations	To embed moral implications in the understanding of the tool

Step 3: Integration of Issue Life Cycle and Strategic Ambiguity Perspectives

We integrated the issue life cycle framework with the ambiguity strategies. We created a matrix that mapped the identified ambiguity types and strategies (second- and first-order categories) against the stages of the issue life cycle and outlined patterns between issue evolution and ambiguity strategies. Furthermore, we identified the constraining mechanisms embedded in the transition to a new stage of the issue, which reflected the developments in the moralization of the issue.

Step 4: Development of a Process Model

We synthesized our findings to develop a process model illustrating the dynamic relationship between issue life cycle progression, issue moralization, and strategic ambiguity evolution. We distinguish between theoretical mechanisms that transfer across contexts and case-specific manifestations. The generalizable aspects of our analysis include the staged progression from technical to moral recognition, the emergence of constraining mechanisms as moralization advances, the evolutionary pattern of ambiguity strategies, and the process of strategic delayed moralization as an organizational response over time, while context-specific elements include the particular concerns raised (e.g., gunshot detection accuracy), the specific stakeholder constellation, and the timing of stage transitions.

Constraining Mechanism in the Emergence Stage of the Issue: Questioning

As the issue transitioned from dormancy to emergence, a constraining mechanism we call questioning came into effect. The emergence stage was a first step in the moralization process of AI in law enforcement, as societal actors began to question the foundational assumptions underpinning ShotSpotter’s existence. The Silvon Simmons case first challenged this faith, bringing scrutiny not only to the technology’s reliability but also to its legal implications. Law enforcement officials, legal experts, and investigative journalists all raised doubts. A San Antonio PD representative noted that “About 80% of the times when ShotSpotter was activated, police could find no evidence of a shooting at the scene” (Public-SAPD-2016). Further doubts were raised by the National Association of Criminal Defense Lawyers (NACDL) which questioned whether “the sounds [are] always gunshots or are they sometimes vehicles backfiring or fireworks?,” noting that “Preliminary reports suggest a level of geographic precision that does not withstand the company’s own detailed analysis” (Public-NACDL-2018). Similarly, a Voice of San Diego investigation found that “the technology is not living up to the company’s own selling points,” and went on to note that “In four years, police have made two arrests while responding to a ShotSpotter activation . . . [only] one of those arrests was directly related to the activation” (Public-Voice San Diego-2020). This low yield – just one relevant arrest out of 584 alerts – reflected serious questions about the system’s practical value.

This questioning destabilized ShotSpotter’s framing of technology as infallible, pressuring them to respond by strategically employing ambiguity to delay future moral scrutiny.

Strategic Ambiguity in the Emergence Stage of the Issue: Unclarity

In response to technical questioning, ShotSpotter moved from silence to employing ambiguous communication. At this stage, the form of ambiguity employed is labelled unclarity, which refers to maintaining surface-level transparency while deliberately avoiding definitive statements that could expose it to moral accountability. Unclarity works by offering vague, technical (Crilly et al., 2012; Eisenberg, 1984) or incomplete information that creates the appearance of openness while leaving key concerns unresolved. Rather than confronting implications raised by the Silvon Simmons case, ShotSpotter selectively released complex technical explanations that were highly fact-based, yet intentionally unclear to non-technical audiences. For example, ShotSpotter kept relying on metrics like detection accuracy, or on the essential contribution of their skilled and well-trained analysts. However, the organization never clarified how accuracy was measured, and while referring to employees’ training as “comprehensive,” they never disclosed what this entailed. When Chicago data showed police found no evidence of gunfire after 88% to 89% of ShotSpotter alerts, the company disputed that this proved inaccuracy, arguing that “someone can shoot a gun but leave no evidence behind” (Corporate-2016b). This vague assurance avoided quantifying the system’s false positive rate. CEO Ralph Clark admitted that the 80% figure was a warranty rather than a real-world metric.

Under oath in a 2017 court proceeding, ShotSpotter forensic analyst Paul Greene offered an evasive explanation acknowledging that “our guarantee was put together by our sales and marketing department, not our engineers. We need to give them [customers] a number . . . We have to tell them something. [. . .] It’s not perfect. The dot on the map is simply a starting point” (Corporate-2017c). When critics raised cost-effectiveness concerns, the company reframed the goal of the technology rather than addressing the issue. For example, CEO Ralph Clark argued that the system was “not developed to lead to arrests” but to help intervene at the origin of gun violence, since “only a small number of individuals are responsible for most of a city’s gunfire and any tools available to get those folks off the street are important” (Corporate-2017b). By emphasizing such an intangible benefit (potentially deterring shooters) and invoking a general principle, this response gave a vague assurance of value instead of directly tackling questions about cost-effectiveness or concrete outcomes (like arrest or conviction rates). Through these strategies of unclarity, ShotSpotter actively delayed the moralization of the issue, reinforcing the narrative that the technology was a neutral, evidence-based tool rather than a socially contested object of concern.

Constraining Mechanisms in the Growth Stage of the Issue: Identification

As the issue entered its growth stage, a new constraining mechanism emerged that we call identification. Unlike questioning, which focused on technical functionality of AI technologies, identification marks a step forward in the moralization process, where stakeholders started to connect ShotSpotter to structural issues such as racial injustice, civil liberties violations, algorithmic bias, and the disproportionate targeting of minorities. These associations were expressed through class action lawsuits and investigations that exposed the potential of the technology to cause harm as an opaque AI tool in policing. Chicago racial justice advocates and local media emphasized how the technology reinforced over-policing in specific communities. ACLU noted that ShotSpotter was “deployed overwhelmingly in communities of color, which already disproportionately bear the brunt of a heavy police presence,” and warned that this “can distort gunfire statistics and create a circular statistical justification for over-policing in communities of color” (Public-ACLU-2021). Community organizers denounced the city’s spending priorities, stating, “This is not a structural investment in the community. . .While ignoring these needs, the city chooses to invest in police surveillance” (Public-BlockClubChicago-2021). In San Diego, residents feared the system would heightened the risk of police violence. One community organizer stated, “Even if ShotSpotter did work, I wouldn’t want this type of technology used in my neighborhood, because it leads to police being actively ready for an armed confrontation” (Public-WTTWNews-2021). Privacy concerns were also raised at the local level. In Oakland, civil liberties groups warned that “placing live microphones in public places raises significant privacy concerns” (Public-ACLU-2021), especially given the potential for these devices to capture conversations beyond gunshot detection. Meanwhile, in Cleveland, the city’s Community Police Commission criticized the lack of democratic oversight, noting they “have found no evidence that these technologies were implemented. . . where the entire community was engaged. . . and their values were represented” (Public-SignalCleveland-2022). Overall, the mechanism of identification is constraining because it creates moral challenges that an organization can no longer delay through unclarity about its functionality and technical features. For ShotSpotter, this shift meant that technical ambiguity and vagueness were no longer sufficient, forcing them to adapt their ambiguity strategy once again in an effort to delay the growing moralization of the issue.

Strategic Ambiguity in the Growth Stage of the Issue: Inconsistency

In response to the increasing moralization of AI in law enforcement, ShotSpotter adopted a new form of strategic ambiguity, which we label inconsistency. Unlike unclarity, which obscures through technical complexity and vagueness, inconsistency creates ambiguity by contradictory (Ashforth & Gibbs, 1990) or changing narratives (Suddaby & Greenwood, 2005) that selectively align with the values or expectations of different stakeholders, without committing to a unified moral position. Throughout this phase, the company’s mission changed from “reduce gun violence and improve the police-community relationship” (Corporate-2020c), to “protecting communities and improving public safety by providing equal protection for all” (Corporate-2020b), to “a proven and cost-effective means of supporting law enforcement” (Corporate-2021b), and finally “Helping to save lives” (Corporate-2021c). Inconsistency is particularly visible in how ShotSpotter frames its role and responsibility depending on the audience. In formal, legal, or contractual settings, the company distances itself from the consequences of its alerts: “ShotSpotter does not warrant that the use of its services will result in the prevention of crime, apprehension or conviction of any perpetrator of any crime, or prevent any loss, death, injury, or damage” (Corporate-2022d). However, in other corporate communications, the company frames itself as an active agent in crime prevention and justice: “ShotSpotter supports law enforcement in reducing gun violence, and it increases the chances of leading to an arrest” (Corporate-2021c). These contradictory claims allow ShotSpotter to shield itself from liability while appealing to stakeholders who value public safety contributions. Similarly, the company alternated between claiming ownership of alert verification process: “All ShotSpotter alerts are reviewed and confirmed by professionally trained acoustic experts located in our Incident Review Center, operating 24/7 to ensure accuracy before alerts are sent to law enforcement” (Corporate-2020a), and distancing itself from the outcome: “ShotSpotter has no control over how law enforcement agencies choose to respond to the alerts they receive” (Corporate-2021c). In this way, inconsistency allowed ShotSpotter to manage the constraining force of identification by fragmenting the moral debate, responding separately to critics, thereby delaying the moralization process.

Constraining Mechanism in the Maturity Stage of the Issue: Embedding

By the maturity stage, the ShotSpotter debate had moved beyond isolated controversies and began to converge into a broader, more unified public understanding of the moral challenges posed by AI in law enforcement. This convergence constrained ShotSpotter’s strategic options as, with a unified moral frame now shared across stakeholders, the organization could no longer deploy inconsistent narratives without contradictions being immediately exposed. We call this constraining mechanism embedding. Reports, campaigns, and formal communications moved concerns into mainstream political arenas. As embedding progressed, the issue was no longer just about ShotSpotter as a product, but about what its use symbolizes: a systemic moral problem in the way AI is applied in law enforcement. Public officials, civil liberties organizations, and national media all adopted this moral frame. U.S. senators urged investigation of ShotSpotter’s “contribution to unjustified surveillance and over-policing of Black, Brown, and Latino communities” (Public-Senator-2024). The ACLU cautioned that ShotSpotter “entrenches racially biased policing and surveillance” (Public-ACLU-2024), while EPIC stressed that despite “mounting evidence of ShotSpotter’s discriminatory impact, there is no indication that its Title VI compliance has ever been seriously assessed” (Public-EPIC-2023). National media described ShotSpotter as “a surveillance technology” marked by controversy over being “allegedly inaccurate, ineffective, and even biased” (Public-NPR-2024). Overall, across federal reports, advocacy group statements, and media investigations since 2021, a pattern emerged: moral and civil society considerations were embedded in how people evaluate AI policing technologies and were now influencing high-level decision-making. This created a new constraint for the organization because the moral frame is now fixed and shared across different actors. For ShotSpotter, this resulted in the need to change its ambiguity strategy again to stop the moralization process from becoming regulatory intervention.

Strategic Ambiguity in the Maturity Stage of the Issue: Moral Splitting

In this phase, ShotSpotter adopted a form of strategic ambiguity we call moral splitting. Moral splitting refers to the manipulation of moral boundaries (Benford & Snow, 2000) not by denying ethical concerns, but by reframing them through emotionally charged, simplified, or strategically personalized narratives, making it difficult to understand good and evil. Rather than engaging with moral criticism in its complexity, ShotSpotter began constructing a binary moral logic: either you support ShotSpotter, or you accept continued gun violence and death. A central component of this strategy was the use of personal moral testimonials, often people from minority communities – those most frequently invoked by critics as being harmed by ShotSpotter adoption. The company amplified voices claiming that “ShotSpotter saved my life” or “we need this technology to protect our neighborhoods” (Corporate-2024d). Moral splitting is also achieved by attacking the accusers (Coombs, 2007) with the purpose of undermining their validity. Specifically, ShotSpotter framed the debate as a stark moral choice between using its technology and allowing unchecked violence. In a 2022 press release, the company warned that the “outrageous allegations created a false narrative that undermines the important work ShotSpotter does every day to help combat the gun violence epidemic” (Corporate-2022c). Additionally, the company stated: “Who wouldn’t want to save lives? [. . .] Who wouldn’t want police rapidly alerted to shootings?” (Corporate-2022c), casting doubt on the motives of anyone opposed. Portraying critics as liars or misleaders was also part of the strategy. CEO Ralph Clark dismissed opposing claims as “false, misleading and specious statements by so-called ‘critics’” (Corporate-2024k), while official communications described critiques as “bogus and dishonest claims advanced by opponents” (Corporate-2024i), and “inaccurate [and] reflect[ing] a fundamental ignorance of policing” (Corporate-2024i). By casting criticism as lies and “mischaracterizations . . . twisted by some media” to “mislead the public” (Corporate-2024b), the firm delegitimizes its critics as dishonest actors undermining community safety. When Chicago’s mayor moved to end the contract, Clark argued “he would choose politics over public safety” (Corporate-2024k), delegitimizing concerns as political maneuvering rather than genuine worry about community harm. Moral splitting was a direct response to the constraining force of embedding adopted by ShotSpotter to strategically enter the moral conversation on its own terms.

Constraining mechanism in the resolution stage of the issue: Codification

As the issue enters resolution, the key constraining mechanism is codification, which entails the translation of moral criticism into specific rules that reduce interpretive flexibility for both societal actors and the organization. Codification works by embedding either moral concerns or the organization’s claims, or a compromise between the two, into enforceable prescriptions that define what is legitimate and what is not. In the ShotSpotter case, this mechanism becomes visible when controversy is no longer negotiated through competing claims by societal actors, but through formal rules that specify the conditions under which the technology can be deployed, if at all. Once AI in law enforcement is treated as a regulated surveillance infrastructure, ambiguity is less useful because agreement is granted- or withdrawn-through rule-based authorization. The BridgeDetroit coverage quotes the appellate opinion’s admonition that “procedural safeguards cannot be ignored or downplayed by government actors as mere technicalities,” and that “strict compliance with procedural safeguards” may be necessary “to ensure that technology serves the people” (Public-TheBridgedetroit-2025). In this way, regulation constrains organizational action by shifting the evaluation of the technology from a moral perspective to a matter of procedural and legal authorization, making strategic ambiguity ineffective.

Strategic ambiguity in the resolution stage of the issue: Absent

In this phase, strategic ambiguity becomes useless because the moralization process has been stabilized into enforceable constraints and decision rules, and even potentially counterproductive because regulation requires transparency and defines precise boundaries (Oliver, 1991). AI in law enforcement is now recognized fully in its moral implication, and organizational communication is compliance-based. This change is reflected in post-2024 organizational discourse that emphasizes transparency, governance architecture, and legal-admissibility framings rather than the ambiguity. For example, in its 2025 SEC filing, ShotSpotter highlights legal-institutional validation directly, stating that its forensic analyses have “survived dozens of challenges . . . under the Frye and Daubert standards of admissibility,” which frames the issue as something settled through formal regulation rather than through contest over moral significance. Furthermore, it explains how “AI outputs are always subject to human review” to ensure alignment with “ethical standards and regulatory compliance” (Corporate-2025c).

Strategic Delayed Moralization

Our article advances understanding of how AI technologies transform from neutral artifacts into objects of moral significance While current literature on AI in society has stressed how AI technologies are perceived as neutral and the risks this entails (den Hond & Moser, 2022; Moser et al., 2022; Lindebaum et al., 2020), we develop a process model theorizing how organizations strategically try to delay this moral recognition.

Strategic delayed moralization operates by exploiting and maintaining the initial faith in technology (Hilgartner & Bosk, 1988) that surrounds AI adoption. During ShotSpotter dormant phase, the technology benefited from its status as what Lindebaum et al. (2020, p. 248) identify as a “supercarrier of formal rationality” – an artifact presumed neutral, objective and apolitical. Our analysis shows that this presumption creates a strategic window during which organizations can embed technologies into society before moral implications are recognized. Strategic delayed moralization captures how organizations actively maintain this window by deploying strategic ambiguity to delay the transition from technical to moral recognition. This shows a fundamental insight for understanding AI in society: the gap between AI adoption and moral recognition is not accidental but actively produced and maintained through organizational strategy. This reveals the organizational dimension of delayed moral recognition (den Hond & Moser, 2022; Moser et al., 2022), while simultaneously showing that organizations fight for the control of this process with societal actors. Importantly, we also show how this process may eventually fail, explaining why moral recognition of AI technologies eventually occurs despite organizational resistance (Lindebaum et al., 2020). As moralization advances, it generates constraining mechanisms that progressively narrow the interpretive space available to organizations, ultimately compelling the moral recognition they sought to delay.

Moralization-Driven Constraints on Strategic Ambiguity

Our article extends issue management theory by positioning moralization as the driving force in issue evolution, rather than treating it as a byproduct of attention or stakeholder pressure. While previous research has examined organizational strategies to manage emerging issues (Mahon & Waddock, 1992) and avoid moral scrutiny (Carlos & Lewis, 2018; Le et al., 2019), our analysis reveals how the moralization process itself generates specific constraints that shape the available strategic responses. The constraining mechanisms we identify explain how organizations lose control over the moralization process through the accumulation of moral discourse, a process that goes beyond understanding of increasing scrutiny or maturity in the traditional issue lifecycle sense (Mahon & Waddock, 1992).

Moreover, our findings reveal that organizations can delay moralization temporarily through ambiguity, but may not be able to prevent it indefinitely. This insight extends recent work on strategic silence and deflection (Carlos & Lewis, 2018; Le et al., 2019) by showing that deflection strategies face not just stakeholder pushback but structural limits imposed by the moralization process itself. This also suggests that a strong public discourse and institutional spaces for moral deliberations are essential control mechanisms on organizational attempts to delay recognition.

The Evolution of Strategic Ambiguity Across Issue Stages

Our article extends strategic ambiguity theory (Cappellaro et al., 2021; Eisenberg, 1984) by showing how strategic ambiguity is used in the context of emerging societal problems. Our approach recognizes that the evolution of societal problems constrains and shapes the available ambiguity strategies over time. This perspective goes beyond viewing ambiguity strategies as merely reactive dialectical relationships between two actors (Cappellaro et al., 2021), to theorize this as a dynamic process that evolves alongside societal debates that constrain organizations and force them to change their strategies. We identify three types of strategic ambiguity – unclarity, inconsistency, and moral splitting – that function as stage-specific responses to evolving moralization constraints. Each transition between these types represents not organizational choice but organizational necessity driven by the progression of societal understanding. Organizations do not abandon unclarity because it has served its purpose; they abandon it because questioning has made purely technical ambiguity untenable. This insight extends strategic ambiguity theory (Cappellaro et al., 2021; Eisenberg, 1984) by showing how ambiguity is used in the context of emerging societal problems and how it is temporally and structurally constrained by the evolution of social meaning. While Cappellaro et al. (2021) demonstrate how actors maintain ambiguity through dialectical struggle with specific opponents, our analysis shows that when the struggle involves society-wide moralization of issues, the opponent becomes the accumulated discourse itself.