Abstract
Helping systems are designed to reduce harm. They are built on the premise that expertise, intervention, and organized response can improve human well-being. This report examines a different question. What happens when those same systems generate harm, and when the process of recognizing that harm introduces risk to the people within them?
Drawing from a small observational sample across health and social service contexts, this report identifies a consistent pattern. As the perceived risk of learning increases, the system's capacity to recognize and respond to harm decreases. These data suggest that harm is not only produced within systems of care, but that systems also shape the conditions under which harm can be seen, named, and addressed. Knowing these conditions is essential for any effort aimed at meaningful repair.
Introduction
Helping systems operate with a shared promise. They exist to care, to intervene, and to improve outcomes for people and communities. Their legitimacy is tied to this function, and their authority is reinforced through training, credentialing, and institutional structure. Over time, these systems have developed increasingly sophisticated methods for identifying problems, delivering services, and measuring results.
At the same time, there is rising recognition that these systems can produce harm. This harm manifests in multiple forms, including clinical errors, bureaucratic burdens, policy-driven inequities, and relational breakdowns between providers and the people they serve. Entire bodies of research have emerged to document these effects, and in many cases, to propose solutions.
Less understood is what happens after harm is named. When individuals within a system begin to question whether the system itself may be adding to the problem, a different dynamic emerges. The issue is no longer only technical or procedural. It becomes relational, institutional, and, at times, existential. This report begins with a simple question. What conditions make it possible, or impossible, for a system to learn from the harm it produces?
Why This Question Now
In the wake of cascading system failures across healthcare, child welfare, housing, and public health — many made visible by the COVID-19 pandemic — the gap between institutional intention and institutional impact has become harder to ignore. Systems are being asked to account for themselves in new ways. This report examines why that accounting remains so difficult, and what must change for it to become possible.
The Conditions of Seeing Framework
Systems do not fail because they lack information. They fail because they lack the conditions to see what they already know.
What Systems Already Know
Across disciplines, there is broad agreement that systems designed to provide care can also produce harm. In medicine, this is called iatrogenesis — a term which captures the unintended negative consequences of treatment and interventions (Illich, 1976; Institute of Medicine, 2000). In public policy, similar dynamics appear as unintended consequences. In economic and historical research, harm is traced via patterns of inequality that continue across generations, frequently rooted in earlier institutional decisions that were normalized at the time.
Reparations literature provides another lens, framing harm as produced, maintained, and transmitted through systems over time. These bodies of work differ in language and emphasis, but they converge on a shared recognition: harm exists within systems that were designed to help. The challenge is not that harm is unknown. The difficulty lies elsewhere.
How Systems Make Harm Legible
While harm is widely acknowledged, its recognition is mediated by institutional structures that determine what can be perceived and comprehended. Systems count on classification, measurement, and reporting mechanisms to define reality. These mechanisms establish categories, assign value, and create the conditions under which something becomes visible within the system.
Historical research offers a clear illustration. Accounting practices during slavery translated human lives into financial categories, enabling systems of exploitation to function as routine economic activity. This did not conceal harm. It reorganized it into a form that the system could process without disruption. Modern systems continue to rely on structured forms of representation. What is counted becomes actionable. What is not counted remains peripheral.
Where Systems Fail to See
Reparations literature stresses the importance of clearly naming harm and establishing mechanisms for responsibility. Policy examination shows the difficulty of enacting these mechanisms in practice, even when there is agreement on the underlying problem. Psychological research introduces another dimension, showing that identity, emotion, and group dynamics affect how individuals and institutions respond when harm is acknowledged.
What remains underdeveloped is an understanding of how systems process harm internally once it is introduced. There is limited exploration of the conditions that allow a system to remain in a learning posture when its own practices are called into question.
What Makes Seeing Possible
Emerging insights suggest that recognition is not purely a function of information. It is shaped by the conditions under which learning occurs. These conditions include the ability to name harm without triggering immediate defensive response, the capacity to hold responsibility without collapsing into blame, and the willingness to tolerate ambiguity long enough for deeper understanding to emerge.
When these conditions are present, systems are more likely to engage in reflective processes that allow for meaningful learning. When they are absent, systems may default to protective responses that preserve coherence at the expense of accuracy. In such cases, learning is constrained not by a lack of data but by the perceived risk of engaging that data.
Joy Repair Contribution
This study builds on existing literature by focusing on the conditions that enable or limit recognition within systems. Rather than asking whether harm exists, it examines whether systems can see what they are doing.
The findings suggest that as learning becomes riskier, systems become less able to recognize harm. This shift moves the focus from individual behavior to systemic conditions. By centering the conditions of learning, this report provides a framework for understanding why systems capable of producing knowledge may still struggle to act on it.
Structural Competency Note
Metzl and Hansen (2014) introduced the concept of structural competency to describe the capacity of practitioners to recognize how systems — including healthcare itself — shape health outcomes. Their framework offers a clinical complement to this report's organizational findings: individual competence is insufficient when the structures governing practice remain unexamined.
Methodology
This report draws on a small, purposive sample of observations across health and social service environments. The goal was not to produce a statistically representative dataset, but to identify patterns that appear consistently across different contexts.
Sources include public-facing professional writing, documented responses to critiques of care systems, facilitated conversations within practice and leadership settings, and applied work conducted through Joy Repair initiatives. The analysis focused on patterns in how harm was introduced, interpreted, accepted, resisted, and translated into action or inaction.
Positionality Statement
This research was developed within Joy Repair, an applied practice organization. The lead researcher is a licensed social worker with direct experience facilitating harm-recognition processes inside health and social service systems. This proximity to practice is an epistemological resource, not a confound.
On Sample Size & Validity
Qualitative and observational research traditions do not require large samples to generate valid findings. Purposive sampling, theoretical saturation, and pattern consistency across diverse contexts are the relevant validity criteria here (Lincoln & Guba, 1985).
Findings
A consistent pattern emerged across contexts. As the perceived risk of learning increased, the system's capacity to recognize harm decreased. This relationship was expressed through five recurring dynamics.
Individualization of Harm
Harm was frequently individualized. When concerns were raised, responses often moved quickly to identify a specific person or event, narrowing the scope of the issue and reducing visibility into broader patterns. This allowed the system to contain the problem without engaging its underlying structure. This mirrors what sociologists call "attribution error at the institutional level" — the tendency to locate systemic problems in individual actors rather than structural conditions.
Ryan, 1976
Reassertion of Authority
Authority was reasserted in moments of uncertainty. Systems tended to stabilize themselves by returning to established hierarchies, reinforcing expertise, and relying on familiar language. While this provided coherence, it often limited the space available for deeper inquiry. This dynamic is consistent with what Argyris (1990) termed "defensive routines" — organizational behaviors that protect existing structures from scrutiny, even when scrutiny is precisely what the situation requires.
Argyris, 1990
Learning Recast as Threat
Learning was reinterpreted as risk when it challenged identity or legitimacy. In these moments, new information was not engaged as an opportunity for understanding, but as a potential threat. This shift reduced curiosity and increased defensiveness, constraining the system's ability to process what was being presented. The literature on cognitive dissonance and motivated reasoning provides individual-level mechanisms for this organizational pattern.
Festinger, 1957; Kunda, 1990
Diffusion of Responsibility
Responsibility became diffuse within complex environments. Without clear mechanisms for collective accountability, responsibility was often displaced or fragmented, making coordinated response difficult. The system retained awareness of harm in parts, but struggled to act on it as a whole. This resonates with research on the bystander effect as an organizational phenomenon — when accountability is shared by many, it is often exercised by none.
Darley & Latané, 1968
Language as a Limiting Structure
Language played a central role in shaping perception. When harm did not align with existing categories or definitions, it became difficult to name, measure, or address. This created conditions in which harm could be experienced but not fully recognized within the system's own terms. Language is not neutral. The categories a system possesses determine the harms it can see.
Bowker & Star, 1999
A system cannot repair what it cannot name. And it cannot name what its language was not built to hold.
Conceptual Framework
Table 1
The Recognition Gradient
Perceived risk of learning vs. system capacity to recognize harm
| Perceived Risk | System Posture | Harm Recognition | Likely Outcome |
|---|---|---|---|
| Low | Curious, Open | High | Learning & Adaptation |
| Moderate | Cautious, Selective | Partial | Incremental Change |
| High | Defensive, Reactive | Minimal | Harm Containment Only |
| Existential | Protective, Closed | Blocked | Harm Reproduction |
Five Recurring Dynamics
How Systems Limit Their Own Sight
Individualization of Harm
System narrows scope to single actor, avoiding structural pattern recognition.
Reassertion of Authority
Hierarchy reinforced. Familiar language deployed. Space for inquiry closes.
Learning Recast as Threat
New information becomes danger. Curiosity replaced by defensiveness.
Diffusion of Responsibility
Accountability fragments across roles. System knows harm in parts but cannot act as whole.
Language as Limiting Structure
If the system has no word for it, it cannot see it. Categories determine what harm is legible.
A system cannot repair
what it cannot name.
Interpretation
These findings suggest that helping systems function not only as sites of care but as environments that regulate perception. The ability of a system to learn from harm is shaped by whether learning can occur without destabilizing core elements such as identity, authority, and coherence.
When these elements are threatened, systems may prioritize internal stability over external accuracy. This does not require conscious avoidance. It emerges through routine processes of sense-making, where maintaining continuity becomes more immediate than engaging complexity.
In this context, harm is not simply overlooked. It is filtered through conditions that determine what can be acknowledged without disrupting the system's functioning. Recognition becomes conditional, dependent on whether the system can absorb what it encounters.
The system is not broken. It is working exactly as designed — to survive. The question is whether survival and accountability can be made compatible.
This interpretation carries significant moral weight. If harm is not an accident of inattention, but a predictable output of systems optimized for self-preservation, then solutions that address only information — more training, better data, clearer guidelines — will consistently fall short. The conditions under which information is received must change alongside the information itself.
Implications for Practice and Policy
If systems are to improve their ability to respond to harm, attention must be given to the conditions under which learning occurs. This includes creating environments where harm can be named without immediate attribution of blame, developing shared approaches to responsibility that reflect the distributed nature of complex systems, and expanding the language available to describe lived experience.
For Practitioners
Supervisors who create genuine safety for reporting harm without retaliation, clinicians who name structural contributors to client suffering, and administrators who protect dissent rather than suppress it are all doing the structural work this framework calls for. The challenge is that this work is often invisible, unrewarded, and at times career-threatening.
For Policymakers
Policy environments frequently incentivize performance over honesty. An accountability system that makes harm-reporting dangerous is not an accountability system. It is a suppression system in accountability's clothing. Policymakers committed to genuine improvement should consider what accountability mechanisms create, not only what they measure.
For Leaders
Audit your organization's accountability structures. Ask not only what you measure, but what your measurement systems make safe or dangerous to report. Accountability that punishes disclosure is not accountability.
For Communities
Communities most harmed by helping systems — Black, Indigenous, and low-income communities, among others — have long named what this report is now framing in organizational terms. This report does not seek to translate lived community knowledge into institutional legitimacy. It seeks to build the case that institutions must become worthy of that knowledge.
Conclusion
Helping systems are designed to respond to harm. This report suggests they also shape the conditions under which harm can be seen.
When learning becomes risky, systems may lose access to the very information needed to improve. The challenge is not only to generate knowledge, but to create the conditions under which that knowledge can be recognized and acted upon.
We do not lack evidence. We lack the conditions to see it clearly, speak it safely, and act on it together.
This report is a beginning. It is an attempt to name a dynamic that many practitioners, community members, and researchers have experienced but struggled to articulate within the language systems provide. The conditions of seeing are not fixed. They can be designed. They can be protected. They can be changed.
The work of repair begins not with more information, but with the courage to build systems that can finally bear witness to what they know.
Join the Work
Practitioners
Bring this framework into your supervision. Ask: what is our system's current risk of learning? What would it take to lower it?
Researchers
Partner with Joy Repair to test, refine, and expand these findings. The Recognition Gradient is a hypothesis. Help us prove it, complicate it, or replace it with something better.
Leaders
Audit your organization's accountability structures. Ask not only what you measure, but what your measurement systems make safe or dangerous to report.
Communities
You have always known this. Joy Repair exists to build toward institutions that are finally ready to listen. We want your experience, your critique, and your leadership.
Table 2
Knowledge Map
Interdisciplinary foundations of the Conditions of Seeing framework
Harm Within Systems of Care
Establishes that helping systems produce iatrogenic harm as a structural feature, not an aberration.
Structural Harm & Systemic Inequality
Documents how institutional patterns reproduce harm across generations and how blame is mislocated.
Reparations & the Necessity of Recognition
Frames harm as requiring active, named acknowledgment before repair becomes possible.
Systems of Classification & Perception
Shows how institutional categories determine what harm is legible — and what remains invisible.
Identity, Cognition & Response to Harm
Explains individual and collective psychological mechanisms that resist harm acknowledgment.
Organizational Learning & Defensive Routines
Identifies how organizations suppress learning to protect identity and coherence.
Expanding System Capacity for Recognition
Points toward structural approaches that make recognition possible without destabilizing systems.
References
- Argyris, C. (1990). Overcoming organizational defenses. Allyn & Bacon.
- Argyris, C., & Schön, D. A. (1978). Organizational learning: A theory of action perspective. Addison-Wesley.
- Bowker, G. C., & Star, S. L. (1999). Sorting things out: Classification and its consequences. MIT Press.
- Creswell, J. W., & Poth, C. N. (2018). Qualitative inquiry and research design (4th ed.). SAGE.
- Darley, J. M., & Latané, B. (1968). Bystander intervention in emergencies. JPSP, 8(4), 377–383.
- Darity, W., & Mullen, A. K. (2020). From here to equality. UNC Press.
- Darity, W., Hamilton, D., Mullen, A. K., & Tippett, R. (2024). The reparations calculus. JEP, 38(1), 3–28.
- Feagin, J. R. (2013). The white racial frame (2nd ed.). Routledge.
- Festinger, L. (1957). A theory of cognitive dissonance. Stanford University Press.
- Illich, I. (1976). Medical nemesis. Pantheon Books.
- Institute of Medicine. (2000). To err is human. National Academies Press.
- Jeacle, I. (2022). Accounting and the management of bodies. AOS, 97, 101295.
- Kahneman, D. (2011). Thinking, fast and slow. FSG.
- Kunda, Z. (1990). The case for motivated reasoning. Psych. Bulletin, 108(3), 480–498.
- Lincoln, Y. S., & Guba, E. G. (1985). Naturalistic inquiry. SAGE.
- Merton, R. K. (1936). Unanticipated consequences of purposive social action. ASR, 1(6), 894–904.
- Metzl, J. M., & Hansen, H. (2014). Structural competency. SSM, 103, 126–133.
- Ryan, W. (1976). Blaming the victim (rev. ed.). Vintage Books.
- Senge, P. M. (1990). The fifth discipline. Doubleday.
- Tavris, C., & Aronson, E. (2007). Mistakes were made (but not by me). Harcourt.
- Weaver, V. M. (2008). Frontlash: Race and punitive crime policy. SAPD, 21(2), 230–265.
McAleavey, M. (2026). The conditions of seeing: Why systems struggle to recognize harm. Journal of Economic Iatrogenesis & Labor Studies, 1(1). joy4repair.com/framework