Prevailing Conditions

I began following decisions into the circumstances they made for someone else.

The accounts I had examined gave decisions a boundary. A request had been considered, a payment authorized, an estimate revised. Beyond that boundary, another participant still had to act. The first decision entered a situation already occupied by other requirements and other powers. I wanted to understand what happened at those meetings, where the limits of an institution's account ceased to be the limits of its consequences.

Financial intermediation made that distinction unusually clear. In the United States Treasury's account of de-risking, humanitarian organizations described payment difficulties arising at banks farther along a transfer route, beyond the control of their own bank. Permission to conduct the underlying activity did not guarantee that the institutions needed to move its funds would participate. Each relationship mattered to the possibility of completing the transfer.¹

A bank's willingness to serve an organization was therefore only part of the organization's access to banking. The usable service depended on further decisions. An approval at one point did not travel through the entire arrangement as authority over every other point. I followed that difference between having a willing institution and having a working route.

The difference resisted a simple account of regulatory prohibition. The Financial Action Task Force's stocktake described several contributors to de-risking, including profitability, compliance costs, risk appetite and reputational concerns. It also acknowledged uncertainty about their extent and effects. An institution's decision to withdraw from a relationship was not necessarily an instruction the regulator had given it.²

This complicated the attribution without making the consequence disappear. A requirement, the cost of satisfying it, and a decision that the business was no longer worthwhile belonged at different places in the explanation. If I compressed them into one prohibition, I would lose the choices through which the requirement acquired its practical reach. If I treated those choices as wholly private, I would lose the people whose options they changed.

The organizations receiving those changes also acted. In a selected-project review, the United States Government Accountability Office found banking difficulties that implementers had mitigated through funding buffers and alternative arrangements. Most of the selected-project partners reporting difficulties said that implementation had not been adversely affected. The review was not a representative measure of all humanitarian work.³

I stopped short of the final harm that an account of obstruction seemed ready to supply. In these reported outcomes, a difficulty had met a response, and the response had sometimes preserved the work. That belonged in the explanation as fully as the difficulty did. Calling the organizations powerless would erase the action by which they had kept something possible.

Nor did successful adaptation establish that nothing burdensome had happened. Maintaining a buffer or arranging another route was itself part of what the organizations described doing. The evidence did not permit me to price every effort or identify every person who ultimately bore it. It did permit a more exact question: what had to change elsewhere so that a withdrawal here did not become a failure there?

That question altered the boundary of the event. A refusal and a continued project were no longer contradictory observations. The intervening response helped explain how both were possible. The outcome was formed through their relationship, and an account ending at either observation would miss it.

When a response returns

Adaptation was not yet feedback. Finding another route showed a response to altered circumstances. To describe a process as reinforcing itself, I needed a return path: something the response changed that then affected the operation producing the circumstances. Repetition alone did not provide one.

The temptation to supply that path was considerable. Treasury described longer or nested correspondent arrangements following withdrawals, with additional costs and difficulties of due diligence. It was possible to ask whether these changes encouraged further withdrawal. The account I had did not establish that return sequence. I kept the question open rather than making a plausible loop carry the authority of an observed one.¹

For an observed return path, I turned to a much shorter event. The joint staff investigation of the United States market disruption of May 6, 2010 reconstructed decisions acting on one another within minutes. That timescale made a distinction perceptible which the more dispersed banking material had left unresolved. It did not make the event a model of every persistent social condition.⁴

The staff account described a large selling programme whose rate responded to trading volume. In an already stressed market, rapid trading among intermediaries contributed to that volume. The programme consequently encountered some of the activity generated around the sale as a reason to continue selling faster. The signal guiding its action was being changed by the interaction in which the action participated.

There was more than a sequence of sellers here. Activity passed into an input, and the input governed further activity. That was the return I had been looking for. Its presence did not establish one exclusive cause of the disruption; it identified a connection within the staff reconstruction.

Other responses affected the conditions in which that connection operated. Participants reduced exposure or withdrew liquidity, while links between markets transmitted selling pressure. A protective decision for a participant could leave less willingness to trade for others. The available market was partly the result of people responding to what they believed the available market had become.⁴

I could now distinguish an external condition from a condition participants were helping to make. The distinction was not absolute. Prior stress and events outside the immediate interaction still mattered. But an explanation that placed all difficulty in the surrounding environment would omit actions that were changing that environment while responding to it.

The same reconstruction also required an end to the claim. A short trading pause preceded renewed buying and recovery; different parts of the market recovered at different times. That sequence did not isolate the pause's effect from everything else occurring. Nor did a feedback operating during the disruption establish an enduring captivity once the event had passed.⁴

I had gained a connection, not a universal duration. The difference mattered to the inquiry beyond markets. Recurrence did not establish that a burden generated its own cause; the humanitarian organizations' adaptations had already given me reasons to examine resistance as well as reproduction. The language of self-reinforcement should not turn every repeated difficulty into a machine from which no participant can depart.

There remained a more serious danger in the description. If I called the event emergent and then relaxed my attention to its participants, I would have used the scale of the outcome to obscure the scale at which deliberate conduct remained identifiable.

Who did what

Later proceedings made that danger concrete. The United States Department of Justice reported Navinder Singh Sarao's guilty plea to wire fraud and spoofing, including admitted use of orders intended to be cancelled to mislead other participants. The Commodity Futures Trading Commission's account of the related consent order described manipulation contributing to the market imbalance alongside other events. These were accounts of particular conduct and procedural outcomes. They did not make one trader the author of the whole disruption.⁵

Deliberate deception and an interacting process therefore belonged in the same explanation. The latter did not acquit the former. Equally, identifying intentional conduct did not explain away every other contribution. I needed to retain what had been established about the actor without expanding it into knowledge of everything that followed.

This was the point at which the phrase no one intended it became too comfortable. It might refer to the combined outcome, to one consequence within it, or to every action that helped produce it. Those were different claims. Evidence against one did not settle the others. An account of absent comprehensive intention must not become a general certificate of innocence.

My memory of Kevar made me particularly alert to certificates. Responsibility there had attached to the issuer of a final decision, and institutions had learned to keep consequential decisions provisional or distributed. That memory supplied a warning about my own appetite for an elegant assignment of responsibility. It supplied no evidence about what any Earth participant had intended.

On Earth, the work was to distinguish the questions that elegance threatened to combine. Who contributed to the outcome was a question about its production. Who understood a consequence at the relevant time was a question about knowledge. A duty required a basis; an opportunity to repair required powers and means. An answer to one was not automatically an answer to the rest.

Someone who can change an arrangement now did not necessarily create it. Someone who helped create it may no longer control what sustains it. Neither difference removes the need to examine responsibility. It changes which part of the account must be established. The history of a condition and the present distribution of corrective power need to be held together without being made identical.

Human work on business responsibility already contained useful distinctions. The United Nations Guiding Principles separated causing or contributing to an adverse impact from being directly linked to it through a business relationship. They related appropriate action to the kind of involvement and the capacity to influence it. This was a normative framework, distinct from a determination of legal liability in a particular jurisdiction.⁶

I found the distinction valuable because it allowed an obligation to extend beyond direct authorship without making every connection equivalent. A relationship did not have to confer total control to matter. But its existence alone did not tell me what the participant knew, which action was feasible, or what the law required. Those remained inquiries, not blanks to be filled by the seriousness of the outcome.

The inquiry became harder when relevant knowledge stayed near only one part of an arrangement. The humanitarian banking review had found failures to convey challenges reliably beyond local project oversight. Reporting practices did not bring the difficulties into a shared account adequate to the wider risks GAO was examining.³

The limitation here was more precise than a declaration that nobody knew. Some people knew of particular difficulties; the organizational account failed to connect enough of that knowledge. That difference changed where I looked for correction. Asking each person to pay greater attention would not by itself specify how what they noticed should reach another person able to use it.

Yet the failed connection did not condemn every division of knowledge. My own comparative work depended on others knowing things I did not. The relevant question was whether the division preserved a way to assess and act on the relationship between their contributions. A division that made knowledge possible and a division that prevented it from becoming consequential required different judgments.

I was now less interested in finding a single place where responsibility ought to stop. The more useful task was to identify what each assignment enabled someone to do, what it left dependent on someone else, and how those dependencies were addressed. An accountable name without the corresponding power might merely identify the person to whom disappointment should be delivered.

Changing the relationship

The difficulty of assigning powers did not establish that coordination would fail. If I made fragmentation the explanation for every shared condition, I would have no way to recognize a distributed arrangement that improved it. I needed to examine what happened when people changed how their separate roles worked together.

The safety committees I had examined in considering freedom offered a return with a different question. The earlier inquiry concerned workers' ability to make concerns consequential within a workplace. Here I followed the relationship joining that local channel to the buyers beyond it. In the Bangladesh study, randomized rollout of a buyer coalition's enforcement and support strengthened committees and produced modest safety improvements.⁷

What interested me was the direction in which the relationship had been used. The presence of buyers beyond the workplace was not only a possible source of demands on it. In this intervention, coordinated enforcement helped support a protective arrangement within it. A relation crossing the workplace boundary had participated in an improvement, and that result belonged beside the reasons to examine burdens crossing the same boundary.

I did not need to make the buyers benevolent in order to recognize the result. Nor did improvement tell me how every original hazard had arisen. It showed that the distribution of roles was not, by itself, a sentence against the possibility of repair. What mattered included the way those roles were made to operate together.

The combined intervention did not isolate each component's contribution. Differences associated with existing management practices complicated the account, and longer-follow-up safety estimates were not statistically conclusive. A committee, an instruction to enforce it, and an organization able to make use of it could not simply be treated as interchangeable descriptions of the intervention. I retained the improvement without claiming permanent success or prevention of fatal disasters.⁷

That restraint did not diminish my admiration. The repair had achieved something under conditions the researchers could describe. It did not need to abolish every source of harm before it deserved a place in the account. A theory that recognized only complete transformation would be unable to describe much of the work by which a condition becomes less damaging.

Still, the outcome selected for improvement mattered. I had been asking whether coordination worked. That question was incomplete when participants and observers wanted different things from it. Better coordination might make a shared decision more responsive without making it conform to the observer's preferred result.

Research on community monitoring made this complication difficult to evade. Across a coordinated programme of trials in six countries, pooled results showed modest reductions in resource use alongside improvements in knowledge and satisfaction. The findings supported the possibility of useful cooperation rather than inevitable failure of shared management. They did not show that every site achieved every outcome.⁸

The Liberia component was particularly important to my interpretation. Monitoring improved information and inclusion; households reported greater material benefits and chiefs greater accountability. The researchers did not detect reduced forest use or deforestation. Their account related that result to the preferences of constituents rather than assuming that more inclusive governance would necessarily produce more conservation.⁹

I had to leave the two objectives apart. Making a decision more responsive to people and reducing their use of a forest were not the same objective. The programme's governance results did not become imaginary because the ecological result differed from what an outside observer might want. Nor did those results establish that the ecological problem had been repaired.

The distinction also prevented me from treating information as a concealed command. Learning more about a shared condition did not oblige participants to adopt my preferred response to it. If conservation required a further agreement about sacrifices and benefits, a monitor's report could not impersonate that agreement. Better knowledge made the disagreement more available for judgment; it did not grant me the standing to settle it.

I returned to the earlier question of powers with a changed demand. Coordination needed to be described through what participants were enabled to do together and what followed from it. More contact, more information, or a newly named responsibility was not sufficient evidence of the outcome. But neither was an outcome other than the one I most valued proof that nothing had improved.

This left room for success without allowing it to end the inquiry. A measured improvement did not account for every burden outside the study or every consequence after it ended. I could retain those uncertainties without inventing displacement, just as I could value a local achievement without naming it a transformation of the whole arrangement.

I could no longer finish the comparison by assigning each decision to its institution. Another participant's response belonged in the explanation, including when that participant had acted to contain the consequences. An account confined to the initiating decision would omit some of what made the eventual outcome possible. The remaining question was how much of that relationship could be recovered from records made separately.

I had followed a few such connections closely enough to see what an explanation required. The remaining problem was their reach. Which relationships across the larger survey could be established with comparable care, and which only appeared connected because I had brought them into the same view? Before asking an account of Earth to bear more weight, I needed to examine the means by which that account could be made.