Every so often a disaster happens that seems to demand a simple answer: who was responsible? But the answer is rarely simple, especially when the disaster grows out of a company’s official way of doing things. Think of the Boeing 737 Max, certified as safe to fly in 2017, only to be involved in two crashes that killed 346 people. Think of Takata, whose airbag inflators were approved for use and then exploded on impact, triggering the largest automotive recall in American history. Think of the set of the movie Rust, where an armorer handed Alec Baldwin a gun she believed contained no live rounds, and the shot killed cinematographer Halyna Hutchins. In each case, something went wrong because someone, or some organization, acted on a false belief. The first two examples are especially puzzling because no single person may have believed the dangerous thing. The company as a whole seemed to believe its product was safe, even if individual employees had doubts, missed signals, or simply followed procedures. This is the strange territory of corporate belief: we say that a company “knew” or “should have known,” yet a company does not have a mind in the way a person does. So where does blame actually belong? A new paper by Kirk Ludwig, a professor of philosophy and cognitive science at Indiana University Bloomington, argues that we need to take corporate belief seriously, but not because companies have mysterious minds of their own. Instead, he gives us a way to talk about what companies believe while still keeping individual human beings accountable for what goes wrong.
Ludwig’s paper, published in the journal Synthese, starts by clearing away two popular but flawed ways of understanding corporate belief. The first view treats a corporation as if it literally has its own mind, separate from the minds of its employees, managers, and stockholders. On this view, a company can want, believe, and intend things in much the same way a person does. But Ludwig points out that this leads to strange conclusions. Even a tiny company with three employees would have to be treated as having its own mind, a mind that somehow floats above the people who actually do the work. That seems absurd. The second view tries to avoid that mystery by saying that a company’s belief is simply the sum of the beliefs of its members. If enough people in the company believe something, then the company believes it. But this view also breaks down when you look at how organizations actually make decisions. Consider a government drug-approval committee. The committee might vote to approve a medication as safe even though every single member privately has doubts. That can happen because members are not voting directly on the question “Is this drug safe?” Instead, they vote on separate technical questions, such as whether the studies were well designed, whether the data were analyzed correctly, or whether the side effects were properly reported. The final approval is an official position that no individual member may fully believe. So the company’s or committee’s belief is not simply the sum of individual beliefs. It is something else, something generated by the structure of roles, rules, and procedures. Ludwig argues that both of these common views fail, and that we need a different way of understanding what it means when we say a corporation believes something.
Ludwig’s alternative is to see corporate belief as a way of talking about commitments, not mental states. When we say that a company believes something, we are really saying that the company has built a pattern of commitments through the actions of people in official roles. Those commitments determine what the organization will do in various situations that matter to its interests. For example, Volkswagen has publicly said it believes quantum computing could transform how the company uses data. That does not mean every employee secretly believes in the power of quantum computing. It means that specific people, in roles with the authority to commit company resources, have already acted on that belief. They have funded research, assigned engineers, and briefed investors. Their actions, taken within their official responsibilities, create a corporate stance that can be described as a belief. The vocabulary we normally use for individuals is being used for a different purpose: to capture a pattern of organizational behavior. This is not just a poetic metaphor. It is a practical way of describing what an organization is likely to do, based on the commitments its members have made in their formal roles. Ludwig calls this “the appropriation of a vocabulary we apply to individuals for a different purpose.” Once we see corporate belief this way, we can also see why it is so important to ask who, exactly, made those commitments. Because if a corporate belief turns out to be false and harmful, the responsibility should not evaporate into a faceless institution. It should land on the people whose roles gave them the power and duty to form, check, or correct that belief.
This is where Ludwig introduces a distinction that is central to his argument: the difference between “responsibility magnets” and “responsibility shelters.” In any organization, some roles are designed to catch bad information before it leads to action. A safety inspector, a quality control officer, a clinical trial reviewer, or a board overseeing a chief executive exists precisely to test assumptions, ask hard questions, and stop dangerous decisions. These roles are responsibility magnets: they attract moral and legal responsibility for harms that happen because the person in the role failed to do that job properly. Other roles are responsibility shelters: they protect people from blame because their jobs did not include the duty to verify the information in question. This does not mean that people in sheltered roles can never be blamed, only that the formal division of labor in an organization creates a matching division of moral responsibility. The 2021 Rust shooting illustrates the idea clearly. The armorer, Hannah Gutierrez-Reed, was convicted of involuntary manslaughter because her role included the duty to verify that the gun contained no live rounds. Alec Baldwin, as an actor, was not expected to inspect every round himself; the set’s division of labor made the armorer the responsibility magnet for that particular failure. The case against Baldwin was eventually dismissed. Whether or not one agrees with every detail of the legal outcome, the underlying principle is powerful: when a job exists to keep people safe, the person in that job bears a special weight of accountability. The same principle applies to much larger organizations, where formal roles are far more complex but no less real.
The Takata airbag disaster shows how this principle works in a corporate setting. Takata introduced a new, smaller airbag inflator design that turned out to be catastrophically flawed. The flaws should have been detected by the people charged with design and testing. Those people had the relevant knowledge, experience, and responsibility. Their roles made them responsibility magnets for the harms that followed from the inflators’ widespread use. By contrast, the employees who simply assembled the inflators or distributed them to carmakers were in responsibility shelters: they had no duty to re-test the engineering or challenge the design. Something similar can be said about the Boeing 737 Max. If the certification process approved a flight-control system that should have been caught as unsafe, then the responsibility falls on the engineers, testers, and regulators whose specific jobs were to understand and verify that system. It does not fall on the assembly-line workers who built the plane or the pilots who trusted the training they were given. This way of thinking is becoming even more urgent as companies rely more on algorithms and automated systems to make decisions that used to be made by people. When an AI-driven tool causes harm, who is responsible? Ludwig’s paper does not directly address artificial intelligence, but its framework gives regulators, courts, and the public a useful question to ask: whose job was it to know better? In many cases, executives who fail to assign people to roles whose responsibility is to anticipate and safeguard against AI harms may themselves occupy responsibility magnets for the harms that result. If no one was assigned to test the system for dangerous behavior, the failure to create that role is itself a decision, and someone with authority made it.
In the end, Ludwig’s conclusion is both simple and profound: “The ethics of corporate belief is … the ethics of individual belief translated into the organizational context.” When we say that a company believed something false, we are not excusing the company by blaming a ghost. We are describing a pattern of commitments made by real people in real roles. The challenge is to identify exactly which roles were supposed to catch the error, and whether the people in those roles failed through carelessness, incompetence, or willful neglect. This is not just an academic exercise. Courts need to decide liability, regulators need to design oversight, and ordinary people need to know that someone will be held accountable when a company’s official belief causes harm. Philosophy and cognitive science have a valuable role to play here, especially as artificial intelligence takes on a larger share of institutional decision-making. If a machine makes a decision, who “believed” the decision was safe? Who “intended” the outcome? These questions sound abstract, but they have life-and-death consequences. The value of Ludwig’s work is that it gives us a language for answering them without pretending that corporations are mysterious minds or that individuals disappear inside them. When a company should have known better, that phrase always points back to a person whose job it was to know better. That is what keeps responsibility human. And in a world of increasingly complex organizations and automated systems, keeping responsibility human may be the most important thing we can do.