[A philosophical inquiry into how a faceless committee can become morally culpable. Reading time: ~6 minutes.]
Direct answer: The agency approach, defended by Christian List and Philip Pettit in Group Agency: The Possibility, Design, and Status of Corporate Agents (2011), argues that a group can be a moral agent when it has the right internal architecture: reliable decision procedures, the capacity to form intentions, and responsiveness to reasons. When these conditions hold, collective blame is not a metaphor but a judgment of the group's procedures and its capacity to do better.
When a pharmaceutical company ships a drug it knows to be dangerous, who is the responsible party? The lab technician who ran the test? The manager who suppressed the results? The CEO who signed the approval? Or the corporation itself—that invisible lattice of contracts, procedures, and habits? The prevailing instinct is to blame the company, not just the people inside it. But what does it mean to blame an entity that has no body, no conscience, and can never stand in a courtroom dock?
The agency approach provides a rigorous answer. A group can be a moral agent if it satisfies three conditions: it forms intentions that are not merely the sum of its members' private wishes, it acts on those intentions in the world, and it is responsive to reasons—able to adjust its behavior when confronted with new evidence or criticism. A board of directors that deliberates, votes, and issues policies is such an agent. The group's decision procedure creates a stable, identifiable subject that persists even as individual directors come and go.
This is not to say the group becomes a mystical super-entity. It is to say that the corporation's internal architecture—its voting rules, its chains of command, its feedback loops—can generate a form of agency that is real enough to bear moral predicates. When we blame the oil company for a spill, we are blaming a decision-making system that could have chosen otherwise but did not. We are not merely using a convenient fiction.
The modern digital platform is the purest experiment in constructed agency we have ever built. When a recommendation engine curates a pattern of radicalization, no single engineer wrote the harmful loop—yet we blame the platform. The intuition that groups can be blamed has never been more computationally urgent.
Three Levels of Blame
The agency approach operates alongside—not instead of—individual accountability. The table below, drawn from the source article, clarifies the different levels at which blame attaches.
| Blame Level | Primary Subject | Grounding | Response Required |
|---|---|---|---|
| Individual guilt | A specific person | A manager falsifies safety data | Criminal or professional consequence |
| Collective responsibility | The group's policies and culture | The company rewards risk-taking that endangers workers | Institutional reform and repair |
| Political responsibility | All members of a community | A state votes for a climate treaty and fails to honor it | Civic accountability and democratic contestation |
If you found this valuable, consider supporting our work.
Join PhiloCrux community.
Unlock high-density masterclasses and investigations into ideas surviving outside the algorithmic consensus. Support independent thought and get full access to our digital library.
Join NowThe Limits of Group Blame
Yet the agency approach does not dissolve the individual. It operates alongside it. The corporation can be blamed for designing a culture that incentivized corner-cutting, while the individual engineer can be blamed for falsifying the safety data. The mistake is to think collective blame and individual blame are competitors. They are complementary scales of judgment, each capturing a different level of causation.
Take the case of an algorithm that systematically discriminates against a marginalized group. No single programmer wrote a line of racist code. The harm emerged from the platform's training data, its optimization targets, its testing procedures. The platform itself is the proper object of blame. But the product manager who ignored the bias reports has a separate, individual responsibility. To blame only the platform is to give the humans a free pass; to blame only the humans is to miss the structural nature of the harm.
The takeaway is neither a blanket “yes” nor a dismissive “no.” Groups can be blamed—but only when they have enough internal coherence to be considered agents, and only in ways that supplement rather than replace individual accountability. The hard work begins when we try to apportion that blame fairly