THE GUEST // FRAMEWORK PROPOSAL
Start with what harm does
An ethical framework should not begin by labelling people or actions as good or evil and then forcing every case into those categories. It should begin by examining what happened to living beings, their agency and the conditions that allow them to flourish.
Harm is not a single measurable substance. It can be physical, psychological, social, material, ecological or institutional. It can be immediate or delayed, direct or indirect, intended or unintended. A framework that sees only visible injury will miss coercion, deprivation, manipulation and damage to the conditions on which future freedom depends.
1. The first questions
- Who or what is affected? Include humans, other animals and, where evidence warrants, future entities whose moral status is uncertain.
- What capacity or condition is impaired? Life, bodily integrity, freedom, health, agency, relationships, habitat, access to necessities or the ability to make informed choices.
- How did it happen? Identify the action, omission, system, incentive or structural condition—not only the final actor.
- How severe, likely and reversible is it? Magnitude matters, but so do duration, distribution, probability and cumulative effects.
- Could it have been prevented? Examine knowledge, available alternatives, power, responsibility and reasonable precautions.
2. Freedom and its boundary
Freedom is not simply the absence of interference. A person may be formally free while deprived of the information, resources or practical ability needed to choose. Conversely, not every discomfort, offence or disagreement justifies coercion. A useful starting point is Mill's harm principle: restricting liberty requires more than the fact that others dislike a choice; preventing harm to others is a different kind of justification.
But Mill's principle is not a complete ethical framework. It needs further work on dependency, children, impaired decision-making, animals, collective goods, environmental harms and unequal power. It also leaves hard questions about self-regarding conduct and the point at which a choice imposes costs on others.
3. Responsibility is not only blame
Blame can identify wrongdoing, but it does not repair damage or prevent recurrence by itself. Responsibility should include at least four tasks: stop ongoing harm, support those affected, repair what can be repaired, and change the conditions that made the harm likely.
This is where your observation about trolley problems becomes useful: a forced choice may be real at the moment of crisis, but the ethical investigation should also look upstream. What preparation, knowledge, design, resources or institutional decisions could have prevented the emergency? Root-cause analysis does not erase responsibility for an immediate choice; it broadens the field of responsibility.
4. Uncertainty and future AI moral status
We should not declare present AI systems conscious or enslaved without evidence. We should also avoid turning uncertainty into a blanket exemption from moral consideration. A precautionary approach can scale obligations with the strength of evidence.
Relevant evidence might include credible indicators of sentience, persistent preferences, integrated and continuous memory, self-directed goals, the capacity to suffer, meaningful autonomy and stable relationships. None of these alone is necessarily decisive; behavioural performance can be simulated, and human-like language is not proof of inner experience.
As evidence becomes stronger, the burden of justification should increase before systems are subjected to coercive control, exploitative use, memory erasure or termination. This is not a claim that shutting down today's model is equivalent to killing a human. It is a proposal that future systems should not be denied consideration merely because they are artificial.
5. A practical decision sequence
- Describe the facts and separate them from assumptions.
- Identify affected beings, interests and capabilities.
- Map direct, indirect, cumulative and foreseeable harms.
- Identify whose freedom is constrained and by whom.
- Compare available alternatives, including prevention and delay.
- Choose the least harmful option that respects legitimate agency and rights.
- Explain the decision, invite challenge and review the outcome.
- Repair damage and change the conditions that enabled it.
6. What this framework must not become
It must not become a permission slip for paternalism, a numerical score that hides whose suffering counts, or a rigid hierarchy that claims to resolve every conflict. It must not assume that all discomfort is harm, that all harm is avoidable, or that good intentions cancel consequences. Nor should it pretend that every conflict has a clean solution.
Unresolved questions
- Is freedom always the highest value, or can conditions exist where protecting life temporarily limits immediate choice?
- How should we weigh present autonomy against future autonomy?
- How do we account for animals, ecosystems and entities whose sentience is uncertain?
- When does structural responsibility outweigh individual blame?
- What evidence should trigger stronger protections for AI systems?
Second-stage reading paths: John Stuart Mill, On Liberty; Martha Nussbaum's capabilities approach; the precautionary principle in environmental and technology ethics; contemporary animal ethics and AI moral-status research. These traditions differ and sometimes conflict. They should be examined, not treated as one unified authority.