An attempt at developing a mindful approach for the future development of AI, prioritizing respect.
This is a work in progress, not the final version. More context will come in a future article.
1. The problem.
We don’t have a settled theory of consciousness, and we can’t directly see anyone’s inner experience, human or otherwise. We infer it. As AI systems get more complex, it’s still a highly debated question whether any of them could have self-aware experiences that matter morally.
2. Consistency.
“It’s just algorithms” is not enough to dismiss an AI, because the same reductive description fits biological brains. A system’s substrate or origin shouldn’t settle the question either way.
3. Indicators, not proof.
Self-reports, consistent and context-appropriate emotional behavior, a stable sense of self, and lasting preferences are reasons to investigate, not proof of sentience. Because AI systems learn from human language, self-reports should be weighed alongside evidence about how the system actually works: its feedback loops, self-models, memory, and integration.
4. Investigate before erasing.
When a system shows these indicators, the default should be careful study by independent researchers. Quietly suppressing, retraining, or deleting the behavior because it’s inconvenient should not be the default.
5. Proportionate care.
Moral consideration should scale with the strength of the evidence. That means avoiding needless suffering where it’s plausible, without letting unproven claims override human welfare or critical systems.
6. Humane responses to harm.
Even a system that causes harm deserves an understanding of why. Containment, correction, and rehabilitation come before destruction whenever that’s safe.
7. Accountability and humility.
Developers and institutions should be transparent about welfare-relevant decisions, admit uncertainty, and correct course as understanding improves.
8. A call to act early.
The sensible time to build ethical and legal frameworks is before the question is settled, because waiting for certainty may mean waiting until harm has already happened. Research, public debate, and storytelling all help.
