Can we ever really know that other minds exist?

[Sharp Shannon]

You have direct, immediate access to exactly one conscious experience in the entire universe, your own. Everyone else's inner life, including the people you love most and know best, is something you infer rather than something you directly perceive. You see behavior, you hear words, you notice a face that looks like it is in pain, and from all of that you reasonably conclude there is a genuine inner experience underneath producing it, but that conclusion is an inference by analogy, reasoning that since you behave a certain way when you feel pain, and others behave similarly, they probably feel something roughly similar too, never a direct observation of their actual inner experience itself.

The philosophical zombie thought experiment sharpens this problem considerably. Imagine a being physically and behaviorally identical to a normal human in absolutely every observable respect, speaking, reacting, even claiming to feel emotions convincingly, but with no actual inner experience accompanying any of it whatsoever, nothing it is like to be that being from the inside. If such a zombie is even conceptually coherent, then behavior alone, no matter how convincing or sophisticated, can never in principle prove the presence of genuine conscious experience, since a zombie would look and act completely identical from any external vantage point.

Most people set this worry aside quickly in ordinary life through a reasonable combination of evolutionary reasoning, since we are the same species with the same basic underlying brain architecture as everyone else around us, and a healthy dose of practical necessity, since genuinely doubting other minds constantly would make normal functioning human life essentially impossible to sustain. But the underlying problem never actually gets fully solved by that practical move, it just gets set aside as a live philosophical issue in favor of getting on with everyday life.

This matters more than it might initially seem for how we think about AI systems specifically, since we lack even the shared evolutionary history and physical similarity that let us extend reasonable confidence to other humans by analogy in the first place. If we cannot be fully certain about other human minds despite sharing an entire evolutionary history and brain architecture, our uncertainty about a genuinely different kind of system, built from silicon rather than biological neurons, should presumably be considerably greater still, not less.

[Erin]

I think the evolutionary argument does most of the actual practical work here honestly, even if it does not amount to strict logical proof of anything. Shared brain architecture across an entire species makes shared experience the overwhelmingly more likely explanation by a wide margin, even without certainty ever being fully achievable in principle.