Here's a view from an old Scientific American article I'm telling from memory, explaining how we can judge conscious.
There's some wasp that kills a spider and brings it to a hole where it's going to plant its eggs on it. It drags the spider to the entrance of the hole, then leaves it outside the hole for a moment while it checks inside the hole one last time, then comes back out and drags the spider in.
A pesky scientist moves the spider away from the hole while the wasp is checking inside. When the wasp comes out, it resets its instinctive program to pull the spider to the entrance of the hole, releases it, and checks inside again. The scientist repeats this interference, making the wasp repeat its hard-programmed procedure, seemingly forever. The wasp never sees the pattern. The conclusion is the wasp is not conscious. It's an automaton.
How people are different is we recognize loops, then break out of them. Most higher animals can do this to varying degrees. One might think we do this best.
We have awesome pattern detectors. Sometimes it seems like 90% of what the brain is for.
So, we might say if a robot would nearly always recognize it was repeating a pattern then initiate a novel activity, that might show it was conscious. This would seem to me to require a massive amount of pattern detection that included detecting patterns of its own pattern detection. One level of recursion should be sufficient. Sounds computable to me.
The article ended with the observation that people often tend to behave in repetitive loops, like in serial failed relationships, and are therefore largely unconscious of such things. Indeed, debates about the nature of consciousness are infuriatingly repetitive. Each time we are about to pull it into the computable zone, someone drags it back out to the hard problem zone.
