The mathematics of intelligence in multi-agent decision problems implication gives us informed guesses about what we should expect to see if an agent has free will versus if they don't. You are arguing with words like illusion, so I don't think you realize this, but what you actually need to provide to have your actual position be the consequence of experimental evidence is evidence of extreme compression being possible. Illusion? It is the evidence for agents, not evidence against agents, according to the theory informed guesses.
There is this story about a village of blind men and elephants. You are like a blind man from that village, saying that the elephant must be abstract because you felt a footprint of the elephant. Try playing with an elephant carving. It predicts footprints when you press it it into the ground. You can totally guess before you touch the elephant on the basis of the carving. The decision to focus on the untestable is a commitment to a hindsight fallacy which experimental investigation would likely refute.
There is this story about a village of blind men and elephants. You are like a blind man from that village, saying that the elephant must be abstract because you felt a footprint of the elephant. Try playing with an elephant carving. It predicts footprints when you press it it into the ground. You can totally guess before you touch the elephant on the basis of the carving. The decision to focus on the untestable is a commitment to a hindsight fallacy which experimental investigation would likely refute.