INNOCENT: True Stories of Human Error — and the Machine That Might Have Seen Truth

Jennifer Thompson was certain. She had memorized her attacker's face while he raped her, deliberately, so she could send him to prison one day. She did - and she was wrong. Ronald Cotton spent eleven years there before a DNA test proved it.He was not alone.INNOCENT walks through more than seventy true cases where absolute human certainty - a witness's memory, an expert's testimony, a jailhouse confession, a government algorithm's clean printout - turned out to be catastrophically wrong. A bite-mark expert whose testimony sent two innocent men to prison while the real killer struck again. A fire marshal whose "science" put a father on death row for a house fire that was never arson at all. A Dutch tax algorithm that branded twenty-six thousand families as fraudsters and helped bring down a government. A facial recognition match that got an eight-months-pregnant woman arrested for a robbery she could not physically have committed.And, running underneath every chapter, the same uncomfortable question: could a machine have caught what the humans missed - or would it just have gotten confidently, catastrophically wrong in a different way?This is not a book about artificial intelligence solving crime. It is a book about certainty itself - where it comes from, how it fails, and what it costs when nobody in the room is willing to stop and ask, one more time, "Am I sure?" Drawing on court records, government investigations, and decades of forensic-science research, INNOCENT traces the same devastating pattern across courtrooms, hospitals, welfare offices, and Silicon Valley product launches: the exact moment confidence stopped being checked.For readers of true crime, forensic science, and the unfolding story of AI in the justice system, INNOCENT is a reckoning with how often "beyond a reasonable doubt" has meant nothing of the kind - and a clear-eyed look at whether the newest tools built to fix that problem are actually ready for the job.