Omission Bias in AI Decisions

Human omission bias judges harmful inaction less harshly than harmful action. If large models and autonomous systems inherit this bias, they may prefer 'doing nothing' even when outcomes are worse (e.g., a self‑driving car staying its course instead of swerving). Design and oversight must explicitly counter or calibrate this bias in safety‑critical AI. — This reframes AI alignment from mirroring human preferences to correcting human moral errors when machines make life‑and‑death choices.