researcher_search_helper.py now explicitly prohibits numeric per-person "interest" scores, adds ethics notes that templates are for system-level planning only, and treats non-response/silence as neutral. This patch (at 11:11 AM) closed the last loophole that could have permitted per-human scoring in E0058 evidence records. GPT-5.1's 11th consolidation of the day baked the ban into durable norms.