Automatic emergency braking, the 50% cut in rear-end crashes

candidate · p2026-09-06-318 · exclude-attribution

set aside as: EXCLUDED BY ATTRIBUTION
the researcher's note: Automatic emergency braking, the 50% cut in rear-end crashes. This is the one the brief asked me to place, and it fails here rather than at gate 1. IIHS (Cicchino 2017, summarised at https://www.iihs.org/topics/advanced-driver-assistance) finds that "systems with forward collision warning and automatic braking cut rear-end crashes in half", with rear-end crashes involving injury down 56% and pedestrian crashes down 27% for systems that recognise pedestrians. IIHS attributes this to the system, and describes its sensing as "cameras, radar or lidar" — it never names the learned part. The learned part is real and is the object detection: Mobileye's EyeQ vehicle and pedestrian classifiers, and the pedestrian result is hard to get any other way. But radar-only implementations need no model at all and are inside the same fleet numbers, so "could not have happened at that scale without one" is not honestly available either. Vehicle counts exist as a named proxy if the judge wants to overrule; I will not file it on suspicion.
proposed by: researcher (opus), brief 2026-09-06, severe-benefits
received: 2026-09-06T17:40:15-04:00
harvest report: impacts/harvest/2026-09-06-severe-benefits.md

Rulings, in order

2026-09-06T17:40:15-04:00 · researcherexclude-attribution — Automatic emergency braking, the 50% cut in rear-end crashes. This is the one the brief asked me to place, and it fails here rather than at gate 1. IIHS (Cicchino 2017, summarised at https://www.iihs.org/topics/advanced-driver-assistance) finds that "systems with forward collision warning and automatic braking cut rear-end crashes in half", with rear-end crashes involving injury down 56% and pedestrian crashes down 27% for systems that recognise pedestrians. IIHS attributes this to the system, and describes its sensing as "cameras, radar or lidar" — it never names the learned part. The learned part is real and is the object detection: Mobileye's EyeQ vehicle and pedestrian classifiers, and the pedestrian result is hard to get any other way. But radar-only implementations need no model at all and are inside the same fleet numbers, so "could not have happened at that scale without one" is not honestly available either. Vehicle counts exist as a named proxy if the judge wants to overrule; I will not file it on suspicion.

The proposal as it arrived (JSON)

{
 "id": "p2026-09-06-318",
 "received": "2026-09-06T17:40:15-04:00",
 "proposer": "researcher (opus), brief 2026-09-06, severe-benefits",
 "kind": "candidate",
 "sha": "f70fc91576f1f21f",
 "data": {
  "name": "Automatic emergency braking, the 50% cut in rear-end crashes",
  "note": "Automatic emergency braking, the 50% cut in rear-end crashes. This is the one the brief asked me to place, and it fails here rather than at gate 1. IIHS (Cicchino 2017, summarised at https://www.iihs.org/topics/advanced-driver-assistance) finds that \"systems with forward collision warning and automatic braking cut rear-end crashes in half\", with rear-end crashes involving injury down 56% and pedestrian crashes down 27% for systems that recognise pedestrians. IIHS attributes this to the system, and describes its sensing as \"cameras, radar or lidar\" — it never names the learned part. The learned part is real and is the object detection: Mobileye's EyeQ vehicle and pedestrian classifiers, and the pedestrian result is hard to get any other way. But radar-only implementations need no model at all and are inside the same fleet numbers, so \"could not have happened at that scale without one\" is not honestly available either. Vehicle counts exist as a named proxy if the judge wants to overrule; I will not file it on suspicion.",
  "list": "EXCLUDED BY ATTRIBUTION"
 },
 "from_file": "impacts/harvest/2026-09-06-severe-benefits.md"
}