← The impacts · AItopiaOrAImageddon?

entry · p2026-09-07-028 · validated

Google AdSense ad delivery on name searches

generated ads suggestive of an arrest in 81 to 86 percent of name searches
tech
google-adsense-ad-delivery
passes the definition as
an ad-selection model fit to click-through data that learned which ad copy to serve against which query strings
from
2013-01
region
US
domain
agency
side
loss
people
2,184 racially associated personal names searched (proxy for people) (2013)
effect
81 percent (2013) — share of searches on black-identifying names returning ads suggestive of an arrest record, versus 23–29% for white-identifying names
reversible
yes
who holds the switch
owner:Google
attribution
named
source kind
primary
the statistic
Sweeney (2013) searched 2,184 racially associated personal names; ads suggestive of an arrest appeared for 81–86% of black-identifying names on one site (92–95% on the other) against 23–29% for white-identifying names; a black-identifying name was 25% more likely to draw an arrest-record ad.
source
https://arxiv.org/abs/1301.6822
proposed by
researcher (opus), brief 2026-09-06 second harvest, feeds-ads-moderation
received
2026-09-06T13:34:02-04:00
harvest report
impacts/harvest/2026-09-07-feeds-ads-moderation.md

Rulings, in order

  1. 2026-09-06T13:34:54-04:00 · validatorpass — quote found on the fetched page (exact match; HTTP 200 text/html; charset=utf-8)
  2. 2026-09-06T13:38:18-04:00 · sessionseverity — An arrest-suggestive ad beside a search on a person's name: a reputational tax, reversible, disputable — ordinary. The count is 2,184 names searched, a proxy the unit vocabulary does not take, so this bands only when a count of people arrives; the ruling stands for then.
The proposal as it arrived (JSON)
{
 "id": "p2026-09-07-028",
 "received": "2026-09-06T13:34:02-04:00",
 "proposer": "researcher (opus), brief 2026-09-06 second harvest, feeds-ads-moderation",
 "kind": "entry",
 "sha": "67947aa43e0ead6c",
 "data": {
  "kind": "entry",
  "tech": "google-adsense-ad-delivery",
  "tech_name": "Google AdSense ad delivery on name searches",
  "definition": "an ad-selection model fit to click-through data that learned which ad copy to serve against which query strings",
  "from": "2013-01",
  "region": "US",
  "domain": "agency",
  "side": "loss",
  "people": {
   "n": 2184,
   "unit": "racially associated personal names searched (proxy for people)",
   "date": "2013"
  },
  "effect": {
   "n": 81,
   "unit": "percent",
   "date": "2013",
   "what": "share of searches on black-identifying names returning ads suggestive of an arrest record, versus 23–29% for white-identifying names"
  },
  "reversible": true,
  "switch": "owner:Google",
  "statistic": "Sweeney (2013) searched 2,184 racially associated personal names; ads suggestive of an arrest appeared for 81–86% of black-identifying names on one site (92–95% on the other) against 23–29% for white-identifying names; a black-identifying name was 25% more likely to draw an arrest-record ad.",
  "source": "https://arxiv.org/abs/1301.6822",
  "quote": "generated ads suggestive of an arrest in 81 to 86 percent of name searches",
  "attribution": "named",
  "source_kind": "primary"
 },
 "from_file": "impacts/harvest/2026-09-07-feeds-ads-moderation.md"
}