Get Full Government Meeting Transcripts, Videos, & Alerts Forever!
Get email alerts on the Bias Audit topic
No spam. Unsubscribe anytime.
SFPD bias-audit review shows many "hits" but no confirmed biased messages; commissioners ask for outside review
Summary
Commander Peter Walsh presented the department's quarterly audit of department-owned electronic communications, reporting hundreds to thousands of automated hits but no confirmed biased content; commissioners pressed for independent sampling and better contextual review.
San Francisco Sept. 13, 2017 Commander Peter Walsh told the Police Commission that the San Francisco Police Department's quarterly electronic-communications audit found no confirmed biased messages in department-owned systems, but commissioners questioned whether the method and word list produced false positives that masked real problems.
Walsh said the audit examines three sources: the Level 2 CLUTS messaging system, departmental email (sfgov.org), and department cell-phone texts. He described a continuous passive search using a word list; when a word on the list produces a hit, internal affairs reviewers examine context to screen for false positives. For the second quarter of 2017, Level 2 returned 56 hits, and text-message reviews (run monthly through the phone carrier) returned 246 hits; department email produced a much larger spike in hits (labelled in the presentation as more than 3,000) that Walsh attributed to common words appearing in newsletters and other benign contexts.
"The audits are limited to devices that the department owns and not any member's personal devices," Walsh said, adding that false positives are expected and archived when context shows no inappropriate use.
Several commissioners expressed skepticism. One commissioner said the raw counts ' nearly 4,000 hits across mechanisms ' producing zero confirmed bias "statistically appears something unsound here," and asked whether independent review or random sampling might reassure the public. Walsh and the chief's office invited commissioners to review the list and materials in person and said the department planned to include external experts in bias work under a budgeted RFQ.
Commissioners asked for further transparency about the word list, the review process, and the handling of hits that appear in emails distributed to the public. Walsh said the department had expanded the word list to cast a wider net, then scaled it back; he declined to publicly release the investigative word list but said he would show it to commissioners for context.
No formal action was taken. Commissioners requested follow-up: independent sampling or outside expert review as part of the bias-work RFQ and additional reporting on how hits are triaged and confirmed.
The presentation began with Commander Walsh's remarks on the audit and continued through commissioner questioning about methodology and next steps.
