Incident 102: Personal voice assistants struggle with black voices, new study shows

Description: A study found that voice recognition tools from Apple, Amazon, Google, IBM, and Microsoft disproportionately made errors when transcribing black speakers.

Tools

New ReportNew ReportNew ResponseNew ResponseDiscoverDiscoverView HistoryView History

Entities

View all entities
Alleged: Microsoft , IBM , Google , Apple and Amazon developed and deployed an AI system, which harmed Black people.

Incident Stats

Incident ID
102
Report Count
2
Incident Date
2020-03-23
Editors
Sean McGregor, Khoa Lam
Applied Taxonomies
CSETv0, CSETv1, GMF, MIT

CSETv1 Taxonomy Classifications

Taxonomy Details

Incident Number

102

Notes (special interest intangible harm)

Voice recognition software's performance varied depending on the speaker's regional accents and race. Speech recognition performed worse for African American vernacular.

Special Interest Intangible Harm

yes

Date of Incident Year

2020

MIT Taxonomy Classifications

Machine-Classified
Taxonomy Details

Risk Subdomain

1.3. Unequal performance across groups

Risk Domain

  1. Discrimination and Toxicity

Entity

AI

Timing

Post-deployment

Intent

Unintentional

Personal voice assistants struggle with black voices, new study shows
theverge.com · 2020

Speech recognition systems have more trouble understanding black users’ voices than those of white users, according to a new Stanford study.

The researchers used voice recognition tools from Apple, Amazon, Google, IBM, and Microsoft to tran…

Racial disparities in automated speech recognition
pnas.org · 2020

Automated speech recognition (ASR) systems are now used in a variety of applications to convert spoken language to text, from virtual assistants, to closed captioning, to hands-free computing. By analyzing a large corpus of sociolinguistic …

Variants

A "variant" is an AI incident similar to a known case—it has the same causes, harms, and AI system. Instead of listing it separately, we group it under the first reported incident. Unlike other incidents, variants do not need to have been reported outside the AIID. Learn more from the research paper.
Seen something similar?

Similar Incidents

By textual similarity

Did our AI mess up? Flag the unrelated incidents

Gender Biases in Google Translate

· 10 reports
COMPAS Algorithm Performs Poorly in Crime Recidivism Prediction

COMPAS Algorithm Performs Poorly in Crime Recidivism Prediction

· 22 reports
High-Toxicity Assessed on Text Involving Women and Minority Groups

High-Toxicity Assessed on Text Involving Women and Minority Groups

· 9 reports
Previous IncidentNext Incident