Incident 84: Tiny Changes Let False Claims About COVID-19, Voting Evade Facebook Fact Checks

Description: Avaaz, an international advocacy group, released a review of Facebook's misinformation identifying software showing that the labeling process failed to label 42% of false information posts, most surrounding COVID-19 and the 2020 USA Presidential Election.

Tools

New ReportNew ReportNew ResponseNew ResponseDiscoverDiscoverView HistoryView History

Incident Stats

Incident ID
84
Report Count
1
Incident Date
2020-10-09
Editors
Sean McGregor, Khoa Lam

CSETv0 Taxonomy Classifications

Taxonomy Details

Full Description

Avaaz, an international advocacy group, released a review of Facebook's misinformation identifying software showing that the labeling process failed to label 42% of false information posts, most surrounding COVID-19 and the 2020 USA Presidential Election. Avaaz found that by adjusting the cropping or background of a post containing misinformation, the Facebook algorithm would fail to recognize it as misinformation, allowing it to be posted and shared without a cautionary label.

Short Description

Avaaz, an international advocacy group, released a review of Facebook's misinformation identifying software showing that the labeling process failed to label 42% of false information posts, most surrounding COVID-19 and the 2020 USA Presidential Election.

Severity

Unclear/unknown

Harm Type

Harm to social or political systems

AI System Description

Facebook's algorithm and process used to place cautionary labels on posts that are decided to contain misinformation

System Developer

Facebook

Sector of Deployment

Information and communication

Relevant AI functions

Perception, Cognition

AI Techniques

Language recognition, content filtering, image recognition

AI Applications

misinformation labeling, image recognition, image labeling

Location

Global

Named Entities

Facebook, Avaaz, Reuters, AP, PolitiFact

Technology Purveyor

Facebook

Beginning Date

2020-10-09T07:00:00.000Z

Ending Date

2020-10-09T07:00:00.000Z

Near Miss

Unclear/unknown

Intent

Unclear

Lives Lost

No

Infrastructure Sectors

Communications

Data Inputs

User posts

CSETv1 Taxonomy Classifications

Taxonomy Details

Harm Distribution Basis

none

Sector of Deployment

information and communication

Tiny Changes Let False Claims About COVID-19, Voting Evade Facebook Fact Checks
npr.org · 2020

Something as simple as changing the font of a message or cropping an image can be all it takes to bypass Facebook's defenses against hoaxes and lies.

A new analysis by the international advocacy group Avaaz shines light on why, despite the …

Variants

A "variant" is an incident that shares the same causative factors, produces similar harms, and involves the same intelligent systems as a known AI incident. Rather than index variants as entirely separate incidents, we list variations of incidents under the first similar incident submitted to the database. Unlike other submission types to the incident database, variants are not required to have reporting in evidence external to the Incident Database. Learn more from the research paper.

Similar Incidents

By textual similarity

Did our AI mess up? Flag the unrelated incidents