Diagnosing Gender Bias in Image Recognition Systems
Bibliographic Data
| ID | 2423958 |
|---|---|
| Authors | Carsten Schwemmer (0000-0001-9084-946X, GESIS - Leibniz Institute for the Social Sciences, corresponding author), Carly Knight (0000-0002-3985-6186, New York University), Emily D Bello-Pardo, Emily Bello-Pardo (American University), Stan Oklobdzija (0000-0002-8082-5824, California YIMBY, Sacramento, CA, USA), Martijn Schoonvelde (0000-0003-4370-2654, University College Dublin), Jeffrey W Lockhart (0000-0003-1519-9588, University of Michigan) |
| Year | 2020 |
| Volume | 6 |
| Publication date | 2020-01-01 |
| Peer Reviewed | Yes |
| Open Access | Yes |
| Type | ARTICLE |
| Venue | Socius Sociological Research for a Dynamic World (JOURNAL) |
| Journal identifiers | ISSN: 2378-0231 • E-ISSN: 2378-0231 |
| Publisher | SAGE Publications Inc (PUBLISHER) |
| DOI | 10.1177/2378023120967171 |
| PMID | 35936509 |
| OpenAlex | W3103585759 |
| Language | DE |
| Citations received | 19 |
| References cited | 52 |
Image recognition systems offer the promise to learn from images at scale without requiring expert knowledge. However, past research suggests that machine learning systems often produce biased output. In this article, we evaluate potential gender biases of commercial image recognition platforms using photographs of U.S. members of Congress and a large number of Twitter images posted by these politicians. Our crowdsourced validation shows that commercial image recognition systems can produce labels that are correct and biased at the same time as they selectively report a subset of many possible true labels. We find that images of women received three times more annotations related to physical appearance. Moreover, women in images are recognized at substantially lower rates in comparison with men. We discuss how encoded biases such as these affect the visibility of women, reinforce harmful gender stereotypes, and limit the validity of the insights that can be gathered from such data
Affect (linguistics · Data science · Geography · Image (mathematics · Machine learning · Scale (ratio · Visibility · Communication · Computer Science · Evolutionary Psychology and Human Behavior · Gender Politics and Representation · Psychology · Social and Intergroup Psychology · Artificial Intelligence
Computer-assisted qualitative visual analysis
Creating large, high-quality geospatial datasets from historical maps using novice volunteers
Auditing the representation of migrants in image web search results
Diversity and language technology
Facing the Competition
Gender bias perpetuation and mitigation in AI technologies
Harm to Nonhuman Animals from AI
The Prevalence and Consequences of Women’s Algorithmic Underrepresentation in Cross-National Political Google Searches
Gender bias in text-to-image generative artificial intelligence
Analyzing radical visuals at scale
Automated Visual Analysis for the Study of Social Media Effects
The user is dead, long live the platform? Problematising the user-centric focus of (digital) memory studies
Bridging the computational and visual turn
Unintended effects of algorithmic transparency
Analyzing Text and Images in Digital Communication
When Neutrality Conceals Bias
Automated Coding of Political Campaign Advertisement Videos
Gender, Candidate Emotional Expression, and Voter Reactions During Televised Debates
Generative Multimodal Models for Social Science
Status generalization
The Managed Heart
Algorithms of Oppression
Human Decisions and Machine Predictions
The Misgendering Machines
Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead
Bias in computer systems
Algorithmic Decision Making and the Cost of Fairness
Excavating AI
Images as Data for Social Science Research
Gender Advertisements
Framed by Gender
Gender Inequality in Product Markets
Who are these people?” Evaluating the demographic characteristics and political preferences of MTurk survey respondents
Answering Mobile Surveys With Images
Instrumental and Expressive Traits, Trait Stereotypes, and Sexist Attitudes
The Matthew Matilda Effect in Science
Using Crowdsourcing Websites for Sociological Research
Gaydar and the Fallacy of Decontextualized Measurement
Narrating histories of women at work
Images that Matter
Striking a Responsive Chord
Status, communality, and agency
A model of (often mixed) stereotype content
Photography and Sociology
Automating Inequality
On the Ethics of Crowdsourced Research
Hearing Gender
The Geometry of Culture
Why Status Matters for Inequality
What Should an Anthropology of Algorithms Do
Social Change and The Persistence of Sex Typing
Machine Vision and Social Media Images
Visual Images of American Society
Doing Gender
The Gender Revolution
Unpacking the Gender System
The End of the Gender Revolution? Gender Role Attitudes from 1977 to 2008
| Unique citing works | 19 |
|---|---|
| Citations per year | 3,8 |
| Citation span | 2021 - 2026 (6) |
| Citation velocity | current |
| Highly cited | No |
| Citation types | Neutral: 19 |