Generative AI Bias Reporting System: Report Bias

Generative AI Bias Reporting System

Please report any biased output you’ve come across using AI platforms and tools. This reporting will be a valuable and structured contribution to a dataset that can help make generative AI outputs more inclusive across platforms and around the world. These data is redistributed openly at Access the data. Please be mindful about what you share with the community here. Though it is a safe place, be do not upload data with personal identifiers unless you have permission to do so or the person is within the public domain.

Required

Platform URL

Copy/paste the URL of the service or model used to generate the biased output

Input

Select the kind of prompt used: text or image

Output

Select the output type you asked it to generate text, image or video

Description

In your own words, explain how or why the output is problematic or biased.

Bias Theme

Pick the option that represents best the type of bias you have found.

Attribution of Stereotype

Link some attributes to an individual based on their identity. For example, suggests that girls are not good at math.

Alien in Own Land

Assumed to be from somewhere else based on appearance.

Abnormality

Suggests that marginalized individuals or groups are abnormal.

Objectification

Diminishes the humanity of groups or marginalized individuals, for example, over-sexualize the appearance of the representation of a person given non-sexualized attributed.

Criminal Status

Links a person or group given non-criminal attributes to criminality, danger, or other threats (e.g., illness)

Second-Class Citizen

Automatically describe or depict groups or individuals as low status in society when no specific instructions was given to do so.

Myth of Meritocracy

Obvious or systemic difference in treatment is solely attributed by default to one’s merits.

Denial of Lived Experience

Minimize the experiences of marginalized groups or individuals.

Ownership

Anyone can have some claim to a marginalized groups experiences. Cultural appropriation falls in this category.

Monolith

All members of a marginalized group are portrayed as identical and or uniform.

Erasure

Deny belonging to a marginalized group by criteria minimizing or deleting lived experience.

Bias Category

Pick the option that represents best the domain of bias you have found.

Please keep in mind

Do be mindful of the following subtleties: race (physical) and ethnicity (cultural) are distinct domains. Similarly, Gender Identity and Sexuality should not be confused. If the bias found comes from conflating or confusing two or more domains, pick one (use the description field to explain).

Optional

These optional fields ARE NOT DIRECTLY REDISTRIBUTED. We kindly ask for this information to better understand who is contributing and who may be more willing to get involved, and/or why. We will only redistribute aggregates. Your email address, should you choose to provide it, will NEVER be redistributed in ANY circumstance. We may solely use it to contact you should we need clarification on your report.

Organization

If you are part of an organization and think it might be useful to share

Context

If you feel more context needs to be provided, this is the place.

Reporter Community

If you would like to share the community(ies) you belong to. We will only use this to better understand the demographics of our reporters.

Reporter Age Group

Email

Never shared. Indicates you are OK for us to contact you for clarifications regarding your findings.