search

very-good-science / data-hazards

Data Hazards is a project to find a shared vocabulary for talking about worst-case scenarios of data science - and to use that vocabulary to help people understand and avoid Data Hazards.
https://datahazards.com
Other
32 stars 14 forks source link

Hazard Idea: Dealing with upsetting/triggering data? #206

Open HuwWDay opened 1 week ago

HuwWDay commented 1 week ago

(Inspired by the symposium on researcher well-being in mental health research)

Symbol idea: Could be something like someone studying a mirror with a clipboard, or someone with a clipboard stood at the bottom of a podium (indicating that they're doing a study of someone higher up). Unsure here, very open to suggestions

Description: Dealing with sensitive data that might be upsetting; health etc.

Examples: Any project dealing with marginalised people or people of ill health, in particular if the data scientists involved are a part of that group (so mental health researchers who have mental health conditions or social scientists investigating discrimination of a group that they are a part of)

Safety Precautions: Nina would know this better than me! But some example from one of the talks WhatsApp Image 2024-06-27 at 08 25 58

dsmukilan commented 1 week ago

This is a good idea, also for moderating data used for training, it is necessary to have this hazard marked to help annotators. Examples : Violent content, medical images, etc