,
7 tweets,
3 min read
Read on Twitter
In new research that guesses someone's appearance from their voice, the data comes from YouTube videos. @jane_c_hu at @Slate talked to me about the ethics & legality of researchers mining public content. This particular case has some interesting nuances... slate.com/technology/201…
The dataset used came from a *different* project, a " large-scale audio-visual dataset comprising speech video clips with no interfering background noises." There are no details about the data collection beyond that the videos were public. looking-to-listen.github.io/avspeech/
One critical thing to consider in ethical decisions about collecting public data is the purpose of the research--what you're going to use the data for. But what if you release a dataset publicly in the spirit of open science? Should you think about how others might use it?
Another researcher is quoted about the decisions she made to use YouTube data, and noted that her IRB "had no precedent" for dealing with this, and therefore she had to set her own ethical standards because she knew no one else would. Yes! We can choose to be thoughtful!
I'll also pose a wrinkle to my suggestion that inform & consent can be pulled apart... @moduloone and I found that many Twitter users just wanted to *know* that they were part of research. But are there times when harm only occurs because of finding out about the research?
And all of this points to a really tricky problem right now in mining public data. Competing values between more open science (making data available, e.g., for replication or innovation) and protecting the rights and privacy of the people who create that data.
Here is my plea to researchers using public data: Write about the decision in your paper. Don't just say "it was public" - when you write about data collection in your methods section, include justification for why your use of this data was appropriate.
