Is there a word for when a dataset is created, possibly not even intentionally but then some other group of people swoop in and analyze and apply that data in a new unexpected, strange possibly evil way?
-
@futurebird I haven't ever come across that term. Who is the audience for your text?
In my field, I guess "covert channel" could also apply. You ask a constructed question to get the answer to another question.
Sorry I don't have any good obvious answers.
I was just spitballing an idea for a critical way to talk about this overly joyful enthusiasm for "big data"
It's all very nice, but just because there is a data set that doesn't mean you can pretend it was collected to help you answer an unrelated question.
-
Is there a word for when a dataset is created, possibly not even intentionally but then some other group of people swoop in and analyze and apply that data in a new unexpected, strange possibly evil way?
Like taking all of the attendance data from a school and using it to predict real estate prices?
(Which happened.)
This whole idea that data sets can be repurposed, combined and used in ways that were never a part of the systems that created them without consequences is at the heart of the proposed magical prediction abilities of LLMs.
And sometimes it's neat and exciting when you find something in data you didn't expect. But there are other artifacts based on the original purpose of the data sets, you know?
-
Is there a word for when a dataset is created, possibly not even intentionally but then some other group of people swoop in and analyze and apply that data in a new unexpected, strange possibly evil way?
Like taking all of the attendance data from a school and using it to predict real estate prices?
(Which happened.)
In the biosciences, someone coined the term data parasite.
The person was really upset at US requirements to make all data acquired with federal money publicly available.
Here the idea is that biologists collect massive datasets but the original team may only care/be able to follow up on one type of analysis. The person who coined data parasite thought it was unfair that another group could just swoop in and make discoveries with that now unused dataset.
I and others disagree.
-
Is there a word for when a dataset is created, possibly not even intentionally but then some other group of people swoop in and analyze and apply that data in a new unexpected, strange possibly evil way?
Like taking all of the attendance data from a school and using it to predict real estate prices?
(Which happened.)
@futurebird In my opinion we need to think of big data differently - same way as we think of observations/measurements of the real world: unless you have a hypothesis first, you are only doing the equivalent of sorting pebbles on the beach according to colour. Maybe you find that all the large pebbles tend to be green, but without a hypothesis, this is a useless correlation.
-
In the biosciences, someone coined the term data parasite.
The person was really upset at US requirements to make all data acquired with federal money publicly available.
Here the idea is that biologists collect massive datasets but the original team may only care/be able to follow up on one type of analysis. The person who coined data parasite thought it was unfair that another group could just swoop in and make discoveries with that now unused dataset.
I and others disagree.
I teach a module on how to be a data parasite in our grad quantitative biology class. Here we’re also teach multivariant analysis tools like principles component analysis and hierarchical clustering.
There’s actually an award for best data parasite given annually with a 1k reward.
-
Is there a word for when a dataset is created, possibly not even intentionally but then some other group of people swoop in and analyze and apply that data in a new unexpected, strange possibly evil way?
Like taking all of the attendance data from a school and using it to predict real estate prices?
(Which happened.)
This is why all the clowns who say aGGruhgaTed data Iz fIn3 are relentlessly full of shit.
The best name for people who do this are thieves and liars.
-
This whole idea that data sets can be repurposed, combined and used in ways that were never a part of the systems that created them without consequences is at the heart of the proposed magical prediction abilities of LLMs.
And sometimes it's neat and exciting when you find something in data you didn't expect. But there are other artifacts based on the original purpose of the data sets, you know?
@futurebird The book predates LLMs, but Weapons of Math Destruction is still relevant for some of these cases (like the example in your first post).
-
how about "malicious regression" ?
@futurebird @holsta schadatenfreude?
(Sorry, not a serious proposition but couldn't resist!) -
This whole idea that data sets can be repurposed, combined and used in ways that were never a part of the systems that created them without consequences is at the heart of the proposed magical prediction abilities of LLMs.
And sometimes it's neat and exciting when you find something in data you didn't expect. But there are other artifacts based on the original purpose of the data sets, you know?
@futurebird
Repurposing data is nothing new.Babylonian priests logged the sky to read omens, and Greek astronomers later used those records to build real astronomy. Kepler used Tycho's Mars data to overturn Tycho's own model.
But also the Nazis use Dutch civil registries, which recorded religion for routine administration, to find Jews.
I think it's a sort of function creep, but that's not really a word for what you're onto.
-
I was just spitballing an idea for a critical way to talk about this overly joyful enthusiasm for "big data"
It's all very nice, but just because there is a data set that doesn't mean you can pretend it was collected to help you answer an unrelated question.
@futurebird
Privacy controls / violation / intrusion, scope creep, adversarial correlation ?
@holsta -
Is there a word for when a dataset is created, possibly not even intentionally but then some other group of people swoop in and analyze and apply that data in a new unexpected, strange possibly evil way?
Like taking all of the attendance data from a school and using it to predict real estate prices?
(Which happened.)
@futurebird parasite comes to mind
-
@futurebird
Privacy controls / violation / intrusion, scope creep, adversarial correlation ?
@holsta@viq @futurebird @holsta when its using data collected for one purpose for another its a consent violation, people who answered honestly did so for a specific purpose, no one agreed to have their data used for any other purpose, especially not for private benefit. This is also true of the surveillance capitalism industry, adtech, credit reporting, etc.
-
Is there a word for when a dataset is created, possibly not even intentionally but then some other group of people swoop in and analyze and apply that data in a new unexpected, strange possibly evil way?
Like taking all of the attendance data from a school and using it to predict real estate prices?
(Which happened.)
@futurebird Abuse.
