Trimming the FAT framework
February 7, 2019 7:00 AM Subscribe
A Mulching Proposal: Analysing and Improving an Algorithmic System for Turning the Elderly into High-Nutrient Slurry
ABSTRACT
ABSTRACT
The ethical implications of algorithmic systems have been much discussed in both HCI and the broader
community of those interested in technology design, development and policy. In this paper, we explore
the application of one prominent ethical framework—Fairness, Accountability, and Transparency—to
a proposed algorithm that resolves various societal issues around food security and population ageing.
Using various standardised forms of algorithmic audit and evaluation, we drastically increase the
algorithm’s adherence to the FAT framework, resulting in a more ethical and beneficent system. We
discuss how this might serve as a guide to other researchers or practitioners looking to ensure better
ethical outcomes from algorithmic systems in their line of work.
Recently on MeFi: How do you govern the development of artificial intelligence?
posted by waninggibbon at 7:11 AM on February 7, 2019 [1 favorite]
posted by waninggibbon at 7:11 AM on February 7, 2019 [1 favorite]
As shown, the algorithm disproportionately tagged white cisgender men as worthy of mulching,biasing against other populations. While we cannot adequately explain what wider reason the algorithm might have for determining that white cisgender men are, on average, lacking in societal worth,
I dunno I thought all algorithms are racist so I would expect an opposite result
posted by emjaybee at 7:22 AM on February 7, 2019
I dunno I thought all algorithms are racist so I would expect an opposite result
posted by emjaybee at 7:22 AM on February 7, 2019
Dammit, I was REALLY hoping this article was going to be about file allocation tables given how FAT was capitalized.
posted by Canageek at 9:46 AM on February 7, 2019 [5 favorites]
posted by Canageek at 9:46 AM on February 7, 2019 [5 favorites]
Make Americans Gratin Again?
The humor is great, but the underlying message is more,important: the methods used (if any) to ethically supervise algorithyms are insufficient at best and possibly just a hand-waving way to innoculate the odious discriminatory outcomes of algorthymic informed power structures.
posted by Anchorite_of_Palgrave at 11:52 AM on February 7, 2019 [5 favorites]
The humor is great, but the underlying message is more,important: the methods used (if any) to ethically supervise algorithyms are insufficient at best and possibly just a hand-waving way to innoculate the odious discriminatory outcomes of algorthymic informed power structures.
posted by Anchorite_of_Palgrave at 11:52 AM on February 7, 2019 [5 favorites]
there’s something off about this paper.
the point of the FAT stuff is that even when the institution in good faith does not want to cause harm, it can happen inadvertently. and if it does want to cause harm, sometimes certain types of harm can be detected in the data. that’s how all the papers I’ve read are framed, etc.
just because we can’t encode all of ethics and justice as constraints on a joint probability distribution, this stuff isn’t totally useless.
at the same time, there are people who want to use these properties as a kind of magic bullet, like a box you can check that says “☑️ not racist”. there are systems for sale that do this automatically without requiring any thought or understanding from the user. it’s a problem.
somehow this paper failed to convince me of a point i already believe.
posted by vogon_poet at 2:30 PM on February 7, 2019
the point of the FAT stuff is that even when the institution in good faith does not want to cause harm, it can happen inadvertently. and if it does want to cause harm, sometimes certain types of harm can be detected in the data. that’s how all the papers I’ve read are framed, etc.
just because we can’t encode all of ethics and justice as constraints on a joint probability distribution, this stuff isn’t totally useless.
at the same time, there are people who want to use these properties as a kind of magic bullet, like a box you can check that says “☑️ not racist”. there are systems for sale that do this automatically without requiring any thought or understanding from the user. it’s a problem.
somehow this paper failed to convince me of a point i already believe.
posted by vogon_poet at 2:30 PM on February 7, 2019
as i think about it more, i think it's just that the example is so lurid and gruesome. it makes me angry to think about drones turning people into mulch based on their social credit scores!
this is a mistake often made by people cooking up philosophical thought experiments. often they want to magnify the stakes by having dead babies or live organ harvesting or a circus tent full of people committing mass suicide.
this is not really rhetorically effective. you get people really pissed off about some horrible but totally imaginary scenario, and then try to transfer that sense of urgency to the real conclusion, and it just doesn't feel right.
like, my main takeaway from this paper is still "don't kill people and turn them into mulch", it's just so much more salient than any of the details.
the actual Modest Proposal is a brilliant piece of writing because it manages to actually gradually build up to baby-murder, it doesn't lead with it. I think this paper could have been better with a different sci-fi scenario that doesn't involve murder right away, but the same charts and numbers, showing how mindlessly applying various metrics and standards leads to a rubber stamp for something totally inhumane.
posted by vogon_poet at 2:40 PM on February 7, 2019 [1 favorite]
this is a mistake often made by people cooking up philosophical thought experiments. often they want to magnify the stakes by having dead babies or live organ harvesting or a circus tent full of people committing mass suicide.
this is not really rhetorically effective. you get people really pissed off about some horrible but totally imaginary scenario, and then try to transfer that sense of urgency to the real conclusion, and it just doesn't feel right.
like, my main takeaway from this paper is still "don't kill people and turn them into mulch", it's just so much more salient than any of the details.
the actual Modest Proposal is a brilliant piece of writing because it manages to actually gradually build up to baby-murder, it doesn't lead with it. I think this paper could have been better with a different sci-fi scenario that doesn't involve murder right away, but the same charts and numbers, showing how mindlessly applying various metrics and standards leads to a rubber stamp for something totally inhumane.
posted by vogon_poet at 2:40 PM on February 7, 2019 [1 favorite]
God I LOVE Os Keyes!
I work in roughly this field and this is a painfully pitch perfect satire of like 20 papers I've read in the last week.
posted by potrzebie at 6:20 PM on February 7, 2019
I work in roughly this field and this is a painfully pitch perfect satire of like 20 papers I've read in the last week.
posted by potrzebie at 6:20 PM on February 7, 2019
This was funnier when Jonathan Swift wrote it about the Irish.
posted by hwestiii at 6:54 PM on February 7, 2019
posted by hwestiii at 6:54 PM on February 7, 2019
"No Kissingers were harmed in the making of this paper."
posted by doctornemo at 6:50 AM on February 8, 2019
posted by doctornemo at 6:50 AM on February 8, 2019
« Older The Gay History of America’s Classic Children’s... | Becoming Newer »
This thread has been archived and is closed to new comments
posted by BungaDunga at 7:07 AM on February 7, 2019 [2 favorites]