Data Ethics Documentation
October 31, 2023 3:30 PM Subscribe
I am looking for codes, policies or other documentation on data ethics, digital ethics, or ethics of ethics of artificial intelligence from any of these:
* Companies focused on data, such as data vendors, etc.
* Law enforcement organizations.
* Local governments.
I am probably not interested in anything academic, if they already have restrictions about human-subjects research. I am also not interested in nonserious answers or anything about lacks of ethics in business or law enforcement, to whatever degree.
* Companies focused on data, such as data vendors, etc.
* Law enforcement organizations.
* Local governments.
I am probably not interested in anything academic, if they already have restrictions about human-subjects research. I am also not interested in nonserious answers or anything about lacks of ethics in business or law enforcement, to whatever degree.
RELX, the parent company of LexisNexis has their “Responsible Artificial Intelligence Principles” [pdf], though if you ask me they don’t really live up to them…
posted by jedicus at 4:01 PM on October 31, 2023 [2 favorites]
posted by jedicus at 4:01 PM on October 31, 2023 [2 favorites]
You might find Ari Ezra Waldman's recent book Industry Unbound to be worthwhile.
posted by humbug at 4:03 PM on October 31, 2023
posted by humbug at 4:03 PM on October 31, 2023
Hi NotLost, I am super curious to see what you learn! I recently did a presentation on the ethics of the Association for the Advancement of Artificial Intelligance for a grad school course. One of my sources was this HBR article Why You Need an AI Ethics Committee, which is tangental to your research. It provides rationale for businesses to create ai ethics for legal/reputational reasons and provides some suggestions as to how to proceed.
posted by boomdelala at 5:37 PM on October 31, 2023 [1 favorite]
posted by boomdelala at 5:37 PM on October 31, 2023 [1 favorite]
Google's Ethical Principles for AI
IBM, Everyday Ethics for Artificial Intelligence
NYC Local Law 144 re: automated employment decision tools (Slate maintains it's weaksauce)
The SEC's Investment Advisory Committee made suggestions in March in re: Establishment of an Ethical Artificial Intelligence Framework for Investment Advisors
posted by Iris Gambol at 7:36 PM on October 31, 2023 [1 favorite]
IBM, Everyday Ethics for Artificial Intelligence
NYC Local Law 144 re: automated employment decision tools (Slate maintains it's weaksauce)
The SEC's Investment Advisory Committee made suggestions in March in re: Establishment of an Ethical Artificial Intelligence Framework for Investment Advisors
posted by Iris Gambol at 7:36 PM on October 31, 2023 [1 favorite]
All departments of health have these and some are published on their websites (in the US, they have lots of highly identified personal health data for their Epidemiology work, but are required to do public data release so they've thought a lot about potential reidentification and combining their data with other datasets and potential risks to patients with release).
I know CSTE is collecting data policies for their DRPASS workgroup right now if you have any professional connections to public health.
I don't work for an agency anymore or I'd share!
posted by esoteric things at 10:59 PM on October 31, 2023
I know CSTE is collecting data policies for their DRPASS workgroup right now if you have any professional connections to public health.
I don't work for an agency anymore or I'd share!
posted by esoteric things at 10:59 PM on October 31, 2023
Perhaps something here might help?
https://www.cdc.gov/surveillance/pubs-resources/dmi-summary/policy-considerations.html
https://www.hhs.gov/hipaa/for-professionals/privacy/laws-regulations/index.html
Or
https://www.networkforphl.org/resources/topics/covid-19/health-data-sharing-and-privacy/
From that last link, this is a discussion of "grey area" scenarios where there's a need to share data but also potential risks of harm from doing so. https://www.networkforphl.org/resources/health-departments-tools-for-navigating-data-sharing-during-the-covid-19-pandemic/ They're the kind of unpleasant situations where ethics become a lot clearer in my opinion (are you truly prioritizing patient privacy or doing it only when it's easy and doesn't add friction to public health work?).
My favorite identity resolution scenario that is a real situation - how do you/your systems deal with medical data with stolen identities, especially those that have been stolen for years and the person knows and uses the ss#/dob of the other person? It's delightfully and shockingly difficult to write algorithms to person match exclusively without some level of human resolution in edge cases.
I'm not sure if that tickles your data/ethics intrigue, but I find it a delightfully hard problem... far harder than it initially appears, and is the sort of edge scenario where applied ethics really get hammered out.
posted by esoteric things at 11:14 PM on October 31, 2023
https://www.cdc.gov/surveillance/pubs-resources/dmi-summary/policy-considerations.html
https://www.hhs.gov/hipaa/for-professionals/privacy/laws-regulations/index.html
Or
https://www.networkforphl.org/resources/topics/covid-19/health-data-sharing-and-privacy/
From that last link, this is a discussion of "grey area" scenarios where there's a need to share data but also potential risks of harm from doing so. https://www.networkforphl.org/resources/health-departments-tools-for-navigating-data-sharing-during-the-covid-19-pandemic/ They're the kind of unpleasant situations where ethics become a lot clearer in my opinion (are you truly prioritizing patient privacy or doing it only when it's easy and doesn't add friction to public health work?).
My favorite identity resolution scenario that is a real situation - how do you/your systems deal with medical data with stolen identities, especially those that have been stolen for years and the person knows and uses the ss#/dob of the other person? It's delightfully and shockingly difficult to write algorithms to person match exclusively without some level of human resolution in edge cases.
I'm not sure if that tickles your data/ethics intrigue, but I find it a delightfully hard problem... far harder than it initially appears, and is the sort of edge scenario where applied ethics really get hammered out.
posted by esoteric things at 11:14 PM on October 31, 2023
Few more links:
https://www.michigan.gov/mdhhs/keep-mi-healthy/mentalhealth/information-sharing-and-privacy (behavioral health data like substance use/treatment. MI dept of health)
https://doh.wa.gov/data-and-statistical-reports/health-statistics/data-request-faq and https://doh.wa.gov/data-and-statistical-reports/health-statistics/data-types-and-systems (regular health data sharing policies, WA dept of health)
I think this ppt could be useful- it's from MI, about health data sharing when it comes to medicaid performance measures, which require sharing patient data (in most cases aggregated so not an individual privavy concern), and that there's a desire to share publicly to make quality of care more transparent to patients (yea, I know): https://www.michigan.gov/mdhhs/-/media/Project/Websites/mdhhs/Inside-MDHHS/Policy-and-Planning/Social-Determinants-of-Health-Strategy/CIE/Resources/Michigan-Multipayer-Alliance-Guiding-Principles-for-Data-Sharing.pdf?rev=1f2654b2a843407ca9ec81a7c8c8b602&hash=50C2E7F0B1ABDA0E0282CE3F8BD0C853
Those look more like the applied guidelines that I've seen that health departments have internally. Fwiw, I'd expect these to be FOYA-able from your jurisdiction, if you're not in a hurry.
posted by esoteric things at 11:31 PM on October 31, 2023 [1 favorite]
https://www.michigan.gov/mdhhs/keep-mi-healthy/mentalhealth/information-sharing-and-privacy (behavioral health data like substance use/treatment. MI dept of health)
https://doh.wa.gov/data-and-statistical-reports/health-statistics/data-request-faq and https://doh.wa.gov/data-and-statistical-reports/health-statistics/data-types-and-systems (regular health data sharing policies, WA dept of health)
I think this ppt could be useful- it's from MI, about health data sharing when it comes to medicaid performance measures, which require sharing patient data (in most cases aggregated so not an individual privavy concern), and that there's a desire to share publicly to make quality of care more transparent to patients (yea, I know): https://www.michigan.gov/mdhhs/-/media/Project/Websites/mdhhs/Inside-MDHHS/Policy-and-Planning/Social-Determinants-of-Health-Strategy/CIE/Resources/Michigan-Multipayer-Alliance-Guiding-Principles-for-Data-Sharing.pdf?rev=1f2654b2a843407ca9ec81a7c8c8b602&hash=50C2E7F0B1ABDA0E0282CE3F8BD0C853
Those look more like the applied guidelines that I've seen that health departments have internally. Fwiw, I'd expect these to be FOYA-able from your jurisdiction, if you're not in a hurry.
posted by esoteric things at 11:31 PM on October 31, 2023 [1 favorite]
Sorry, you got an area I did a bunch of research on for a previous job:
Ahrq and office of minority health (part of hhs or office of civil rights) have been concerning experts and developing recomendations about AI and bias since like 2017. I attended a meeting in 2023 but this is all I can find of their guidelines. I thought they were excellent, though not being appreciated by ONC/CMS/EHRs - they had something about purpose/intent of the AI being developed that was great, but I'm not digging it out :( https://effectivehealthcare.ahrq.gov/sites/default/files/wysiwyg/impact-healthcare-algorithms-racial-ethnic-disparities-March-2.pdf
UK https://www.openaccessgovernment.org/disparities-against-minority-ethnic-groups/130356/ and a report https://www.minorityhealth.hhs.gov/news/new-report-provides-recommendations-improving-health-equity-racial-ethnic-and-tribal
No surprise, but specific to LLMs in med https://apnews.com/article/ai-chatbots-racist-medicine-chatgpt-bard-6f2a330086acd0a1f8955ac995bdde4d
I'll stop now :)
posted by esoteric things at 11:47 PM on October 31, 2023
Ahrq and office of minority health (part of hhs or office of civil rights) have been concerning experts and developing recomendations about AI and bias since like 2017. I attended a meeting in 2023 but this is all I can find of their guidelines. I thought they were excellent, though not being appreciated by ONC/CMS/EHRs - they had something about purpose/intent of the AI being developed that was great, but I'm not digging it out :( https://effectivehealthcare.ahrq.gov/sites/default/files/wysiwyg/impact-healthcare-algorithms-racial-ethnic-disparities-March-2.pdf
UK https://www.openaccessgovernment.org/disparities-against-minority-ethnic-groups/130356/ and a report https://www.minorityhealth.hhs.gov/news/new-report-provides-recommendations-improving-health-equity-racial-ethnic-and-tribal
No surprise, but specific to LLMs in med https://apnews.com/article/ai-chatbots-racist-medicine-chatgpt-bard-6f2a330086acd0a1f8955ac995bdde4d
I'll stop now :)
posted by esoteric things at 11:47 PM on October 31, 2023
Apologies if you've seen this, or if it's n/a: Executive Order on Safe, Secure, and Trustworthy Artificial Intelligence, 10/30/2023 Fact Sheet
posted by Iris Gambol at 3:22 PM on November 4, 2023 [1 favorite]
posted by Iris Gambol at 3:22 PM on November 4, 2023 [1 favorite]
« Older Genealogy mystery: Find that death certificate! (U... | A candy colored corn they call the candy corn Newer »
This thread is closed to new comments.
posted by NotLost at 3:40 PM on October 31, 2023