Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Facebook was in talks with hospitals about a proposal to share data

cnbc.com

Post reply on HN
Print view View original

111–120 of 120 posts

Re: Facebook was in talks with hospitals about a proposal to share data

#111

Post by philipodonnell »

chickenfries wrote: post #100
philipodonnell wrote: post #91

Earlier quoted context omitted.

> the result of having HN filled with these Facebook non-stories is that I'm going to stop taking an interest in them. > the public at large do feel that they are surprising and shocking. Not OP, but I don't think the point is that its a non-story to the public at large, I think the point is that its a non-story to anyone reading HN, which is decidedly not the public at large. I get where they're coming from. If I go…

> If I can't trust HN to filter for only truly important stories The HN frontpage is a pretty poor proxy for "importance." It's simply whatever the users of HN find interesting. AFAIK, there is no "only upvote truly important" stories rule.

That's fair. I was using important to mean "worth spending the time to read" as my interests are fairly well aligned with the HN community (at least as far as my interest in Facebook-related news). Though over time audiences change, so as more of these non-stories continue to proliferate, perhaps HN is becoming more targeted to the interests of the general public and I should accept that or move on.

Re: Facebook was in talks with hospitals about a proposal to share data

#112

Post by 411mrc »

gethesmane wrote: post #94
marcell wrote: post #6

The pile on continues... (1) This is probably a project by a small "research" group at Facebook. The goal of that group is probably to publish papers in a Psych journal or something like that about how they were able to correlate anonymized medical data with Facebook feed updates. Tech companies have these research groups for prestige, they are not central to the company mission. (2) According to the article, the pro…

(1) This is probably a project by a small "research" group at Facebook. With blessings from on high, no doubt. (2) According to the article, the project was never actually started. Oh yes it was -- the FB spokesperson said it was in the "planning stages". That's quite definitely a form of "starting" (especially for large companies). If a local waste disposal company were to acknowledge that it was in the "planning" s…

"Hey, major waste company, how about you partner with Facebook and let us sift through people's trash using AI robots before it goes in the landfill? It's for....uhhh, science...yeah, thats the ticket"

Re: Facebook was in talks with hospitals about a proposal to share data

#113

Post by leesalminen »

794CD01 wrote: post #108
ShabbosGoy wrote: post #105

Earlier quoted context omitted.

It’s easy to say that when you are financially secure and comfortable.

What kind of person is facebook interested in hiring who can't become financially secure and comfortable working elsewhere?

Agreed. If you can get hired as an engineer at FB, you can get hired as an engineer at a less shitty company.

If FB started hiring unskilled laborers and promised to train them up to be a software dev, I could see this argument having some weight. But, AFAIK, they don’t.

Re: Facebook was in talks with hospitals about a proposal to share data

#114

Post by a-priori »

Joeboy wrote: post #44
blub wrote: post #13

Earlier quoted context omitted.

Facebook has been screwing with people's privacy for years, so it's only fair that they are now criticised for years on end. This has got to be a new Facebook apology meme: the pile-on. Until Facebook and privacy is regulated in the US, like GDPR does it for the EU, the pile-on should continue. Must continue, no matter how uncomfortable it is for the pro-Facebook, pro-ads or pro-spyware people.

For me, the result of having HN filled with these Facebook non-stories is that I'm going to stop taking an interest in them. As far as I can tell the situation with Facebook is substantially the same as it's been since it started. Users willfully broadcast information using Facebook, and sometimes Facebook uses that information in ways its users didn't intend or expect. Among the recent slew of dramatic stories I hav…

Is it not shocking to you how blatantly Facebook breaches people's expectations about the use of their data?

People share data with Facebook for a particular, immediate benefit to themselves. I share my location so my friends can see where I am, I post my photos so my friends can see what I'm doing and who I'm with, I share my contacts so I can find my friends, etc. In and of itself, this should be fine and safe to do. The problem comes when Facebook takes the data that was given to them for one purpose, in one context, and they use it for another purpose now or in the future.

I can't make informed consent when it comes to data, because the real value of data only comes from when it's aggregated with other data -- either my own over time, or other peoples'. I can't know what incremental effect this datum has when it's combined with everything else Facebook knows about me, and all their other users, and run through their current or future machine learning algorithm. So, it's impossible to know whether it's in my interest to disclose any particular bit of information to them.

Details that are innocuous to human eyes can be very salient to algorithms. I might disclose a set of data points and never make any connection between them. I might mention I feel tired on one day, and write with a negative tone on a few other days, and wake up (i.e, open Facebook for the first time in the morning) later than usual. Without knowing this, that's enough information for Facebook to make a confident inference that I'm depressed, an inference that amounts to discovering private information I never intended to reveal. Of course, they don't disclose that they know that about me, but they do use it against me. They may target ads for anti-depressant drugs, or they may invisibly bias my news feed to have more negative content.

That's even assuming I'm aware that I'm disclosing information at all. If I log into Facebook to see what my friends are up to, then close the tab and start browsing the web, Facebook knows where I go on the web any time I visit a page with Facebook comments even if I don't post any comments. The content of the page, combined with other data they know about me and the other visitors to the site, can be combined to make inferences about me, my interests and hobbies, my sex or sexual orientation, race, socioeconomic class, medical conditions, vices, and so on.

We can't expect every person to become experts on data analysis so they can fully understand the implications of disclosing their data.

Re: Facebook was in talks with hospitals about a proposal to share data

#115

Post by throwaway84742 »

froindt wrote: post #48
downandout wrote: post #42

Earlier quoted context omitted.

Just an aside...of the many things I have entered into Facebook over the years, I am 100% certain that I have never given them my height or pre-existing medical conditions.

There may be some other ways to link it up, at least with a degree of confidence. It really depends on what information is shared from the medical community and by the patient on Facebook. A couple points of speculation: * Facebook may possess a machine learning algorithm which can estimate weight from pictures. Getting within 5 pounds would eliminate most other people. * Facebook could make photos of you and estimat…

Not to mention location metadata on pics taken at or near doctors offices, and “private” messages in Messenger which they admitted they read.

Re: Facebook was in talks with hospitals about a proposal to share data

#116

Post by Xylakant »

d33 wrote: post #63
Xylakant wrote: post #37

Earlier quoted context omitted.

Even if you salt the hashes. The problem is that the number of inputs is limited and it's trivial to enumerate over the input values. Let's take a contrived example: We have the data of a small, entirely made-up island where only two families live, so we have two surnames. Let's name them Foo and Bar. Now, they have an entirely funny tradition, they all get first names based on the order in which they were born (1).…

Doesn't it only apply if the salt is known or constant? If I passed you hash(x) + aes(random_salt), would this attack work?

The salt must never be a constant, the entire point of a salt is that two identical inputs do not hash to the same value. However, it must be stored alongside the hash, so that you can later verify the hashed value. Many modern password hash functions (bcrypt for example) do store the salt as part of the hash.

Re: Facebook was in talks with hospitals about a proposal to share data

#117

Post by cmpxchg »

The initial partnership was with Stanford hospitals. Many in the HN community live in the Bay Area and may have used these medical facilities. If you are concerned, other boards are suggesting patients can contact:

James Laflin, Stanford School of Medicine Ombudsperson: jlaflin@stanford.edu / 650-498-5744

David Entwistle, CEO Stanford Healthcare: 650-723-4000

Re: Facebook was in talks with hospitals about a proposal to share data

#118

Post by itbeho »

marcell wrote: post #6

The pile on continues... (1) This is probably a project by a small "research" group at Facebook. The goal of that group is probably to publish papers in a Psych journal or something like that about how they were able to correlate anonymized medical data with Facebook feed updates. Tech companies have these research groups for prestige, they are not central to the company mission. (2) According to the article, the pro…

They deserve a pile on. I don't want any social media company having any access to my health data, especially if I haven't authorized it.

(1) There is no guarantee they will accurately associate that "anonymized" data with my profile.

(2) There is no guarantee they will "do no harm" with that data. It's a way to run-around existing HIPAA protection and something a lot of organizations would pay for if they could.

Re: Facebook was in talks with hospitals about a proposal to share data

#119

Post by d33 »

Xylakant wrote: post #116
d33 wrote: post #63

Earlier quoted context omitted.

Doesn't it only apply if the salt is known or constant? If I passed you hash(x) + aes(random_salt), would this attack work?

The salt must never be a constant, the entire point of a salt is that two identical inputs do not hash to the same value. However, it must be stored alongside the hash, so that you can later verify the hashed value. Many modern password hash functions (bcrypt for example) do store the salt as part of the hash.

That's not the point. Salt is constant, but different for each entry. They can encrypt the salt and when they share it with hospitals, those can't reverse the hash but FB can. Doesn't it solve the problem?

Re: Facebook was in talks with hospitals about a proposal to share data

#120

Post by RoyTyrell »

sidlls wrote: post #65
RoyTyrell wrote: post #52

Earlier quoted context omitted.

Insurance companies already have tons of relevant medical data on their clients, there's no reason why whatever FB can provide them is going to be miraculously more telling than actual medical data such as their current and past diagnoses and medical conditions.

Insurance companies are perhaps surprisingly often ignorant about their own medical data and what it means for their clients or patients (except for what it means in terms of revenue). Entire profitable businesses (who aren't insurers) sell analytics products based on these data to companies because insurers don't have the means or know how to.

> Insurance companies are perhaps surprisingly often ignorant about their own medical data and what it means for their clients or patients (except for what it means in terms of revenue).

You may very well be correct, but how do you know that?

> because insurers don't have the means or know how to.

Again, where do you get your insight on the inner-workings of insurance companies in general from?

I'm not disagreeing with you outright. I just wondering if you're guessing or you have actual knowledge to the one or more health insurance companies and their actuarial and analytics processes.

Post reply on HN