|
|
|
|
|
|
by btown
1701 days ago
|
|
|
What are the institute's thoughts, if you've formed them, on end-to-end encryption, especially as it applies to social media where the line between group and group-text blurs? I feel it's an incredibly nuanced topic that's become incredibly polarizing in recent days with some of Haugen's comments. On the one hand, in favor of E2EE, companies can and will use the content of messages, if they have access to them, to micro-target suggested content to users, and this can lead to increased levels of misinformation being promoted to people who have engaged with misinformation. And of course there's the government surveillance angle, which is an entirely separate story! But if you remove the signals in that content by encrypting in a way that is opaque to the platform, do you substantially reduce the ability to microtarget? Very possibly not, given the amount of graph data the social media company has anyways about group members independent from the content itself. And encryption gives the social media company the ability to wash its hands of any responsibility or awareness of content. Assuming it were easy to technically achieve (which is a huge leap, to be fair!) do you think it better serves the definition of integrity you've adopted, that a social media platform have the majority of its content end-to-end encrypted, or not? |
|
|
I don't think we have gotten an institute stance on very much relating to E2EE. Our community advisory board (composed of integrity workers) is the moral core of the organization. So far, when we say "the institute has a stance on X", that has meant "the advisory board signs off on X, and that X represents a good faith consensus view of workers in the industry".
We don't yet has a doc that lays out why we think integrity and privacy can coexist nicely. Speaking only for myself, I think the answer lies in careful design. As an example, you could see, via research and experimentation on FB Messenger, that messages that are forwarded in chains of > N are just empirically overwhelmingly likely to be bad faith, spammy, etc. You could then take that finding to WhatsApp, Signal, etc, and then bake in changes to the UX that make it slightly more annoying to forward messages if they've been on a reshare chain of N/M. That kind of stuff.
There's also some consideration to group size -- if a group is 5000 people large on, say, Telegram, it might be encrypted, but it's no longer really private. Maybe it should be treated differently? Unclear, let's think and research about it.
I think a rough consensus we might move towards is treating messaging differently than broadcast, and also treating broadcast features inside of messaging apps differently than straight up messaging themselves.
But again, those are just some of my more idle thoughts. There are members and fellows who are better experts on this particular subject than I am.
Does that make sense? Is that helpful?