Fact-checking chatbot “Meiyu” shuts down dubious family texts
wsj.com
wsj.com
His uncle is 100% correct. Adding a bot to a conversation thread is very likely literally connecting a hose up to your family discussions and sending all of them to a random startup. You wouldn't add in a random human fact-checker, and even with the illusion that the only entity reading your texts is a robot, there's no guarantee that will always be the case.
"Ms. Hsu, the developer, said Meiyu isn’t designed to collect personal data on users." 1) objectives change, 2) ownership changes, 3) it very likely collects personal data whether it's meaning to or not. Brilliant. I literally could not think of a better target for a state actor than something like this, which could give pretty deep insight into what people are thinking about you when they talk privately.
Is that the same as "doesn't"? And does that apply to all the other components that get access to the chat logs it sends back home?
I thought even then they didn't count it as "collected" until a human sees the result.
So they could ~~collect~~ record all your conversations, ~~analyze it with a computer~~ have a computer analyze it, and then then only have a human look at it if they think they will be able to justify having collected it.
Yeah, I was never quite clear on whether it meant "scraped for data" or "scraped for data and then that data was used". I tried to look it up before I posted, but since "we don't collect..." seems to have been a lie under any of those definitions, I'm not convinced it actually had a clear 'technical' meaning.
Of course, if reading algorithmic output without specific records doesn't count as seeing the data, there's always the possibility that info is making it all the way to the 'trigger a drone strike' step without ever being "collected"...
To heck with "Chinese traditions", "openly disputing facts with elder relatives" does not go down great in the West either, unless you've got a particularly feisty uncle who likes political debates.
Literally my first thought on seeing the headline was "wow, sounds like a great way to get disowned" and that was me projecting my American context onto it.
Have these people never heard of the phrase "pick your battles"? A bot has no tact, and will pick fights with fairly trivial nonsense and deeply problematic lies with the same assiduousness. Lots of things are not, strictly speaking, true. Not all false beliefs are damaging in the same way. Being technically correct is the worst kind of correct.
I was envisioning the opposite - like posting anything critical of the government's handling of Tiananmen Square would have your little uncle spybot posting "corrections" to the chat explaining the government sanctioned view.
===
The examples given are "contrarian bot" rather than "fact checking bot."
e.g., "The doctor quoted in this post does not have proper qualifications." Fine, but that doesn't mean he's wrong.
e.g., “The internet has a lot of information on drinking water. Doctors say not all of it is credible.” No kidding. But this doesn't mean "stay hydrated" was bad advice.
I'm having difficulty finding the value of something like this is. I suppose if you're too timid to push back against other people it might be nice to have a bot to do it. Like the example where it's culturally inappropriate to push back on your elders, so you introduce a piece of software to contradict everything they say.
Fine, maybe that feels good in a strange way, but it seems dysfunctional, passive aggressive, and unproductive. The bot isn't giving advice, it's just saying "that's YOUR opinion" to everyone else's advice
===
- Can you find the same stuff on a more-reputable site? (Perhaps the bot doesn't know about those "user blog" sections sometimes hosted under news-site domains?)
- Can you find a different attribution for the same quote that doesn't throw a 'credentials' flag? (Perhaps less information is better, because writing "a doctor said" will impede checking up on the speaker's credentials?)
- How about just contorting the phrasing of sentences until the bot can no longer extract anything useful about their meaning? What does it do with double negatives and nested clauses?
I don't want this bot in my group chats, but I could imagine using it as an adversary to refine a trolling bot. If you wanted to train up a realistic version of Shiri's Scissor, this might be an effective way to craft posts that are especially hard to refute.
Alert fatigue kicks in easily under the best circumstances, and "an annoying bot cluttering a group text" is hardly the best of circumstances. If the bot limited itself to high-confidence assertions that a claim is factually false, I'd be interested. (I wouldn't use it, the social and privacy aspects guarantee that, but I'd be interested in the tech.)
But "not all information about water is credible" isn't fact-checking; it's a tool you could duplicate with keyword recognition and a list of random stock refutations. Maybe I'm overzealous, but I'd consider removing an actual human from a group-chat if they consistently made that sort of post.
1. both auntie "Meiyu" and "CoFacts" are open source projects, we share the source code on GitHub.
2. can't disagree the search results might be manipulated, but for now, there's just limited volunteers to "clarify" most of rumors on CoFacts.
3. Logs: that's a certain point. Since I deployed this project on Heroku (which remains limited logs) I still have no time to format the logs. this comment about log storage makes sense to me, I won't rebut this point of view is nothing to me, it matters,will update it for sure.
I think wsj actually skip most of background of why we do this. Since not only Taiwan's election has been effected by deliberatedly manipulated rumors, endless medical misinformation. furthermore, most of medias own strong biases in Taiwan, everyday these's (even quite rough) fake news, bombing everyone's brain day after day until you give up to clarify anything.
And these medias own resources, centralized to publish fake news, to fight with them in fact is quite hopeless, unless we try to decentralized our information, allow everyone got the chance to verify these and speak it loud.
So Meiyu, imo at least it clarifies rumors / misinformation for you, and repeat it tirelessly. It's annoyed for everyone I can totally understand (That's another reason I name it as Auntie "Meiyu" by very common senior generation name, I'd like This tiny service becomes a bit heartwarming and friendly), but we must go for it. nor only someone's life / health would be misled by misinformation, but even our island would be ruined by elaborate politic rumors as well.
I suppose the idea is that in a largish group chat, refuting a post is personal while adding the bot is a general change. But most of the examples look blatantly targeted, and even absent that it's an implicit declaration of "you people are full of it and need correction".
Contradicting people can at least be done tactfully, picking battles and making sure you're definitely right before starting them. Adding a "nuh-uh" bot strikes me as emphasizing insult instead of accuracy; it's on the same level as rolling your eyes at someone's comments but refusing to actually disagree with them.
Or https://i.kym-cdn.com/photos/images/newsfeed/001/191/035/135... in bot form.
I get the noble cause to combat false information right at the roots, especially when people are too polite to do it themselves, but making a bot to be rude and obnoxious for you is still rude and obnoxious.
very interesting..
They could make people believe that WW2 never happened or some other garbage. It all depends on who'll be verifying whats true and whats not.
"Drink a lot of water!" "Put on socks before you go to bed!" "Netty pot!" "Drink lots of tea!" All of these are kind of common knowledge. I think the act of encouraging people to rest up and take care of themselves is just a generic way to show you hope they get better. And maybe, if they really are so sick that they aren't thinking clearly, it'll then serve as a reminder to get some tea.