Boffins: You'll see, TWITTERNET! We'll get the TRUTH out of you... While even emergency services have embraced Twitter as a way of conveying information to people quickly, the premature obituaries and malicious rumours seen on the social network every week show that you cannot trust what others tweet. That's why computer boffins from five European universities have started work on a lie …
They also need to see how rumors are re-tweeted. The speed with which (true or false) news travels could provide some useful insight into how "believable" the rumor actually is. The best lies and hoaxes always look a lot like truth or contain some modicum of truth (social engineering relies on similar tactics).
I'd also like to see a similar system applicable to internet news, forums and other info sources (i.e. blogs), which are rapidly becoming as bad or worse than traditional journalism for spreading incorrect or downright false information.
Brave new world
Actually one of our major use cases is internet forums - and we're particularly interested in detecting mis/disinformation in the context of health advice. Not only related to things like the huge MMR debacle or the unsubstantiated links between alzheimers and aluminium, but other rumours - and catching them when they're nascent, before there are enough voices to give a perception of authority to a wrong message.
You handle a trusted account going rogue by building a 'forgetting factor' into the model. You can weight the forgetting factor to guard against things like someone using low-value trades to build up a good reputation before blowing it all on a big heist or for sensors in a network going a bit wonky and returning bad readings.
Rosie
When people post stuff about how crap certain politcal parties are but always forget to mention another one, you can be sure they are usally sock puppets for the one that never gets mentioned.
You have to troll hard* to get them to admit to being Kippers.
*using different shiny lures
Much of the world's gone mad!
Addiction to Twitter, Facebook, etc. has spread with such contagion that heroin/cocaine dealers must be reviewing their marketing strategies and wondering what they've done wrong!
I like many others abandoned these cesspits over three years ago, regulators and do-gooders should do the same.
disinformation - 5%
misinformation (retweets of disinformation) - 94%
truth - 1%
percentage of 'truth' that is relevant, or that normal people actually give a shit about - 0%
percentage of time wasted whilst using twitter - 100%
percentage of time and money wasted doing this research - 100%
(as is my time writing this - but I feel a bit better for it)
The road to Hell is paved with good intentions.
I can accept that the idea behind this project is a good intention, unfortunately I cannot see it end well.
Either it never does work and, like Artificial Intelligence, will always be a few decades away, or somehow, someone does bend space/time / deal with the Devil and produces a working solution that is proven to always be right.
At that point we are one step away from never being able to utter another word on the Internet under pain of perjury. Nor will we ever skirt our fiscal duties as all our administrative declarations will be permanently wired to the worldwide Lie Detector. To summarize : we will have made the worst possible social tool one can think of.
Given that this whole idea will most likely work about as well as web translators (and probably only in English), I think the Truth Hydra (tm) will remain dormant for the time being.
Either it never does work and, like Artificial Intelligence, will always be a few decades away, or somehow, someone does bend space/time / deal with the Devil and produces a working solution that is proven to always be right.
Or it will operate somewhere between the two ridiculous extremes you pose, and possibly be useful.
And, incidentally, you're wrong about AI, except in the case of "strong AI", which relatively few reputable researchers have shown much interest in after the heyday of "symbolic manipulation" approaches waned. AI has produced a great deal of successful work which has proven useful in commercial and scientific applications.
Sounds like a text book case of trying to solve a problem which doesn't exist. Or in this case, to think of ways to structure and organize a medium which exists only because of its lack of usual hierarchical structure and artifices of quality selection. All these things take time and a lot of subjective context. And it's not like algorithms would be trusted by crowds to add fairness. Otherwise the same technology could way easier be used for blogs and newspapers (more context there usually). But the solution for all that is called "common sense" although crowds in the midst of revolutions are not known for it.
"But the solution for all that is called "common sense" although crowds in the midst of revolutions are not known for it."
Hmm, 'common sense' has problems in it usually being only common to one or a very few people.
Not that common and rarely sensible.(it's a form of social bullying)
If people plan on checking over social media for pattterns then it's just as easy to introduce false trails.
Gosh, if only some of the many researchers working in this area had considered the obvious and facile objections that the keen intellect of the Reg's armchair experts have perceived!
With such genius readily available in the Reg forums it's a wonder anyone else even bothers to try.
This sounds like the sort of thing that trust and reputation models (TRMs) are meant to deal with. Calculating the reliability or otherwise of a peice of information is (one of) the use cases they have been proposed for.
I haven't got my (freshly completed) dissertation to hand but I have a few (academic) references in there that cover it. I thought it sounded like 'one of those' when I heard the news on the beeb this morning. Three years is probably do-able, provided there are sufficient people available.
Rosie.
Yes, along with sentiment analysis, influence models, information-entropy estimators, and similar techniques. There are any number of research projects (academic and commercial) looking to sift the social-media wheat from the chaff; I've done some of that myself, in a small demonstration-purposes way, and I know other people working in the area.
This particular sort of work does go beyond simple stuff like product-review sentiment analysis - to do it properly, you need to be quite sophisticated in determining things like tone and footing, which means relatively deep analysis of linguistic structure, identifying things like implicit nominal predicates and determining conversational entailment. Those sorts of things are still cutting-edge research. But you can do a decent first approximation with algorithms that are well-established and easy to scale out using platforms like Apache UIMA or homegrown systems build on top of stock map/reduce engines like Hadoop.
NOW we know the truth about why touchscreens are being pushed so hard! They want to get closer to your fingernails in case they need to pull a few to get your attention. I knew all along it was one way of building a fingerprint database for everyone in the world but this is just going too far!
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
