Slashdot Mirror
Real-Time Fact Checking With "Truth Teller"
The Washington Post has announced a prototype news application called "Truth Teller", that displays “TRUE" or “FALSE” in real time next to video of politicians as they speak. The Knight Foundation-funded program automatically transcribes speeches and checks the statements against a database of facts. From the article: "For now, the early beta prototype has to be manually hand-fed some facts, and thus only works on topics it has been specifically designed to recognize. Since Congress has yet to pass a budget, and financial discussions are prone to widespread lies and misstatements, Truth Teller is being piloted on the issue of tax policy."
So basically they've made a static page with the word "FALSE" on it.
Slashdot social media options: AIM, ICQ, Yahoo, Jabber and Mobile Text. Why no MySpace?
Wow. This will never be abused. Now we are spoon feeding low information voters?
if(lips_move)
then
display("FALSE");
Science is all about firing a drunk pig out of a cannon just to see what happens.
A good politician, just like a bad philosopher, only makes statements that are beyond truth and falsity.
... it's a sticker that reads "bullshit", you slap it on the screen permanently.
Joke aside, who will debunk the dunkers ? Everything we know is false, for vaster and more elaborate definitions of "false" as science progresses.
Maybe we deserve this world ?
that displays âoeTRUE" or âoeFALSEâ in real time next to video of politicians as they speak.
Few statements can be classified as "true" or "false" exactly. There is always some fraction of bullshit, but the fraction varies:
Politifact has
True --The statement is accurate and thereâ(TM)s nothing significant missing.
Mostly True -- The statement is accurate but needs clarification or additional information.
Half True -- The statement is partially accurate but leaves out important details or takes things out of context.
Mostly False -- The statement contains some element of truth but ignores critical facts that would give a different impression.
False -- The statement is not accurate.
Pants on Fire -- The statement is not accurate and makes a ridiculous claim.
And even this may be too coarse-grained.
True
I think I'd rather IBM's Watson, I think it's shown a lot of promise in natural language parsing and I think it would do a better job than anything The Washington Post can come up with.
Have you ever heard of the concept of garbage in - garbage out? If you ask the majority of people they think that things like return on investment as 'facts', but yet infinite growth in a finite system is impossible, and all economic activity is premised on energy and thermodynamics.
When politicians say things like 'high taxes are hurting business' will the system be 'hand fed' the appropriate answer, regardless of mountains of evidence showing otherwise because it discourages 'non-productive consumption', and that the high taxes pay for government infrastucture, welfare, and investment?
Similarly there have been lots of propaganda referred to as facts in terms of tax policy, by the likes of the 'chicago boys' and people like milton friedman et al, however these people don't believe that economics can be studied empirically, and tax policy as an extension of economic policy.
I have had my share of problems with my local oregon newspaper distorting facts of even its 'politifact', and generaly attacking the institution of government itself as bad, so that it can meet the expectations of the patrons which keep it in business.
Is is any suprise that news media that are conservative make way more ad revenue per viewer than liberal, say for example rachel maddow or the daily show vs fox primetime, even when they have better age demographics of viewership for advertisers?
I had been thinking "wouldn't it be cool if" for a while about something like this, but I would like to see it taken a few steps further. Though it would be rather difficult, wouldn't it be cool if there was a system like this which detected bad/underhanded debating tactics such as straw man, Ad hominem, cherry picking and so on.
That'll be great. Often, such as during the primaries while there are still 8 candidates, or when the two candidates are otherwise tied, I eliminate candidates based on relative honesty. I was wishing for a real-time politifact scroller. They all stretch the truth, of course, but some WAY more than others.
This, or later versions, could really be a boon to voters who aren't really interested in politics, so they often don't know an "obvious" lie when they hear it. For example, in some polls most Obama voters didn't recognize the name of the then-current vice president. How are they supposed to judge the veracity of a candidate's statements when they have no interest in, and little knowledge of, politics? (Not saying they are dumb, they just spend their time on things other than politics.)
I had envisioned some invited experts typing quick notes like "factually false" into a chat type system, but if a machine can be more objective, great.
Fox version!
I bet it'll be a Bill O'Reilly animation shouting...
IAIFARSIJDPOOTV - I Am In Fact A Reality Star; I Just Don't Play One On TV
The thing about politics is that people do not agree on what the facts are, except in the most obvious, simple cases.
Also, people agree even less on the applicability of a certain "fact" to a certain problem.
This will accomplish very little. It may even be counterproductive because it may classify the speech of a true visionary as a lie because the thing is just to dumb.
I suspect a mechanical Turk, or perhaps a piece of performance art.
Sound like total rubbish to me. Politicians do lie sometimes and they even occasionaly tell the truth but mostly they bend the truth out of all proportion. If they make a statement its not TRUE or FALSE usually the answer would be "WELL... ITS COMPLICATED.. it depends how you look at it" In the UK we have a radio show dedicated to statistics called More or Less http://www.bbc.co.uk/programmes/b006qshd These folks can spend half the show discussing the truth behind a single political statement and then sometimes dont come to a firm conclusion
hmmm.
Wouldn't it be almost perfect to just say "false"?
if (person.getJobs().containsKey(Jobs.POLITICIAN)) {
return false;
}
Until they publish the fact database for everyone to see, how can we tell if it's just more editorializing disguised as fact-checking? The example they give on their demo page, "The Recovery Act saved or created millions of jobs and cut taxes for 95 of the American people" is not encouraging. The "saved or created" statistic was widely panned at the time it was first used because these terms were invented by the Administration, they are not standard employment terms that can be verified with empirical data. Later attempts to find the data behind this claim turned up many dubious sources.
Saying one has the "the" truth implies that the facts are undisputed. If we saw the actual fact database, my guess is that something like 10-15% of the statements would fall into that category.
"We receive as friendly that which agrees with, we resist with dislike that which opposes us" - Faraday
"The Knight Foundation-funded program"
I'm sure they have a better AI that can determine the truth...
Why broadcast major events or debates live? Why not delay it, have someone fact-checking the tape, making indications and then broadcasting it to the viewers? Or, perhaps some sort of post-game analysis?
Do we really trust news outlets to be the fact-checkers?
Back in the day, "fact checking" was something a newspaper or such ran against their own product. Readers' Digest was famous for their fact checkers.
If they want to check a speech in real time, why don't they run it on their own reporting and opinion pages?
Going around piously checking everyone else's facts is more creating news than anything else.
Since a fair number of their 'fact checkers' get their facts wrong, who is going to check the truth teller facts?
The difference is that the T/Z interaction details are not facts.
How much money was allocated to the military in the last ten years in budgets?
How much money was spent by the military in the last ten years?
What percentage of the revenue of the federal budget is collected as income tax on those making in the top X% of income earners?
How many people receive social security who also have assets greater than $2,000,000?
Are there more or fewer naval ships in service today than in 2000?
Do the bottom 50% of income households pay zero income taxes? zero federal taxes? zero taxes?
How many days of vacation has the president taken in the past 4 years? Has the president taken more vacation days per year than the previous president?
How many firearms are purchased in the US in a year?
How many intruders are shot by firearms owners defending their property or person?
How many suicides are the result of firearm use? Of poison use? Of jumping off buildings?
What is the national average price of hamburger?
All verifiable facts from reliable, independent sources. Based on the ability of Watson to parse, search, and manage data, I think it's also possible to determine if the data is in question.
As someone who routinely fact checks and is appalled at the gross inaccuracies out there (not just the twisting or cherry picking, but simply wrong) I, for one, welcome our new robotic fact checking overlord(s).
Is it just my observation, or are there way too many stupid people in the world?
No - it's easy: What would you allow your 8 year old (or 12, or 16 year old) to tell you as true?
The top two are TRUE and the rest are FALSE. You provide a "human" error band of a few percent on actual facts (the federal government spent 2.5 Trillion Dollars last year would be true, and I'd take anything from 2.4 to 2.7 as an acceptable answer, since 2.67 was requested and 2.49 was approved).
Why should we allow half or partial truths from our leaders, when we don't accept them from our children?
Is it just my observation, or are there way too many stupid people in the world?
"I had been thinking "wouldn't it be cool if" for a while about something like this, but I would like to see it taken a few steps further. Though it would be rather difficult, wouldn't it be cool if there was a system like this which detected bad/underhanded debating tactics such as straw man, Ad hominem, cherry picking and so on."
(Note to mods: the following strange reply has a couple of layers, so be careful what mod category you use!)
We're talking about full Strong AI, aren't we? Political speech is one of the most difficult categories of speech to process! So wait, we're asking a *computer program* ... built in *three months* ... to fact check *tax policy*?
But then we turn around and say that 30 *years* worth of attempts at AI can't figure out the transcript you had with that girl at the bar? And Siri can't figure out what you meant by "tell me where the nearest RPG game store is that doesn't require getting on ____ train on the transit system"?
Yet we want to believe that *in real time* a computer can deduce a statement like "the economy will have less money in the hands of citizens because some of the payroll tax cuts expired and were not replaced."
My first Journal Entry ever, in 8 years! http://slashdot.org/journal/365947/aphelion-scifi-fantasy-horror-poetry-webzine
It is called Knight Foundation Car - K.F.C.
It also produces very good chicken for Carnivores.
Suppose everything they say is true, and is reported as true, and we all sigh with relief. And then one day, "FALSE" appears on our TV screens. What then?
my .02
As a wise man once pointed out, you can fool all of the people some of the time and some of the people all the time, but you can't fool all of the people all of the time. Rupert Murdoch's news empire has made a fortune for him by following this dictum to the letter. Murdoch has figured out which people he can fool all the time -- angry white males who live in the US. And as another wise man once pointed out, a man will hear what he wants to hear, and disregard the rest. Murdoch isn't worried about his angry white male revenue stream abandoning him simply because somebody fact-checked his propaganda -- Murdoch knows that angry white males will ignore anything that disrupts their vision of the world, and embrace anything that endorses it.
And -- just to affirm the Rule of Three -- another wise man once pointed out that the truth is out there. We already have fantastic sites like snopes and factcheck that eviscerate Murdoch's untruth stream in near real-time. People who can't be fooled all of the time and who don't always disregard what they don't want to hear can take comfort in the fact that sites like these are out there and are accessible to them any time their common sense alerts on one of Murdoch's "facts."
I don't think that anyone is telling you that a computer WILL do it today, only that it would be what is useful. Logical fallacy detection would be, bar none, the most useful tool in debate because it would save so much fucking time if a buzzer went off when something someone says is clearly a bunch of shit from a logical standpoint.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
And I applaud the concept. The wrinkle is of course that some "facts" are highly political in of themselves.
For some reason, they decided not to go with its original name, which was "Fact Hunt".
The problem/irony with this is it's being published by the Washington Post, a company that is not trustworthy.
USA lawmakers will soon pass a law making this a felony.
You must gather your party before venturing forth.
I think such a system should be a required filter before any journalist can publish an article. This filtering system should also red flag words that imply speculation on the part of the author such as "might," "if," or "maybe." And to take that a step further, television news should also have a system that identifies when file footage has been inserted into a report for dramatic effect.
Sometimes they lie, like when Romney told workers that Chrysler was going to be shipping all their jobs overseas to china.
http://www.politifact.com/truth-o-meter/statements/2012/oct/30/mitt-romney/mitt-romney-obama-chrysler-sold-italians-china-ame/
Anyone else getting asked to log into the admin of their wordpress when viewing their about page?
Some people die at 25 and aren't buried until 75. -Benjamin Franklin
There's a reason why, when testifying, you're instructed to "Tell the Truth, the Whole Truth, and Nothing But The Truth".
This only covers the first part and, partially, the third. Telling the Whole Truth is the detail in where the devil lies. Without it is how people can state a single true fact ("Created 4 million jobs!") that doesn't, individually, comport with reality ("U3@9% / U6@14%").
Even this list of "verifiable facts" can have multiple true answers. Politicians can pick and choose their assumptions to fit the propaganda they want to spread.
For example, the amount of money allocated or spent by the military depends an awful lot on what your definition of the military is. Does it include the Coast Guard? National Guard? American Legion? Local Police forces?
How many firearms are purchased in a year? What exactly is classified as a firearm? BB guns? Paintball guns? Airsoft guns? Tanks? Missiles? Do you exclude firearms purchased by the military?
Price of hamburger? Retail or wholesale? Bulk or in patties? Already cooked? Sold at fast food restaurants like McDonalds? Sold at higher-end restaurants like Outback?
And how do you really get a reliable independent source that tells you something as complicated as the number of intruders shot by firearms owners defending their property or person? Even judges and juries have problems determining if these incidents are actually self-defense or not. The probability that the viewer's criteria and the politician's criteria and the fact checker's criteria match for what should really be considered self-defense are pretty slim. Well... maybe not. If you are a Washington Post reader and the politician is a Democrat, the probabilities go up considerably.
Politicians can pick their assumptions so that their statements are TRUE for that particular set of assumptions.
The real problem with this entire scheme is that the Washington Post is going to want to use their own set of assumptions when they assess statements. They are going to want to pick the assumptions that show that their favorite politicians are flashing the TRUE indicator a lot, and their least favorite politicians are flashing the FALSE indicator. TRUE and FALSE have no provisions to explain that the politician is not including air rifles as a firearm and the fact checker is. This is useless from a standpoint of someone who wants to know what the issues really are. It is extremely useful if you are a political newspaper and want to create controversy.
So we should hold them as absolute on all matters of truth?
The problem here is confusing factually accurate with "true". There are lots of facts that are accurate but support a narrative that is misleading or outright false. What most of these "fact checkers" are actually doing is deciding whether the facts sited support the media's own narrative. It is an effort to shut down discussion, not contribute to it.
I think a David-McCandless-style graphic (for instance: https://www.youtube.com/watch?v=pLqjQ55tz-U) superimposed behind or beside the speaker would prove more helpful than TRUE or FALSE. If the speaker is discussing a segment of a national budget - Welfare Reform, for instance - a graphic could highlight the size of the speaker's subject relative to the other elements of the budget and within the total budget. If the speaker is discussing Stranger Killings, this could be highlighted in the context of total deaths per year. Such context might be useful in determining the relative societal weight of the issue, irrespective of truthfulness or accuracy.
The main problem with a binary "true" or "false" is that most things end up being some fuzzy shade of gray that isn't entirely fictional, but profoundly distorts the truth by presenting nominally-correct facts in ways that are egregiously misleading.
They really need to present two axes: technical accuracy, and contextual manipulation. In the upper-left corner (blatantly false, intentionally-manipulative), you have Michelle Bachman ranting about something she made up 15 minutes earlier.
In the lower-right corner, you have a bored, socially-handicapped individual who's brutally honest to the point of hurting his own cause who sees no need to manipulate context, because he already researched his facts, and he knows beyond doubt that he's right & anybody who disagrees with him is an idiot. For obvious reasons, such an individual could never, ever win an election for anything, because he'd piss off even people who admired his honesty and mostly agreed with him.
In the upper-right corner, you have someone who's read "How to lie with Statistics" from cover-to-cover 400 times, and regards it (along with "The Art of War") as one of the great works of human literature. He works as the spokesperson for some high-ranking elected official, or does public relations for AT&T.
In the lower-left corner, you have someone who's kind of a disorganized mess. He unintentionally gets facts wrong all the time, but his motives are largely sincere. Just about everything he says gets ranked as "lie" because the numbers aren't quite right, even though many would argue that he's probably the least evil of the four.
One day, a politician will tell the truth.
Unfortunately, this case was never tested, resulting in a catastrophic failure. We can't really blame the developers, the probability was way below 10^-9 per hour, which is considered to be an acceptable failure rate, even in critical systems.
Most Republicans I know, when faced with facts against their leaders, would simply blow it off, saying I'm listening to much to the "leftist media"
However, on the same token, when I presented Democrats with facts against what their leaders said, they'd tell me to "stop being a sheeple listening to faux news"
The fact of the matter is is that facts don't matter, people will believe who and whatever is more convenient to them.
What do I know, I'm just an idiot, right?
.. I'd love to see voice stress analysis as a standard feature in video apps and smart TVs, activating a discreet little red blob in a corner of the screen when analysis indicates a lie.
LOL. How will they handle half-truths? Political pundits and politicians tend to tell those a lot.
They may have an app that does that, but the results are BULLSHIT.
The media themselves are usually telling lies, so we are supposed to just believe them when they say somebody is telling the truth?
... Republicans almost always lie, and Democrats almost always tell the truth.
sigfault (core dumped)