Algorithm Rates Trustworthiness of Wikipedia Pages

paleshadows writes "Researchers at UCSC developed a tool that measures the trustworthiness of each Wikipedia page. Roughly speaking, the algorithm analyzes the entire 7-year user-editing-history and utilizes the longevity of the content to learn which contributors are the most reliable: If your contribution lasts, you gain 'reputation,' whereas if it's edited out, your reputation falls. The trustworthiness of a newly inserted text is a function of the reputation of all its authors, a heuristic that turned out to be successful in identifying poor content. The interested reader can take a look at this demonstration (random page with white/orange background marking trusted/untrusted text, respectively; note "random page" link at the left for more demo pages), this presentation (pdf), and this paper (pdf)."

Light Bulb Moment

Someone should make a wikipedia entry for this algorithm to see how trustworthy it is.

Re:Light Bulb Moment

Sounds crappy. Let's say you expose some important misdeed. You're likely to be edited out by an army of paid staff who keeps an eye on the 'net. (don't tell me I'm paranoid because i saw it happening and read about stuff like that in the news, even slashdot). You are not contributing much else to wikipedia because you simply wanted to expose what's in your knowledge, so you'll end up with a low karma.

Anyway, i guess it'll be another pagerank or slashdot filter affair. People trying to beat it, devs trying to make it better.

The plus is, there is not only wikipedia. You can always search the rest of the web.
The minus is, you search the rest of the web with google which is equivalent if not worse.

We need a good search engine on top of a tor network, and bandwidth to make it run smooth. Not many other way to achieve real net freedom.

Re:Light Bulb Moment

Sounds crappy. Let's say you expose some important misdeed. You're likely to be edited out by an army of paid staff who keeps an eye on the 'net

Nope. If you post one misdeed and that gets edited out, such is life but shouldn't affect your credibility that much because everyone is always getting edited out a few times in the long run.

However, if you edit hundreds or thousands of different articles and people leave you alone, o great guru, you're good.

Wikipedia's ultimate strength depends on the community's desire for good information, readiness to stomp on crap, and will to contribute. Conversely, Wikipedia would decay if people didn't give a rat's ass about Wikipedia and let it go to ruin like an unweeded garden. This mechanism of quality control needs to be applied down the hierarchy of categories, subcategories, and articles. It's understandable that certain areas will have more pristine content overall while other areas will be populated with childish and wanton ideas. Thus, a contributor evaluation program can be tested.

Seems a bit dangerous

>If your contribution lasts, you gain 'reputation,' whereas if it's edited out, your reputation fails

And the editor wars start ...

Re:Seems a bit dangerous

Yeah, that is a bit of a "dangerous" way to go about rating the content, however I think it could be a step in the right direction. If this can be improved, perhaps the site will gain a better reputation in the eyes of professors. Now, I don't doubt that there is a lot of misinformation on the site (intentional or otherwise); however, a good deal of the information I have used for research papers or to quickly check something seems to be confirmed elsewhere (texts, journals, etc).

Hmmm... A reputation metric...

It'd be nice if it could be generalised to other sites...
 

Godwin's Second Law

Every paper touting automatic adjustments for gaming the system becomes obsolete the moment it is published.

(Godwin didn't publish this, but I might get around to editing his Wikipedia entry to say that he did).

7 years???

I've been noticing some of the edit histories for articles that are 5 years old on Wikipedia stop well before 5 years ago. Were some of the edit histories been lost or deliberately truncated?

Doesn't take into account common myths

So, if there is a myth that a lot of people believe is true, then it will stay up there as it is not challenged. So, it still gets reputation, and therefore more credibility, making it more likely that the myth will be perpetrated.

Also, if someone hasn't noticed something that is wrong on an esoteric entry, it will also be given credibility, and once again be more likely to be considered to be fact.

While you could add voting to the algorithm to have people vote on whether it is true, that still gets destroyed by someone who just votes because they think it's true, not because they have verified it.

Either way, it potentially gives additional credibility to something that may be very wrong.

It doesn't have to be perfect

No algorithm, except maybe personally checking every single article yourself, will ever be perfect. I suspect that the stuff you talk about will be very rare exceptions, not the rule. In fact, one of the reasons that it is so rare is because people who know what the actual truth of a matter is can post it, cite it, and show it for all to see that some common misconception is, in fact, a misconception. This is much better than, say, a dead tree encyclopedia where, if something incorrect gets printed, it will likely stay that way forever in almost every copy that's out there. (And, incidentally, no such algorithm can exist, since dead tree encyclopedias generally don't include citations and/or articles' editing histories.)

The goal wasn't to create a 100% perfect algorithm, it was to create an algorithm that provides a relatively accurate model and that works in the vast majority of cases. I don't see any reason this shouldn't fit the bill just fine.

Seems to work ...

Seems to work, the entire page turned orange.

hmmm...

They should just call it wiki-karma.

#REDIRECT

It appears they include #REDIRECT pages; the very first page the random link took me to was Cheliceriformes [ucsc.edu], with the #REDIRECT line in orange. Seems an easy way to gain trust, once a redirect is created it is hardly ever changed.

I dunno about this system.

Does it take into account magnitude of error corrections? If major portions of someone's articles are being rewritten, that's a good reason to de-rep them. If someone makes a bunch of minor spelling or trivial errors, then that's not necessarily a reason to do so.

And, of course, there is the potential for abuse. If the software could intelligently track reversions and somehow ascribe to those events a neutral sort of rep, that would probably help the system out.

As it stands, they're essentially trying to objectively judge "correctness" of facts without knowing the actual facts to check. That's somewhat like polling a college class for answers and assigning grades based on how many other people DON'T say that they disagree with a certain person in any way.

I suspect this heuristic measures....

the relative controversy of the item being edited.

If I edit a history page of a small rural village near where I live, I can guarantee that it will remain unaltered. None of the five people who have any knowledge or interest in this subject have a computer.

If I edit an item on Microsoft attitude to standards, or the US occupation of Iraq, I'm going to be flamed the minute the page is saved, unless I say something so banal that noone can find anything interesting in it.

But my Microsoft page might be accurate, and my village history a tissue of lies....

Tuned for Subject Matter

Sounds like a worthy start to the process of introducing more trustworthyness into Wikipedia entries, but this maybe needs tuning for content type too.

Afterall just because someone is a reliable expert at editing the wikipedia entries on Professional Wrestling [wikipedia.org] or Superheroes [wikipedia.org] doesn't necessarily mean we should trust their edits on, for instance, the sensitive issues of Tibetan sovereignty [wikipedia.org].

Unpopular but neutral points of view?

I realize that an encyclopedia by definition will always emphasize the established majority opinion about any given subject. But it seems that this tool might strengthen majority opinions beyond what is reasonable. If you happen to edit an article by adding valid but unpopular dissenting points of view, and the other contributors are sufficiently boneheaded, you lose karma (or whatever the tool calls it) for no good reason. This might then easily develop a life of its own, and you are screwed.

Tyranny of the majority

Although this method will certainly help filter pranks and cranks, it won't help if the "consensus" among wikipedia authors is wrong. If a true expert edits a page, but the masses don't agree with the edit, they will undo the expert's addition and give the expert a low reputation. Thus, the trust rating becomes a tool for maintaining erroneous, but popular ideas.

That said, I can't help but believe that this tool is a net positive because it makes points of debate more visible. One could even argue that it literally highlights the frontiers of human knowledge. That is, high-trust (white) text is well known material and highlighted (orange) text represents contentious or uncertain conclusions.

Re:Tyranny of the majority

Yes, this system demonstrates the correlation between the content and the majority opinion, not between the content and the correct information (assuming such objectively exists).

Of course, if you take as an axiom that the majority opinion will, in general, be more reliable than the latest random change by a serial mis-editor, then the correlation with majority opinion is a useful guideline.

Something that might be rather more effective, though perhaps less practical, is for Wikipedia to bootstrap the process much as Slashdot once did: start with a small number of designated "experts", hand-picked, and give them disproportionate reputation. Then consider secondary effects when adjusting reputation: not just whether something was later edited, but the reputation of the editor, and the size of the edit.

This doesn't avoid the underlying theoretical flaw of the whole idea, though, which is simply that in a community-written site like a wiki, edits are not necessarily bad things. Someone might simply be replacing the phrase "(an example would be useful here)" with a suitable example. This would be supporting content that was already worthwhile and correct, not indicating that the previous version was "untrustworthy".

Algorithms are handy

when things can be quantified and measurable. I've always wondered about the algorithm of a brand's worth. What is the logo's value, in relation to the slogan, and the consumer experience?

For instance, Google has a strong brand, despite their hideous logo and "Don't be evil" slogan, because the consumer experience is so good. Coca-Cola, on the other hand, score big with their logo's distinctive cursive script, despite ongoing critisms of its health effects and numerous allegations of wrongdoing by the company. And their product just isn't that good.

Man, I would loves me an algorithm for that.

A reasonable first step...

...but call me when there's a tool to measure the truthiness of an article.

Goddamn...

How did they pass up the chance to name this algorithm "Truthiness"? [wikipedia.org]

Don't Care.

I might give a damn if Wikipedia editors had any actual interest in keeping articles truthful.

I do not trust wikipedia on any "divisive issue"

unless it is consistent with what I already know to be true or have had time to verify against other sources.

too many zealots rule certain categories and unfortunately too many of the same are the very powers that be.

This will promote one thing

Groupthink.

Maybe in the future

What we really need is some sort of algorithm that compares new information to that which is already stored. It then could test hypotheses to gain further understanding. Unfortunately a machine with enough processing power to run this "critical thinking and understanding" algorithm would be impossible to build with today's technology. We would need a new type of processor that has maybe billions of "organic neurons", it would need to be equipped with highly sophisticated sensors, a method of self transportation, self-healing and even it's own energy production system which could harvest energy indirectly from the Sun. We can only dream of such technology being available to everyone.

Should be called "stability"

"trustworthiness" doesn't enter into whether something gets edited out, for precisely the same reason a need for this is perceived at all: it can be edited by anyone!

Nice idea, but ...

what if I am very trustworthy but can't spell or use proper grammar?

Isn't this old news?

For a site that prides itself on being up their with announcing new things, this is really pretty much old news.

Algorithm doesn't prove what it thinks it does

That algorithm is a model that does not match real world data. It might be useful to measure who has protection from the bureaucracy, but it won't and can't decipher how true something is simply by how many times and at what frequency people scribble over it. This algorithm is psuedo-scientific, by assuming a premise without investigating the veracity of said premise, and then running away with it as if it were a proven one.

It's progress over edit counts

One big problem with Wikipedia has been that editor status, and promotion to "adminship", is based on edit counts, the number of times someone has changed something. The editors with huge edit counts don't generally write much; it takes too long. Slashdot karma is a more useful metric than edit counts, but Wikipedia doesn't have anything like karma.

I'd suggested on Wikipedia that we needed a metric for editors like "amount of new text that lasted at least 90 days without deletion". This UCSC thing is a similar metric.

And now it'll no longer work

now the algorithm is publicly known, anyone can make pages that abuse it somehow ^^ and then they can update it to counter those abuses etc... let the war begin!

Never confuse popularity with factual truth

They might be somewhat correlated, on a statistical basis, over
many cases, but there are many individual cases and times
when the currently popular view is wrong and the lone
wolf opinions are later proven to have been correct.

This algorithm would seem to be more of a popularity contest
than a truth finder. I think we have to be very wary of
the truth by mass agreement theory.

Hint: Remember the "weapons of mass delusion" ?
I bet someone commenting that the US government is lying
through their teeth about it would have been re-edited
pretty quick.

PageRank

The algorithms looks very similar to the Google's Pagerank. Take edition time as inverse of links to/from, and the whole concept looks very similar. The question is, PageRank was terribly biased once people started to automate cross linking, will this algorithm performs better against biased editors?

WIKI's top 100

Does anyone else find this list hilarious? http://hemlock.knams.wikimedia.org/~leon/stats/wik icharts/index.php?wiki=enwiki&ns=articles&limit=10 0&month=08%2F2007&mode=view [wikimedia.org] WikiCharts -- Top 100 -- 08/2007 Views per day Percent Title 389 659 ± 1% 3.7114% 1. Main Page 17 773 ± 3% 0.1693% 2. Harry Potter and the Deathly Hallows 11 368 ± 4% 0.1083% 3. Wiki 10 995 ± 4% 0.1047% 4. Harry Potter 9 649 ± 4% 0.0919% 5. Transformers (film) 5 286 ± 6% 0.0504% 6. Naruto 5 173 ± 6% 0.0493% 7. Wikipedia 4 427 ± 6% 0.0422% 8. Deaths in 2007 4 119 ± 6% 0.0392% 9. United States 3 827 ± 7% 0.0365% 10. Harry Potter and the Order of the Phoenix (film) 3 714 ± 7% 0.0354% 11. Sex 3 616 ± 7% 0.0344% 12. List of sex positions 3 584 ± 7% 0.0341% 13. Hypertext Transfer Protocol 3 535 ± 7% 0.0337% 14. The Simpsons 3 519 ± 7% 0.0335% 15. YouTube 3 486 ± 7% 0.0332% 16. Bleach (manga) 3 422 ± 7% 0.0326% 17. Guitar Hero III: Legends of Rock 3 227 ± 7% 0.0307% 18. List of characters in the Harry Potter books 2 838 ± 8% 0.0270% 19. The Simpsons Movie 2 789 ± 8% 0.0266% 20. List of Konoha ninja 2 789 ± 8% 0.0266% 21. List of Akatsuki members 2 773 ± 8% 0.0264% 22. Optimus Prime 2 692 ± 8% 0.0256% 23. Harry Potter and the Half-Blood Prince 2 676 ± 8% 0.0255% 24. Seven Wonders of the World 2 514 ± 8% 0.0239% 25. Chris Benoit 2 449 ± 8% 0.0233% 26. Harry Potter (character) 2 416 ± 8% 0.0230% 27. 50 Cent 2 368 ± 8% 0.0226% 28. Megatron ... so all wiki users are nerdy harry potter fans interested in sex? ha!

Do they track popularity of topics?

I could submit nonsense on a variety of obscure topics, with low odds that anybody will find and correct them, thereby building up a great reputation. I wonder if their system accounts for that.

This is starting to sound like Karma for wikipedia.

I found another flaw in the algorithm

On reviewing the demo, it would seem that the untrusted i.e. frequently
changed sections are essentially the sections that people care about
or (in our present society) have more knowledge about or insight into,
so people want to tweak those sections.

So ironically, the algorithm will flag as untrustworthy the most relevant
sections of articles. The "don't know - don't care" parts will be virgin
white and pure.

Compliance vs Compression

This algorithm is measuring compliance with the Wikipedia dispute processing norms -- not "trustworthiness". A better measure of "trustworthiness" of a passage is its consistency with the rest of the body of human knowledge -- which is most strictly measured by the degree to which it is not a special case within a compressed representation of that knowledge. This is the basis of the Hutter Prize for Lossless Compression of Human Knowledge [hutter1.net]. The Hutter Prize is currently using a 100M sample from Wikipedia as its corpus.

Spelling Mistakes?

What I want to know is if it is smart enough to distinguish edits that correct spelling and grammar mistakes from those that change content.

In particular I'm worried that the system will undervalue the information from people whose edits are frequently cleaned up by others even if that content is left unchanged.

Pseudonyms?

Wouldn't one find an entry, for example, about the early history of the WWW by the genuine Sir Tim Berners-Lee to be considerably more trustworthy than one by signed by some anonymous WikiWonderBoy?

I don't think the algorithm takes that into account.

Other uses in software development

It could also be useful for the trust worthiness of code from various authors! - Poorly written/designed code will probably have more faults and require more edits.

Re:If

Whereas the implementation of "+1 funny" will be the end of the information age.

Re:algorithm

No, actually you could use a dummy account to smear anyone's reputation by constantly re-editing their page back to whatever you want. By fighting over the content of a page, you effectively decrease the reputation of both parties. Since the dummy account isn't a real person, you are safe to throw it away after you are finished.