Hacker News Reader

Comment by hliyan | Hacker News Reader

I was hoping the next generation of wikis (especially as used by wikipedia) would introduce an underlying 'fact unit' [1] layer on which articles are built, effectively separating edit wars from verification wars.

[1] Not the same as a 'knol', which as defined by Google was not a unit. A real knol would be a single atomic fact.

View on HN

The idea of "atomic" units of fact turns out to be really slippery. In the end you're just shifting consensus to questions about what things really are verifiable facts. And then, what is a valid statement you can build on those facts.

Wittgenstein and Russell tried to build philosophical atomism and it fell apart.

http://plato.stanford.edu/entries/wittgenstein-atomism/#DisL...

You're making a common mistake, which is to think of Wikipedia as a collection of facts. It's a collection of verifiable statements from reliable sources, or reasonable summaries thereof. "Reliable" is a collective judgment call, and even "verifiable" can be tricky sometimes if it's not an online resource or if we're summarizing.

So what you really want is a kind of source tracing. You want to know that when one statement is put into question, everything that followed from that statement is also questionable.

I don't think you are contradicting the parent. It's useful to acknowledge the slipperiness in some form.

With encyclopedic content, the consensus 'fact' is useful, plus the level and strength of general agreement over time is useful, plus the alternative 'facts' are useful.

Wikipedia is too much a collection of the most obstinate authors' latest opinions. It's saving grace is the 'History' tab, which adds a lot of value, at least for me.

Try listing the atomic facts (ignoring, for simplicity's sake, their underlying facts, or any references to authoritative sources) for the following statement. As an exercise that may possibly illuminate why that hasn't happened.

"37 minutes ago hilyan was hoping that the next generation of wikis (especially as used by wikipedia) would introduce an underlying 'fact unit' [1] layer on which articles are built, effectively separating edit wars from verification wars.

[1] Not the same as a 'knol', which as defined by Google was not a unit. A real knol would be a single atomic fact."

I will interpret a lack of responses as evidence for failure. ;)

This should be compiled to a flat list of facts:

    (declare hilyan-hope)
    (declare footnote-constraint)
    
    (let [t universal-timestamp]
         (said AndrewOMartin t
               (status hoping hilyan
                              (- t (minutes 37))
                              hilyan-hope))
               (said google (undefined-past)
                     (let [knol (gensym)]
                          (not (unit knol))))
               (let [knol (gensym)]
                    (single-atomic-fact knol)))
    
    (def hilyan-hope
         (let [next-wiki-generation (gensym)
               fact-unit (gensym)
               article (gensyn)]
                 (introduces next-wiki-generation fact-unit)
                 (underlies next-wiki-generation fact-unit)
                 (footnote-constraint fact-unit)
                 (underlies fact-unit article)
                 (cause fact-unit (separate verification-wars edit-wars))))
    
    (def footnote-constraint [fact-unit]
      (let [knol (gensym)]
           (!= knol fact-unit)))

The DSL could be improved :)

Edit: So a couple of things about this exercise: yes, it took me more time (and space) than your phrase. On the other hand, I specified some extra information. Some ambiguities about your phrase needed to be explicit, so that's a win for a more formal language. Also, there's some more effort involved, so I could have the time to think more about what I say. It would be great to have an interactive tool to think and write in a more formal language, even if it is not meant to be compiled to assembler.

I have a side project that touches this tangentially (the code is made up, though!), but I can't find a lot of time for it :(

Does anybody know if something like this exist? Somethig like a wiki/REPL/mindmap sandbox for concepts.

Take a look at Attempto Controlled English.

Thanks!

I need an IDE for this.

I've been really interested in this area. Is your side project online? Even if it is just made up code, I'd love to see what you've been thinking about.

Cracking this problem could bring about an entirely new era of debate & communication.

For those of you trying to represent this, it will be important to explicitly represent context. A fact exists only in a context. The above statement is a good example. Another is, "Bill dreamed that he thought his dog was playing in the front yard." Which could be extended with "but it was actually in the back yard." and further extended with "When he woke up he found that his dog was licking his face."

Rather than creating more wiki, the solution to creating a better Wikipedia is to create a system with less wiki. A large part of Wikipedia’s problems stem directly the “article” format of most wiki pages (and Fereated Wiki maintains the same page metaphor). This gives presentation problems in that it forces the web page to look like a book page, but more importantly, allows any editor to veto any content on the page. This gives rise to real or imagined experts believing they “own” the page, and from that groupthink, harassment, bias naturally follow. See my post: http://newslines.org/blog/wikipedias-13-deadly-sins/

Sites like Larry Sanger's infobitt.com attempt to break down news stories into their facts, by letting contributors asses the factuality of each piece of information. However, I believe that fact checking current news stories is over-rated. Why do we need to fact-check the old story, when a more accurate news story will come along anyway? Again this stems from the wiki, formatting the page as an article, rather than, say, a stream of information.

There are other solutions that can mitigate problems that arise from using wikis, such as assigning anonymous editors to check content, and paying editors to contribute. I have implemented many of these on my site, Newslines (http://newslines.org) which crowdsources news, but uses each news event as the primary data type.

The absolutely vital importance of fact-checking current news stories, AKA "the old story", is in forming a measure of the reliability and veracity of the sources of this information, which can be factored in to future reporting.

Well, there's Wikidata: www.wikidata.org

Reply on news.ycombinator.com