Thread by @schachin

"Knowledge-Based Trust: Estimating the Trustworthiness of Web Sources"

If @Google is testing this in the SERPS (no evidence that they are now w Dr Oz, Dr Axe, & David Wolfe w/ top 10 placements on health/medical searches), but if so? Shame on them.

(see next tweet for link)

Knowledge-Based Trust: Estimating the Trustworthiness of Web Sources
arxiv.org/pdf/1502.03519…

They look at these "facts" using "triples" so any large corpus of similar published articles could alter what is seen as a "fact"

So why is that bad?
It could create a normative corpus based on current "thought".

Think about the "Study" that showed vaccines caused autism.
That then showed up in a "million" articles?

Then Google checks the triples and interprets it as the "fact"

What would have happened to the work that showed it was not?
Possibly? Never seen.

And what about what is currently considered "fact" but morally wrong? If this was 200 years ago, it would pull pro-slavery & anti-female normative views as "fact".

Would abolitionists or suffragists have their views surfaced or would they have been seen as "not factual"?

And in the areas of health? They change sometimes daily, there would be no good way to make sure new & innovative research was ever seen.

I refer you to my thread on Omega Oils & the changing views on those.
What is "fact" here?
threadreaderapp.com/thread/1161757…

PLUS add this @defcon talk on fake science.

"...Fake Science. This talk presents the findings a team of investigative journalists, hackers and data scientists who delved into the parallel universe of fraudulent pseudo-academic conferences and journals"

Bad actors could quite literally create fake facts with little worry the alternatives would be seen.

Most of the research is circa 2015, so hopefully @Google understood the ramifications of trying to determine "fact" & kept it to research --- because it is dangerous.

However, here is ONE paragraph I could see being used.

".. KBT can identify sources with trustworthy data, even though they are tail sources w low PageRanks"

They could avoid the "fact" issue & just surface content that is trustworthy, but w few links. That could be a good use.

Again nothing in the results reviewed shows this is true NOW, but if being tested? Shame on @Google

Creating opportunity for bad actors to create "truths" using methods already available would be far more destructive than fake news bec the real info would be buried on page 10.

Not to mention this can be scammed even more easily than links ... It would in fact be easier to do w a little scripting, some AI content writing, and a good set of PBNs. The same methods used to scam the link profile would be used ---->
Then add fake science AND BOOM "truth"!

For those interested in the debate? Here is a good read on the paper & a good discussion with SMEs in Knowledge Graphs and Google in the comments seoskeptic.com/knowledge-base…