Yesterday the credible web group got together to do some though excercises as we get closer to building a credibility rating engine.
I would rather link the credibility rating to a person canonical website than a social media silo but the tools we are trying to develop are directed more at social networks and journalistic publishers.
Still having your own place online is the first step to credibility and I think we should try to "attach" ratings to a person canonical ID. Not having a website, while a reflection of priviledge, is in itself a loss of credibility.
We then tried this thought excercise and didn't finish in the hour call because this stuff is hard. Common problems we had : should credibility be domain specific? is action required? how big of a scale do we need to ensure enough variance but not threaten internal consitency?
Some quick bytes:
- you can't separate expertise from domain
- we discussed relying on self affirmed expertise. "I know Sandro. I know he is an expert in web standards. Sandro says he is an expert in auto mechanics. Therefore I trust Sandro with my car." This to me is way too fraught with engendered and cultural bias. A white male from the dominant narrative is way quicker to label themselves an expert
- The action kept throwing me. Must we have an action be involved on a credibility rating?
- The US Georgia Governor said, "We just learned how contagious covid-19 is" I now trust him less. Not going tochange my actions. I would love to give him 4 pinnochio noses in my credibility browser extension
Is 1st degree and 2nd degree network effect enough?
I feel like I kind of already include credibility statements for the people I follow: https://
- each person is arranged by domain of knowledge. I do include their first person expertise statements.
- I have four different levels of trust "acquaintance, colleague, friend, muse" I use xfn for this like rel="muse"
This isn't a solution for the cred web work, just another example.. It isn't json-ld. It parses into JSON well so converting it to extra data to stick somewhere in the file wouldn't be hard once there are tools to try http://
I follow Sandro and he is more than acquaintance. He follows Subbu and says he is an expert in journalism. Therefore when I follow Subbu he gets some credibility points toward journalism. The more 2nd degree points Subbu gets the higher Subbu's credibility rating on a goven domain.
This is a little suspect to confirmation bias. Hard to eliminate the "being human part" and the idea that there is ever an objective credibility rating is nonsensical.
Can we annotate for credibility?
Back to the Kemp example. If I read this: https://
<blockquote> <a class="h-card" href="https://gov.georgia.gov/" tag="dumbass"> Georgia Gov. Brian Kemp </a> said he wasn't aware that asymptomatic people could transmit coronavirus as he announced he was preparing to issue a state-wide shelter-in-place order. <cite> <a href="https://www.newsweek.com/georgia-governor-didnt-know-asymptomatic-people-transmit-coronavirus-1495695"> Georgia Gov. Brian Kemp Says He Didn't Know Asymptomatic People Could Transmit Coronavirus </a> </cite> </blockquote>
If enough 2nd degree people gave Brian Kemp a dumbass rating in my network then my browser would then add four pinochios, four pants on fire, whatever, after any Brtian Kemp mention that also linked to any georgia.gov page
Sorry for not using JSON-LD. I do not know it well enough yet. Easier for me to think in way where I can make my text and metatdata human readable at the same time. I will try and figure out how to mark up my example and add the data back in again as JSON-LD.