Re: credibility networks (was Re: Is Alice, or her post, credible? (A really rough use case for credibility signals.))

This document was also discussed at the verifiable credentials meeting this
week:
https://docs.google.com/presentation/d/1jn9DjM-wlZT1B9moBP23qhiB2FZc_H8oqXIZN222a9U/edit#slide=id.ge4a5a0fed4_0_18

I know there is a lot of crossover to the verified credential group here
but I think if we are developing trust signals directed at the author we
should develop our spec to align with VC.

On Wed, Aug 18, 2021 at 12:10 PM Greg Mcverry <jgregmcverry@gmail.com>
wrote:

> We have been playing with the concept of vouch over in the indieweb world:
> https://indieweb.org/Vouch
>
> Different stack since based on webmentions but the workflow pretty much
> the same.
>
> The goal is to create semi-private posts for community members vouched by
> others and as a trust network.
>
> XFN pretty defunct but I use rel="muse" on my poetry follower list as a
> trust signal
>
> https://indieweb.org/XFN
>
>
>
> On Wed, Aug 18, 2021 at 11:35 AM David Karger <karger@mit.edu> wrote:
>
>> We've been working for a few years on this kind of trust network.  I
>> recognize the subject-dependence of trust, but I think that trying to work
>> that into systems being developed now is too ambitious.  Right now the
>> value of a trust network can be demonstrated more effectively by starting
>> with a simpler system that works in terms of generic credibility rather
>> than subject-specific.  What you want are people who know what they know
>> and don't claim to know more.   Yes, you'll lose out on your friend who
>> knows everything about global warming but is anti-vax, but I think there
>> are enough generally trustworthy individuals to drive a network of
>> assessments.
>> On 8/18/2021 9:46 AM, connie im dialog wrote:
>>
>> As an additional thought, perhaps to bridge the exchange between Annette
>> and Bob, and Sandro: one aspect that I see missing in the scenario below is
>> the underlying knowledge/perspective framework or approach that ties
>> signals together: could be understood as a schema or rubric.  This is a
>> different way to tie signals together from trust networks, and is probably
>> underlying those relationships.
>>
>> What I mean by this is: all of the signals proposed are meant to be
>> understood as potential indications of credibility, but they only gain
>> meaning when some of them brought together in a specific interpretive
>> framework.  Implicit in the development of many of the current signals
>> proposed is belief, or trust, in a scientific method of evidence and
>> evaluation of claims using methods such as verifiability. It's also tied to
>> things like expertise and the development of professions.
>>
>> This framework of knowledge is different than a moral order that trusts
>> inherited wisdom, or tradition, for example.  (I'm going to sidestep the
>> elites for now since the power dynamic depends on what kind of elite one
>> is.) Just because they are different does mean that they can't in fact
>> share one or more signals, but the dominance of certain signals over others
>> I think varies.  And because we aren't always consistent, we may hold both
>> of these or more frameworks given a certain context or topic.
>>
>> So I guess I see Bob's suggestion as much in the line of a number of
>> crowdsourced wisdom projects, which can be valuable.  When you think of
>> historical or even current examples, such as genocide reporting, it's very
>> critical to include as many on-the-ground reports as possible, even as
>> those claims also need to be validated as much as possible. In these
>> contexts, there are many indications of what makes for credible witness
>> reports which isn't the same as expertise.
>>
>> But in some cases, on some topics, you can't go with any crowd
>> <https://wearecommons.us/crowd-wisdom-public-wisdom-regarding-misinformation-at-large/>.
>> That is at least if you hold to for example a scientific method of
>> evaluation and validation.  As with Annette, I have no problem with
>> deferring to expertise understood in this framework, and think it's even
>> worth being explicit about the theoretical framework: X claim works if you
>> believe or agree with Y approach.
>>
>> My assumption in the cases of when something is complicated, or new to me
>> is to agree with Sandro but to add on a little more: if he tells me someone
>> is good at something, I'll likely think that someone is good, but what's
>> driving this is trust from experience in his knowledge about certain things
>> at certain times at certain topics (back to the framework or approach).
>>
>> Thoughts?
>>
>> One article that I recently came across seems related --  I just started
>> working through it -- is "Beyond subjective and objective in statistics" by
>> Andrew Gelman and Christian Hennig with a number of responses including by
>> L.A. Paul so sharing in case of interest
>> https://www.lapaul.org/papers/objectSubjectPerspectives.pdf
>>
>> --connie
>>
>> On Tue, Aug 17, 2021 at 10:53 PM Sandro Hawke <sandro@hawke.org> wrote:
>>
>>> It seems to me we can unify these views using credibility networks. We
>>> can let anybody say anything about anything, as long as we only propagate
>>> that content only along credibility network links. I'll simplify a bit
>>> here, saying a "good" source is one which should be believed or one which
>>> has interesting and non-harmful content.
>>>
>>> So let me see content from sources I've personally assessed as "good",
>>> and also from sources my software predicts will be "good".  If I say
>>> Clarence is good, and Clarence says Darcy is good, and Darcy says Edward is
>>> good, then show me Edward's content, sure.
>>>
>>> On the other hand, if there is no one in my network vouching for Edward
>>> in any way, I'm not going to see his content. Essentially, total strangers
>>> -- people with whom I have no positive connection, direct or indirect --
>>> are blocked by default. I'm talking here about content appearing in search
>>> results, news feeds, comments, annotations, etc.  If I ask for something
>>> specifically by URL, that's a different matter. Whoever gave me that URL is
>>> essentially vouching for the content. If they give a link to bad content, I
>>> can push back.
>>>
>>> This general approach subsumes the trust-the-elites model. If someone
>>> only says they trust pulitzer.org, then they'll get an old-media/elite
>>> view of the available content.  If they only say they trust infowars.com,
>>> they'll get a very different view.
>>>
>>> My hope is most people have an assortment of sources they find credible
>>> and the software can help them flag where the sources disagree.
>>>
>>> (This is what I was prototyping in trustlamp. Many details remain to be
>>> solved.)
>>>
>>>     -- Sandro
>>>
>>>
>>>
>>> On 8/17/21 8:46 PM, Annette Greiner wrote:
>>>
>>> I don’t think I have the solution, but I offered my comment to help
>>> better define what would be a reasonable solution. Another way to think
>>> about it is that the signal should not be game-able. As for what you refer
>>> to as “elites” and “hierarchies”,  I have no problem with harnessing
>>> expertise to fight misinformation. Turning up the volume does not improve
>>> the signal/noise ratio.
>>> -Annette
>>>
>>> On Aug 17, 2021, at 2:44 PM, Bob Wyman <bob@wyman.us> wrote:
>>>
>>> On Tue, Aug 17, 2021 at 4:37 PM Annette Greiner <amgreiner@lbl.gov>
>>> wrote:
>>>
>>>> I don’t think this is a wise approach at all.
>>>>
>>> Can you propose an alternative that does not simply formalize the status
>>> of existing elites and thus strengthen hierarchies in public discourse? For
>>> instance, the existing Credibility Signals
>>> <https://credweb.org/reviewed-signals/> (date-first-archived,
>>> awards-won, ..) would seem to provide useful information about only a tiny
>>> portion of the many speakers on the Web. By focusing on the output of
>>> awards-granting organizations, while not providing signals usable by
>>> others, they empower that one group of speakers (those who grant awards)
>>> over the rest of us. Can you propose a mechanism that allows my voice, or
>>> yours, to have some influence in establishing credibility?
>>>
>>> We are seeing now that fraudsters and misinformation dealers are able to
>>>> gain traction because there is so little barrier to their reaching high
>>>> numbers of readers.
>>>>
>>> Today, the "bad" folk are able to speak without fear of rebuttal.
>>> Neither the fact-checking organizations nor the platforms for speech seem
>>> to have either the resources needed, or the motivation required, to
>>> usefully remark on the credibility of more than an infinitesimal portion of
>>> public speech. How can we possibly counterbalance the bad-speakers without
>>> enabling others to rebut their statements?
>>>
>>> In any case, the methods I sketched concerning Alice's statements would
>>> empower formal fact checkers as well as individuals, For instance, a
>>> "climate fact-checking" organization would be able to do a Google search
>>> for "hydrogen 'only water-vapor
>>> <https://www.google.com/search?q=hydrogen+%22only+water-vapor%22>',"
>>> and then, after minimal checking, annotate each of the hundreds of such
>>> statements with a common, well formed rebuttal that would be easily
>>> accessed by readers. Organizations could also set up prospective searches,
>>> such as a Google Alert, that would notify them of new instances of false
>>> claims and enable rapid response to their proliferation. I think this would
>>> be useful. Do you disagree?
>>>
>>> Any real solution must not make it just as easy to spread misinformation
>>>> as good information.
>>>>
>>> I have rarely seen a method for preventing bad things that doesn't also
>>> prevent some good. The reality is that the most useful response to bad
>>> speech is more speech. Given more speech, we can discover methods to assist
>>> in the process of separating the good from the bad. But, if we don't
>>> provide the means to make alternative claims, there is little we can do
>>> with the resulting silence. False claims will stand if not rebutted.
>>>
>>> It must yield a signal with much much less noise than the currently
>>>> available signals.
>>>>
>>> What "currently available signals?" Other than platform provided
>>> moderation and censorship, what is there?
>>>
>>> Increasing the level of he-said/she-said doesn’t help determine what is
>>>> reliable information. Adding to the massive amounts of junk is not the
>>>> answer.
>>>> -Annette
>>>>
>>>> On Aug 16, 2021, at 11:52 AM, Bob Wyman <bob@wyman.us> wrote:
>>>>
>>>> The thrust of my post is that we should dramatically enlarge the
>>>> universe of those who make such claims to include all users of the
>>>> Internet. The result of enabling every user of the Web to produce and
>>>> discover credibility signals will be massive amounts of junk, but also a
>>>> great many signals that you'll be able to use to filter, analyze, and
>>>> reason about claims and the subjects of claims.
>>>>
>>>>
>>>
>>>
>>
>> --
>> connie moon sehat
>> connieimdialog@gmail.com
>> https://linkedin.com/in/connieatwork
>> PGP Key ID: 0x95DFB60E
>>
>>
>
> --
> J. Gregory McVerry, PhD
> Assistant Professor
> Southern Connecticut State University
> twitter: jgmac1106
>
>
>
>

-- 
J. Gregory McVerry, PhD
Assistant Professor
Southern Connecticut State University
twitter: jgmac1106