What if we instead could provide the tools to establish a soundness graph. Evaluate arguments on their own merits rather than their source.
Even Arguments based on false premises can be sound. Just as arguments can be based on true premises and still be unsound. Helping people identify which is which should at least raise the quality of disagreements.
When its premises are false, an argument is always unsound. But it can still be valid. A sound argument is one that is both valid and has true premises.
About builiding a soundness/validity graph, I've dabbled with building a webapp for that (though the graph is more implied than visual). It's still very basic, but if someone has ideas where exactly it should go or how it could be engaging to a community of critical thinkers, please contact me.
My own thinking is that it has to be less of an app and more of a protocol/federated thing augmenting existing channels. Think something like a github bot doing automated reviews with an SO-like community of meta-data authors annotating news articles and such
As a shortcut, trust graph is actually pretty good. Consider this example monologue:
This guy says really interesting things about cars, but my good friend Sally the Car Mechanic says he's talking nonsense; she's an expert in the domain, so I'll approach the new guy with huge dose of scepticism.
Note how trust graph implicitly takes care of known unknowns and unknown unknowns - I know little about cars, Sally knows a lot, so she's able to evaluate the situation better than me. Note also how it handles intent - Sally is my good friend, I trust her to have my best interest in mind, so I know what she's saying is her real opinion, and not e.g. trying to keep me as a customer of her workshop.
Intent is hard to judge, but is unfortunately very important when dealing with information that's not directly and independently testable (which is most of them, including especially conclusions drawn from testable facts). Trust graphs, or Evidence-based Ad Hominems™, are a very powerful shortcut for evaluating information.
Something like code reviews / peer review, but more broadly applied
People only trust sources with which they agree with at first place. Communities believing in a certain narrative will not re-evaluate anything, and all communities believe in specific narratives. The web just made that fact more obvious. No amount of technology of fact checking tool will change that.
Only the people who understand that the truth is often a valuable asset will find these tools valuable as well.
People trust sources which confirm their preexisting biases.
If someone says something that contradicts the information you already have, you should only change your belief proportional to how often they are right and you are wrong. And the only way to estimate the degree of relative correctness is to observe how often they agree with you ...
Now I'm just thinking aloud (awrite?), but maybe a general strategy to convince someone would be to find an authority that they trust more than themselves, or to become such an authority yourself. And the last part is hard, since you can only disagree with them if they will easily change their mind, thus changing their estimate of relative correctness in your favor.