While I agree with several of your conclusions here, and you bring up very good points about there being no limiting factor on frivolous negative ratings, I think your solution is lacking and is only adding more complication to an already over complicated system. The system needs to be simplified, not added to. My suggestion is to have a standard of evidence of theft, violation of contractual agreement, or violation of applicable laws before leaving a negative rating. This solves the primary issue of there being no limiting factor to false or frivolous negative ratings with minimal effort, creates a simpler system, and less counter productive side effects.
The end user does not need to understand how the calculations are made. The end user only needs to know they should leave a negative rating if they strongly believe the person to be a "scammer" and a positive rating should be left if you have affirmative reason to believe the person to be trustworthy AND you should not leave a positive rating if there is a good reason to believe the person is a scammer (this is a lower threshold than to leave a negative rating, but will prevent a scammer from completing a small number of trades to erase the 'trade with extreme caution' warning. This particular proposal is designed to ensure there is something semi-resembling consensus for any labeling of someone as a scammer.
I don't think the requirement to present evidence of
xxx alone is sufficient to prevent frivolous negative ratings. I am following both your and OgNasty's separate disputes with Vod, following what is being posted, it is clear that Vod is in the wrong, yet he continues to have many trust inclusions despite his lies and deception -- I believe this is because he has sufficient allies that he supports and who in turn support him regardless of either being right or wrong, they all simply refuse to criticize eachother no matter how outrageous they behave. I believe if it is required to publicly post evidence of "xxx" to post a negative rating, unless the administration gets involved in trust disputes, which I know theymos does not want to in ~all cases, those who are currently giving frivolous ratings will continue doing so, and will present troll posts as evidence, and others will continue to support these people out of fear of getting tagged themselves, or getting excluded from the trust network.
I also believe that someone displaying (multiple) red flags of being a scammer should be grounds for being labeled as a scammer. A brand new user asking for a loan without collateral is fairly clearly intending on scamming (even though their scam attempt is so transparent that labeling these people a scammer is probably unnecessary). If someone were to explicitly say they intend on stealing money once given the opportunity, I don't believe they would meet your criteria for receiving negative trust, but I believe others should be warned about this person's intentions. marcotheminer is another good example of someone I believes should be tagged as when he recently returned from a long hiatus, he started taking out loans in escalating sizes, and eventually tried taking out a $xx,000 loan, which is constant with someone trying to engage in an exit scam (he was actually tagged by many before this, which was not appropriate). Someone running what is very clearly a ponzi should also be tagged.
Yes, someone displaying a "red flag" of being a scammer is very subjective. My proposal would address this problem by requiring there to be something resembling consensus for someone to be labeled a scammer. If I label someone as a scammer because they are displaying x red flag, and you disagree this red flag means this person will scam in the future, you can counter my negative rating, and others can chime in via their own ratings, and the system will only display a 'trade with extreme caution' warning if there is consensus, while leaning towards not displaying the warning, constant with the principal of the presumption of innocence.