This is a special post for quick takes by James Miller. Only they can create top-level comments. Comments here also appear on the Quick Takes page and All Posts page.
Reputation stops being verifiable beyond 2 degrees of separation.
At 1 degree you observed their behavior directly.
At 2 degrees someone you trust observed it.
At 3+ degrees it's pure performance, reviews can be bought, testimonials cherry-picked, social proof manufactured.
Humans broke away from the constraints of Dunbar's limit and while communities stayed small enough, reputation tracking sufficed, but the introduction of the internet and global connectivity has expanded our effective 'communities' to effectivly billions.
This is why every digital reputation system (LinkedIn endorsements, Trust Pilot scores, follower counts) fails at scale: they collapse verification radius to zero.
We're trying to run reputation-based trust in networks where nobody can actually verify anyone's claims. We rationally offload agency and accountability (for verification of trust) onto institutions who are themselves, perpetrators and participants in this incentive driven, optics obsessed disfunction.
I’ve built a diagnostic engine that assess governance and incentive documentation to surface the structural conditions that 'allow' for the rational behaviour of actors to occur, resulting in externalised harm.
The core of the engine is an Incentive First Framework, that strips narrative, intent, optics, culture and general governance theatre from the equasion and just simple asks:
"Given the structure, what behaviour is rational?"
I have ran the engine through a number of high profile back tests e.g. Boeing 737 MAX, Grenfell Tower and Post Office Horizon etc. and found that the qualifying questions the engine produces would reliably had surfaced the opacity or lack of enforcement/feedback integrity that could have corrected the behaviour.
All analyses were run with names redacted during diagnosis, no hindsight, no narrative framing, just structure based on the 'at the time' documentation.
Two questions:
1. What other documented cases (perhaps with inquiry) could be good for more back testing and that sufficient documentation evidence exists in the public domain?
2. If verified, could this have utility for private equity due diligence, insurance underwriting, and regulators who want early warning signals rather than post-hoc inquiries?