Concerns about the Fisherman incentivization model

[DragonDmoney]

Describe the bug
I'm not sure that the current Fisherman incentive model is suitable for trustless Fisherman (castnet)

Concerns

Contrary to ServiceNodes, Fishermen rewards are not affected by the content of the samples and TestScoreMsgs as long as they are not null.
The current fisherman model rewards all Fishermen the same amount regardless of the quality of sampling. Once we reach Cast-net and there are trustless Fishermen, what incentive is there for Fishermen to give accurate test scores? What is stopping a fisherman from giving terrible test-scores to nodes run by another actor, and giving perfect test scores to their own nodes? Typically in economic environments like this, if there is no incentive model to do a task well, then the task will not be done well. Imagine if Chainlink rewarded all nodes the same regardless of their accuracy, and its core, fisherman, and chainlink perform very similar tasks: reporting off-chain data on the chain.

Solution

These reasons lead me to propose another economic model, similar to the Chainlink model, where fishermen would be rewarded based on how similar their results are to the rest of the fishermen. Over time, these Fishermen would be given a ReputationScore, based on how accurate their data is, and their reward system and impact on ServiceNode's report cards would depend on their ReputationScore. I haven't fully thought out the exact specifications of this, but this is a significant change, and I'd like to hear your thoughts before I draft up a new specification.

Origin Document
https://docs.pokt.network/v1/utility

Creator: DragonDmoney (Pierre)

[Olshansk]

Just wanted to link to the V1 Contributor hour presentation from 06/23/2022 discussing this topic.

I think this proposal has the foundation of potentially very meaningful changes but requires more research & discussion so tagging it as an enhancement for now.

[DragonDmoney]

Note: this issue is continued by

#7