New RSC Reward Algorithm

[yattias]

• [ ] Reward algorithm should be versioned.
• [ ] Implement reward mapping associated with each action

Open Questions:

• Identify exact rewards associated with actions. e.g. 1 upvote = 1 RSC Citation + open access paper = 100 RSC ....

• Should every author be eligible for rewards or just specific authors in the MVP ?

• How do we handle the small subset of people that already claimed papers?

• Should every user be able to claim an author and receive RSC or just bioRxiv bioengineering subject area (initially)?

[yattias]

@TylerDiorio / @joycesticks - @koutst and myself discussed a few things:

1. We will not be creating escrows for each user automatically. This is due to various reasons:
   • There are lots of escrows to be created ~700K for bioRxiv bioengineering
   • If we make a mistake, recalculating is quite "expensive" time wise
   • Instead, we want to calculate an estimated value which the user will see. After claiming, they will receive the actual value

2. We would like to divide rewards among all authors evenly for the MVP. Reasons:
   • It is "easier" and feels like the right thing to do given we don't have all the data
   • More of an incentive for other authors to claim their profiles and join the platform

[joycesticks]

Hey Kobe,

1 makes perfect sense to me.

For #2, I personally feel strongly that we should give the full rewards to the 1st author of the paper. My thought process:

This MVP is to test creating a financial incentive for high-quality research behaviors. For this sprint, I'm not too worried about creating an incentive for first authors to refer their co-authors.

If we find that people are excited about the author rewards feature, and we are see that a bunch of 1'st authors are claiming their profiles, then we should set aside time to build a referral feature that rewards these first authors for referring their co-authors (to claim their own individual first author papers).

[yattias]

@TylerDiorio

Need your help with the folloowing

Identify exact rewards associated with actions.

• 1 upvote = 1 RSC
• Citation + open access paper = 100 RSC

[koutst]

@TylerDiorio I was looking at the excel file provided regarding rewards. It seems like Zipf's law is calculated on the author level. Would it be possible to calculate it on the paper level? This would make it possible for us to provide the amount that was paid out for a specific paper so that the first author could potentially distribute rewards to the other authors.

For the additional weekly(?) drip, will the Zipf's law look at total new citations since the previous week?

[koutst]

The Zipf's law calculations are actually on the paper level.

[yattias]

@TylerDiorio Can you share the multipliers and how they translate into rewards in here once you completed analysis?

[TylerDiorio]

For the drip rewards, we'll need to consider strict adherence to the vesting schedule. Here's my initial logic (will expand later) using 'days' as the time frame to check for new citations and estimate emissions via upvotes; we could certainly use a few hour, few day, or week range instead.

Brief:

1. Count all the upvotes per day
2. Subtract upvote RSC from Total RSC available for rewards
3. Count all the new citations per day
4. Weight the citations 2x if pre-registered, 3x if open data
5. Sum the weighted citations (citation-points) per day
6. Divvy remaining Total RSC over weighted citations

Poorly formatted equations:

Total RSC Rewards (from vesting): X RSC/day RSC per upvote: 1 RSC per upvote Total RSC Rewards after upvotes = (X RSC/day) - (1 RSC/upvote) (Y upvotes/day) RSC per citation: [(X RSC/day) - (1 RSC/upvote) (Y upvotes/day)] (Z citation-points/day) citation-points: Z citation-points = (1 citation-point/citation) (2x if pre-registered) * (3x if open data)

Limitations:

• This works for low amounts of RSC given to upvotes relative to the amount available, but the RSC per upvote should eventually also follow strict vesting rather than flat rates of RSC/upvote