Abstract
We describe a split incentive payment for prosocial acts in a peer-to-peer tipping system, in which each rewarded act receives two equal components from two different authorities. The first component is set by an algorithm that scores the act against a group-level rubric the family's own stewards author and revise (the acts the family says it values more of); in the deployed web application this is a deterministic scoring algorithm, with an on-device AI evaluator planned for native applications. The second component is set later by a group member, who chooses how much to tip onward from a restricted pass-forward balance that can be given to others but not spent. Equal weighting steers behaviour toward acts that are both collectively and individually valued. The components are designed to correct each other: the peer component acts as human-in-the-loop fraud validation of the gameable algorithmic component, since a person who knows the actor is unlikely to tip a fabricated act, and the algorithmic component acts as an impartial floor under the favouritism-prone peer component. Because earning both requires modelling what others value, it also creates an incentive for mutual attention. A nested variant with an autonomous AI agent as the upper-level planner is also disclosed. Stated limits: collusion between members defeats the peer check; the combination may inherit both failure modes rather than cancel them; reward salience may crowd out intrinsic motivation; the evidence is one pilot family over one month, and the fraud-validation and handoff predictions are untested.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Ly, Thon, "The Two-Layer Reward: Pairing an Algorithmically Set Reward with a Peer-Set Tip as Mutual Checks Against Gaming and Favoritism in a Prosocial Incentive System", Technical Disclosure Commons, (September 29, 2026)
https://www.tdcommons.org/dpubs_series/11887