Schneider and Beccuti Examine Cardano DRep Rewards and AI Voting

Manvir Schneider, Senior Research Scientist at the Cardano Foundation, and Juan Beccuti, a researcher in economics and blockchain governance, presented their research program during the first Research Exchange on October 19. Their discussion covered DRep compensation, the use of large language models in governance, and the economic incentives supporting Cardano’s network.

The series invites community feedback while research develops, continuing the Foundation’s collaboration with Input Output and other research institutions.

Rewarding Informed DRep Decisions

DReps receive delegated voting power from ADA holders and participate in Cardano governance. Designing compensation for that work requires a way to encourage representatives to assess proposals rather than simply register votes.

Schneider described the risk of rewarding participation without accounting for effort. A representative could vote randomly or copy another participant’s position while still meeting an activity requirement.

The researchers are studying mechanisms intended to support informed decision-making. Schneider explained that one model they are examining would require individual votes to remain hidden until voting closes.

That requirement would change when delegators and other observers could inspect representatives’ choices. It concerns temporary confidentiality during a voting period, rather than permanently concealing the outcome.

The researchers identified community acceptance of that arrangement as a question for further discussion. No activation of DRep payments or change to Cardano’s voting process was announced.

Shared AI Models Can Produce Correlated Judgments

Another research topic examines large language models as delegates or decision-support tools. Governance proposals can require substantial reading, making automated summaries and analysis a potential aid for representatives.

Schneider highlighted a risk when many participants use the same model with similar information and context. Their reasoning and decisions may become correlated, reducing the independence of their assessments.

That distinction matters when evaluating collective decisions. Several votes based on one analytical source do not necessarily provide the same diversity of judgment as independently formed opinions.

The session presented this as an open research question. It did not establish that shared AI models currently determine Cardano governance outcomes or quantify their influence on existing votes.

[NEWSBREAK]

Staking Parameters and Long-Term Incentives

Beccuti outlined two follow-up sessions examining how parameter changes affect pool operators and delegators.

The first would address parameters directly connected to reward distribution. Reducing a pool’s declared fixed cost, for example, can initially reduce the operator’s receipts while making the pool more attractive to delegators. The net effect depends on subsequent delegation decisions and the existing distribution of stake.

A second session would examine longer-term monetary parameters, including the use of reserves to fund rewards and the sustainability of those incentives over time.

The broader program also includes block-propagation behavior and incentives for validating and adopting Leios endorser blocks.

For the parameter sessions, the researchers plan to compare design objectives with current market conditions. Pool operators and delegators are invited to challenge assumptions using their actual costs, revenues, and responses to changes in rewards.