Signed and social information value

VOIAGE’s experimental signed-social evaluator keeps harmful information, distributional transfers and selective sharing visible. It never clips a negative private value to zero.

The input freezes a complete finite joint-world probability law, named agents and roles, signal topology, sharing designs, nonanticipative policies, costs, transfers, rights receipts and a declared welfare aggregator. Each design is compared with its named comparator. Centralized designs optimize only across their finite catalog; fixed, declared-response and verified-equilibrium designs retain the supplied policy. A verified equilibrium is a checked finite catalog, not a general game solver.

voiage --format json calculate-signed-social-information \
  signed-social-information.json --output result.json

The governed fixture returns a recipient value of -2, a stakeholder value of 4, and a declared social value of 2 cardinal-welfare units. The result also contains the pre-transfer, transfer, cost and post-transfer ledgers, complete ties, policy switches, winners, losers, avoidance diagnostics, externalities and Blackwell-check applicability.

Blackwell-style nonnegativity is checked only for an aligned centralized comparison with the same selector and a verified refinement whose baseline catalog is embedded. Costs, conflicting objectives, fixed responses, rights constraints or a missing refinement assurance can make that theorem inapplicable; the result lists the reasons.

The example is informed by Li and Pozzi’s negative-information construction and bounded team-decision literature such as Cao’s collaborative learning and decision-making framework. It is synthetic contract evidence, not an empirical reproduction. Bayesian persuasion, mechanism design, rational inattention and general game solving remain adjacent.