Implementation and information value

The experimental implementation_information_value evaluator represents current, specific and post-sample implementation as conditional distributions over realised actions. Uptake can vary with the uncertain state, intended policy and observed sample signal; the calculation does not assume that information and implementation are independent.

import json
from pathlib import Path

from voiage import implementation_information_value

request = json.loads(
    Path("specs/frontier/implementation-information/v1/fixtures/normative/input.json")
    .read_text()
)
result = implementation_information_value(request).to_contract_dict()
print(result["gross_components"])

The result contains the full current/perfect-information by current/perfect-implementation matrix, optional specific-implementation and sample-information cells, gross and net components, complete policy ties, decision switches and two exact decomposition residuals. The interaction term is retained rather than allocated silently to information or implementation.

EVEIm and EVSEIm are reported only as terminology candidates. The contract is fixture-backed and Python-only; it is not evidence of stable maturity, cross-language parity, causal uptake identification or release readiness.