Documentation

Linglib.Theories.Pragmatics.RSA.Questions.Basic

RSA Questions: Shared Infrastructure (Stub) #

@cite{groenendijk-stokhof-1984} @cite{hawkins-etal-2025} @cite{van-rooy-2003}

Shared infrastructure for RSA models of question-answer pragmatics.

Status #

The ℚ-based RSA evaluation infrastructure (RSA.Eval, RSA.Eval.sumScores, RSA.Eval.normalize, RSA.Eval.getScore) has been removed. This file retains type definitions and structural properties that do not depend on the old API. RSA question-answer computations need to be re-implemented using the new RSAConfig framework.

def RSA.Questions.expectedValue {α : Type} (f : α) (dist : List (α × )) :

Expected value of a distribution

Equations
Instances For

    RSA question-answer model parameters.

    • αQ :

      Questioner rationality

    • αR :

      Respondent rationality

    • β :

      Action-relevance weight: 0 = informativity, 1 = action-relevance

    • costWeight :

      Response cost weight

    Instances For
      Equations
      • One or more equations did not get rendered due to their size.
      Instances For
        Equations
        Instances For

          Default parameters

          Equations
          Instances For

            Generic response type: taciturn, elaborated, or exhaustive.

            Instances For
              Equations
              • One or more equations did not get rendered due to their size.
              Instances For
                def RSA.Questions.informationValue (valueBefore valueAfter : ) :

                Van Rooy's UV(C) = V(D|C) - V(D).

                Equations
                Instances For
                  def RSA.Questions.basicInformativity (responseSetSize : ) :

                  Informativity as inverse of response set size.

                  Equations
                  Instances For
                    theorem RSA.Questions.beta_tradeoff (params : Params) :
                    params.β = 01 - params.β = 1 params.β = 0

                    beta = 0 yields pure informativity.

                    RSA iteration level.

                    Track the depth of pragmatic reasoning:

                    • L0 = literal listener
                    • S1 = pragmatic speaker (responds to L0)
                    • L1 = pragmatic listener (responds to S1)
                    Instances For
                      Equations
                      • One or more equations did not get rendered due to their size.
                      Instances For