Documentation

Linglib.Phenomena.Reference.Studies.CohnGordonEtAl2019

@cite{cohn-gordon-goodman-potts-2019} — Incremental Iterated Response Model #

@cite{dale-reiter-1995}

Cohn-Gordon, R., Goodman, N. D., & Potts, C. (2019). An Incremental Iterated Response Model of Pragmatics. Proceedings of the Society for Computation in Linguistics (SCiL) 2, 81–90.

The Model #

The incremental RSA model extends the standard RSA framework to word-by-word production. The speaker produces referring expressions incrementally, choosing each word to maximize the listener's posterior probability for the target:

S1^WORD(wₖ | [w₁,...,wₖ₋₁], r) ∝ L0(r | w₁,...,wₖ)^α

The full utterance probability factors via the chain rule:

S1^UTT-IP(w₁,...,wₙ | r) = ∏ₖ S1^WORD(wₖ | [w₁,...,wₖ₋₁], r)

L0 uses extension-based incremental semantics (§2.2): given prefix c,

⟦c⟧(w) = |{u ∈ U : c ⊑ u ∧ ⟦u⟧(w) = 1}| / |{u ∈ U : c ⊑ u ∧ ∃w'. ⟦u⟧(w') = 1}|

where U is the set of complete utterances and ⊑ is the prefix relation.

Formalization #

This is the first formalization to use RSAConfig's sequential infrastructure:

The domain is Figure 1 from the paper: 3 referents (red dress, blue dress, red hat), 3 words (red, dress, object), 3 complete utterances (dress, red dress, red object). Costs are 0 for all words.

Findings #

#FindingStatus
1Adjective-first preferred for target R1 (Figure 1c)rsa_predict
2Noun preferred after adjective for R1 (Figure 1c)rsa_predict
3R2 must start with "dress" (Figure 1c)rsa_predict
4R3 must start with "red" (Figure 1c)rsa_predict
5Uniform fallback after "red" for R2 (§2.2)cases w <;> rsa_predict
6L1 anticipatory implicature: "red" → R3 (Figure 1d)rsa_predict
7Incremental model prefers bare noun over modified NP (Figure 1e)rsa_predict

Words available to the incremental speaker (Figure 1a).

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

      Referents in the reference game scene (Figure 1a).

      Scene: {red dress (R1), blue dress (R2), red hat (R3)}

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

          The three complete utterances in the scene (Figure 1a): "dress", "red dress", "red object".

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

            Utterance-level Boolean semantics: conjunction of word applicability.

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

              Count of complete utterance extensions of pfx that are true of r.

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

                Count of viable extensions: complete utterances extending pfx that are true of at least one referent.

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

                  Extension-based incremental semantics (§2.2):

                  ⟦pfx⟧(r) = trueExtCount(pfx, r) / viableExtCount(pfx)

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

                    Incremental RSA for the Figure 1 reference game.

                    This is the first RSAConfig to use the sequential infrastructure (Ctx, transition, initial). The model produces referring expressions word-by-word, with each step choosing the next word to maximize L0's posterior for the target referent.

                    Architecture:

                    • L0_at(ctx): literal listener given prefix ctx + next word
                    • S1_at(ctx): speaker choosing next word given prefix ctx
                    • trajectoryProb: chain-rule product of S1_at probabilities

                    Parameters: α = 1, cost = 0 for all words, uniform priors.

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

                      Qualitative findings from the incremental RSA model.

                      • adj_first_for_target : Finding

                        The incremental speaker prefers the adjective "red" first when referring to the target R1 (red dress).

                      • noun_after_adj : Finding

                        After producing "red", the speaker prefers the type noun "dress" over the generic "object".

                      • noun_only_for_r2 : Finding

                        For R2 (blue dress), the speaker must start with "dress" — "red" has zero incremental semantics for R2.

                      • adj_only_for_r3 : Finding

                        For R3 (red hat), the speaker must start with "red" — "dress" has zero incremental semantics for R3.

                      • uniform_after_red_for_r2 : Finding

                        After "red" for R2, no extension is true — the speaker is indifferent between "dress" and "object" (uniform fallback).

                      • listener_anticipation : Finding

                        After hearing "red", L1 infers the target is more likely R3 (red hat) than R1 (red dress) — an anticipatory implicature.

                      • incremental_prefers_bare_noun : Finding

                        At the utterance level, the incremental model assigns higher probability to "dress" than to "red dress" for R1 — diverging from the global model which prefers "red dress".

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

                          Finding 1 (Figure 1c): The incremental speaker prefers "red" first when referring to R1 (red dress).

                          S1(red | [], R1) = 4/7 ≈ 0.57 > S1(dress | [], R1) = 3/7 ≈ 0.43

                          Mechanism: "red" narrows the extension set to {red dress, red object}, both true of R1 (trueExtCount = 2, viableExtCount = 2 → meaning = 1). "dress" narrows to {dress}, true of R1 (meaning = 1) but the L0 posterior for R1 is lower because "dress" also applies to R2.

                          Finding 2 (Figure 1c): After producing "red", the speaker prefers "dress" over "object" for R1.

                          S1(dress | [red], R1) = 2/3 ≈ 0.67 > S1(object | [red], R1) = 1/3 ≈ 0.33

                          "red dress" uniquely identifies R1 (only R1 is a red dress), while "red object" is ambiguous between R1 and R3.

                          Finding 3 (Figure 1c): For R2 (blue dress), the speaker must start with "dress" — "red" never applies to R2 (it's a blue dress), so all extensions of "red" have zero semantics for R2.

                          S1(dress | [], R2) > S1(red | [], R2)

                          Finding 4 (Figure 1c): For R3 (red hat), the speaker must start with "red" — "dress" never applies to R3 (it's a hat), so the only extension of "dress" (= "dress" itself) has zero semantics for R3.

                          S1(red | [], R3) > S1(dress | [], R3)

                          Finding 5 (§2.2, uniform fallback): After "red" for R2, no complete utterance extension of "red" is true of R2 (blue dress). The paper states: "probability is evenly distributed over all choices of word."

                          S1(dress | [red], R2) = S1(object | [red], R2)

                          Both equal 1/2 because the meaning function returns 0 for all R2 extensions of "red", yielding uniform L0 → uniform S1.

                          Finding 6 (Figure 1d): After hearing "red", the pragmatic listener L1 infers that the target is more likely R3 (red hat) than R1 (red dress).

                          L1(R3 | red) = 7/11 ≈ 0.64 > L1(R1 | red) = 4/11 ≈ 0.36

                          This is an anticipatory implicature: "red" is the ONLY word available for R3 (S1(red|[],R3) = 1), so hearing "red" raises R3's probability. For R1, the speaker could have said "dress" instead, so "red" is less diagnostic. This foreshadows @cite{sedivy-etal-1999}'s finding that listeners draw contrastive inferences from prenominal adjectives.

                          Finding 7 (Figure 1e): The incremental model prefers "dress" over "red dress" for R1 — the key divergence from the global RSA model.

                          S1^UTT-IP(dress | R1) = 3/7 ≈ 0.43 > S1^UTT-IP(red dress | R1) = 8/21 ≈ 0.38

                          The global model prefers "red dress" (more informative). The incremental model prefers "dress" because it is produced in one step with probability 3/7, while "red dress" requires two steps: S1(red|[],R1) · S1(dress|[red],R1) = 4/7 · 2/3 = 8/21 < 9/21 = 3/7.

                          Map each finding to the model prediction that accounts for it.

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

                            The incremental RSA model and @cite{dale-reiter-1995}'s Incremental Algorithm (IA) solve the same problem — generating referring expressions for a target among distractors — via structurally parallel mechanisms:

                            PropertyD&R IAIncremental RSA
                            ProcessingSequential (attribute-by-attr)Sequential (word-by-word)
                            SelectionDeterministic (fixed order)Probabilistic (soft-max)
                            Q2/CostNone (No Brevity)None (s1Score = L0)
                            StateRemaining distractorsCtx = word prefix
                            TerminationAll distractors ruled outChain rule product over words

                            Both operate in the No-Brevity regime: D&R's IA includes any discriminating attribute without brevity optimization; the incremental RSA's s1Score l0 _ _ w u := l0 u w has no cost term. D&R's fixed PreferredAttributes order is generalized by RSA's probabilistic ranking, which emerges from the L0 semantics at each step.

                            The key difference: D&R is deterministic and may produce non-minimal descriptions (as shown in DaleReiter1995.cups_non_minimal), while the incremental RSA is probabilistic and assigns lower total probability to longer utterances via the chain rule product (Finding 3: incremental_prefers_bare_noun).

                            Both the incremental RSA and @cite{dale-reiter-1995}'s Incremental Algorithm operate in the No-Brevity regime (strength = 0) — the weakest Q2 interpretation. Both enforce Q1 (each word/attribute must contribute to identifying the referent) without Q2 (brevity) pressure. D&R's deterministic fixed-order selection is generalized by the incremental RSA's probabilistic word-by-word production.