Discourse Commitments #
@cite{brandom-1994} @cite{gunlogson-2001} @cite{krifka-2015} @cite{bring-gunlogson-2000} @cite{romero-2024}
Shared types for modeling public commitments in discourse, used by multiple theories of assertion (Krifka, Brandom, Gunlogson).
A CommitmentSlate is an agent's public discourse commitments — the
propositions they have publicly committed to (which may differ from
their private beliefs). This separation is crucial for:
- @cite{krifka-2015}: commitment ≠ belief; lying = commitment without belief
- @cite{brandom-1994}: commitments are normative statuses tracked by scorekeepers
- @cite{gunlogson-2001}: source-marking distinguishes self-generated from other-generated commitments
An agent's public discourse commitments: a list of propositions the agent has publicly committed to.
Following @cite{krifka-2015}: the commitment slate tracks what an agent is publicly committed to, which may diverge from what they privately believe (as in lying, hedging, or performing).
The propositions the agent is publicly committed to
Instances For
The empty commitment slate: no public commitments.
Instances For
Add a commitment to the slate.
Instances For
Retract a commitment (remove first occurrence).
Not all theories support retraction. Stalnaker's CG model has no retraction mechanism; Krifka and Brandom do.
Instances For
Convert commitments to a context set: the worlds compatible with all committed propositions.
Equations
- s.toContextSet w = ((s.commitments.all fun (p : BProp W) => p w) = true)
Instances For
Check if the slate entails a proposition (holds at all compatible worlds).
Equations
Instances For
Empty slate is trivial: all worlds are compatible.
Adding a commitment restricts the context set.
Adding a commitment entails the added proposition.
The source of a discourse commitment.
@cite{gunlogson-2001}: commitments are marked by their epistemic source.
.selfGenerated: the agent generated the commitment from their own evidence.otherGenerated: the commitment originates from another participant
The source determines challengeability: self-generated commitments can be challenged by the addressee; other-generated commitments can be challenged by the speaker.
- selfGenerated : CommitmentSource
Commitment generated from agent's own evidence/beliefs
- otherGenerated : CommitmentSource
Commitment attributed to another participant
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- Core.Discourse.Commitment.instBEqCommitmentSource.beq x✝ y✝ = (x✝.ctorIdx == y✝.ctorIdx)
Instances For
A commitment tagged with its source.
- content : BProp W
The propositional content
- source : CommitmentSource
How the commitment was generated
Instances For
A source-tagged commitment slate.
- commitments : List (TaggedCommitment W)
The tagged commitments
Instances For
The empty tagged slate.
Instances For
Add a tagged commitment.
Instances For
Get all self-generated commitments.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Get all other-generated commitments.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Convert to a plain (untagged) commitment slate.
Equations
- s.toSlate = { commitments := List.map (fun (x : Core.Discourse.Commitment.TaggedCommitment W) => x.content) s.commitments }
Instances For
Convert to context set (ignoring source tags).
Equations
Instances For
Contextual evidence bias.
Expectation about p induced by evidence available in the current discourse situation. Used as:
- A felicity condition on rising declaratives
- A bias dimension for polar questions
This type is shared between assertion theory and question bias theory because the same notion of contextual evidence governs both: evidence for p licenses rising declaratives about p and positive polar questions about p.
- forP : ContextualEvidence
Current context provides evidence for p.
- neutral : ContextualEvidence
No contextual evidence either way.
- againstP : ContextualEvidence
Current context provides evidence against p.
Instances For
Equations
- Core.Discourse.Commitment.instBEqContextualEvidence.beq x✝ y✝ = (x✝.ctorIdx == y✝.ctorIdx)
Instances For
Equations
- One or more equations did not get rendered due to their size.