Epistemic Profile Layer #
@cite{floyd-2018} @cite{gawne-spronck-2024} @cite{tournadre-2008}
Connective layer bundling evidential source, epistemic authority (egophoricity),
and mirativity into a unified EpistemicProfile. Bridges these feature-geometric
dimensions to the model-theoretic level (KContext / ContextTower).
Motivation #
@cite{gawne-spronck-2024} identify 10 concept areas forming a coherent epistemic
domain. Linglib already covers evidential source (Core/Evidence.lean), epistemic
modality (Modality/Kernel.lean), and mirativity (Core/Evidence.lean), but these
are scattered with no connective tissue. The main gap is egophoricity -- the
dimension of WHO has privileged epistemic access (speaker vs addressee vs third
party), which the glossary argues is independent of evidential source.
Design #
EpistemicAuthority fills the egophoricity gap. EpistemicProfile bundles it
with EvidentialSource and MirativityValue for unified epistemic specification.
epistemicAuthority bridges epistemic authority to ContextTower by resolving
from the origin (speech-act context), since egophoric marking reflects the
actual speech-act participants, not reported ones.
Not Yet Formalized #
Engagement, logophoricity, and epistemic vigilance would require substantial new infrastructure (discourse model, binding theory extension, RSA listener model extension respectively) and are left for future work.
Epistemic authority: WHO has privileged access to the propositional content. Egophoric systems (@cite{tournadre-2008}, @cite{floyd-2018}, Gawne & Spronck glossary 2) grammaticalize the distinction between:
- ego: speaker has privileged access (1st person knowledge, volition, intention)
- allocutive: addressee has privileged access (2nd person questions)
- nonparticipant: neither speech-act participant has special access
- ego : EpistemicAuthority
- allocutive : EpistemicAuthority
- nonparticipant : EpistemicAuthority
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- Core.Epistemicity.instBEqEpistemicAuthority.beq x✝ y✝ = (x✝.ctorIdx == y✝.ctorIdx)
Instances For
Epistemic profile: bundles the three orthogonal epistemic dimensions that the glossary identifies as cross-linguistically relevant.
- source: WHAT kind of evidence (Aikhenvald's 3-way)
- authority: WHO has privileged access (egophoric dimension)
- mirativity: WHETHER the content is expected (DeLancey's dimension)
- source : Evidence.EvidentialSource
- mirativity : Evidence.MirativityValue
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
Instances For
Determine epistemic authority from a ContextTower, resolving from the origin (speech-act context). Egophoric marking reflects the actual speech-act participants: ego if the knower is the speaker, allocutive if the knower is the addressee, nonparticipant otherwise.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Ego authority with direct evidence: the canonical "strong assertion" profile. The speaker saw it themselves -- maximally authoritative.
Equations
- Core.Epistemicity.strongAssertion = { source := Core.Evidence.EvidentialSource.direct, authority := Core.Epistemicity.EpistemicAuthority.ego }
Instances For
Non-ego authority with inferential evidence: the canonical profile for epistemic 'must' (@cite{von-fintel-gillies-2010}). Speaker infers, no privileged access.
Equations
Instances For
In ego-authority contexts, direct evidence is the default source. This captures the generalization that 1st-person volitional claims ("I'm going to leave") use ego marking, not evidential marking.
In allocutive contexts, evidential source is typically irrelevant -- the addressee's authority overrides source distinctions. (Tibetan -payin ego vs -pa'dug non-ego; Akhvakh -eri ego vs -ari non-ego)
Equations
- Core.Epistemicity.allocutiveProfile s = { source := s, authority := Core.Epistemicity.EpistemicAuthority.allocutive }
Instances For
Epistemic authority is invariant under tower push: egophoric marking reflects speech-act participants (from origin), not embedded ones.