Future Papers

Already Absorbed by the Current Spine

Several earlier future-paper ideas have now become part of the published or drafted spine. The two-gap paper absorbs the general E0-EP and EP-CA distinction. Warranted Uncertainty absorbs the discipline of not knowing and the first treatment of diagnostic slack. The long-form Books shelf now absorbs the broad methodological argument for predictive power, Bayesian bookkeeping, induction, and revisable ontology. The rationalization-engine paper absorbs the biased-expert case. Institutional Credencing absorbs the first organizational extension. The AI prosthetic paper absorbs the first AI-mediated inquiry treatment.

Future work should therefore avoid restating those claims at the same level of generality. A new paper should either formalize a narrower mechanism, apply the model to a new domain, build a usable practice, or test a hard objection that the current papers leave unresolved.

• |Do not rewrite the two-gap model; specify a new gap, threshold, or repair method.
• |Do not merely repeat warranted uncertainty; show how uncertainty should be represented, taught, audited, or bounded.
• |Do not merely repeat AI as prosthetic; isolate a concrete AI-mediated failure mode or alignment practice.
• |Do not merely repeat institutional credencing; design a procedure, audit artifact, or case taxonomy.

Independent Paper Candidates

These papers are close enough to strengthen Credencing but narrow enough to stand on their own. Each should have a distinct question, diagnostic target, and practical output.

• |Excess Core Irrationality: distinguish raw EP-CA mismatch from excess mismatch after warranted slack, pressure, access limits, and skill limits have been accounted for.
• |The Epistemology of Belief Integrity: give a philosophical account of what it means to follow what one takes the evidence to show.
• |Bayesian Updating Without False Precision: translate Bayesian structure into rough priors, likelihood comparisons, and confidence bands without fake numerical sharpness.
• |Priors as Disciplined Memory: defend priors as revisable background knowledge rather than mere bias or inherited prejudice.
• |Likelihood Literacy: focus entirely on the neglected question of how expected the evidence is under rival hypotheses.
• |Base Rates, Identity, and Statistical Resistance: examine why base rates are rejected when they threaten hope, fear, identity, or group narratives.
• |The Evidential History of a Posterior: treat final credence as answerable to the update path that produced it, not as a free-floating confidence state.
• |Epistemic Akrasia and the Fear of Updating: analyze cases where an agent senses that confidence should move but cannot bring assigned credence along.
• |Identity-Protective Credencing: isolate the mechanisms by which belonging, self-concept, and public commitment distort CA without necessarily damaging EP.
• |The Phenomenology of Certainty: explain why certainty can feel like evidence even when it is only closure, fluency, or social reinforcement.
• |Credence Grammar: build a public language for confidence bands, update conditions, capped confidence, and live alternatives.
• |Assertion, Confidence, and Social Risk: study why public speech rewards overstatement and makes calibrated uncertainty socially costly.
• |Disagreement Without Diagnosis: develop norms for disagreement that avoid premature classification of other agents as irrational.
• |Category Confidence: ask how much confidence we should assign to the categories used to frame claims, not only to claims inside those categories.

Institutional and AI Project Track

The institutional and AI papers now exist as first statements, so the next work should be more procedural. These projects should produce checklists, audit trails, templates, case taxonomies, or decision protocols.

• |Suppressed Evidence and Official Confidence: classify ways internal evidence can fail to become public confidence, from ignored signals to strategic delay.
• |Audit Trails for Belief: design a record format for predictions, confidence levels, evidence sources, updates, and reasons for revision.
• |Epistemic Governance: specify institutional designs that notice reality earlier, protect dissent, lower update costs, and preserve uncertainty records.
• |The Incentive Structure of Denial: map how organizations reward false confidence and make accurate updating expensive.
• |Delegated Credence Protocols: design handoff rules for when AI may gather evidence, summarize evidence, recommend confidence bands, or trigger human review.
• |Model Fluency and False Credence: explain why polished AI outputs create overconfidence when source quality and uncertainty remain unclear.
• |Human-AI Epistemic Alignment: define when users should trust, doubt, defer to, contest, or override model outputs.

Teaching and Tool Projects

Some future work should become classroom practice or site tooling rather than another long paper. These projects would make Credencing easier to teach, test, and reuse.

• |Teaching Credences Before Conclusions: make confidence, uncertainty, and update conditions part of answer formation before students are rewarded for correctness.
• |Calibration Exercises for Classrooms: build forecast journals, confidence scoring tasks, likelihood-comparison drills, and post-update reflections.
• |Discussion Groups as Epistemic Training: treat group discussion as practice in humility, updateability, and non-defensive disagreement.
• |The Honest Novice as an Educational Ideal: defend low-tool, high-updateability learners as a genuine achievement, not a deficit state.
• |Scenario Builder: let users create, save, compare, and export E0-EP-CA-SD cases from the interactive lab.
• |Credence Audit Worksheet: turn the model into a printable and editable sequence of evidence, perception, confidence, uncertainty, and update-condition prompts.
• |Likelihood Literacy Module: add a guided practice path focused on comparing evidence under rival hypotheses.

Selection Rule for New Work

A future paper or project should be accepted into the roadmap only if it adds a distinct unit of work. It should answer a question the current site does not already answer, introduce a new diagnostic distinction, produce a usable tool, or apply the framework to a domain where the variables behave differently.

The test is simple: if the proposed work can be summarized as “the two-gap model, but again,” it should be folded into an existing page. If it creates a new practice, measurement problem, language protocol, institutional artifact, or AI-use rule, it belongs here.