Warrant

feature · paper proposed

Declaration 01abf03d6f99 has not been accepted by anyone. · no paper read under it

How well does the tree's argument support each claim?

A capability that revises existing artifacts. It runs once, and then it is part of the schema.

Part of Tree — What does the paper assert, and how do its claims depend on each other?

zmainen/claim-graphs#21 — the decision is taken there; the propagation is taken here.

How it works

Assembles, for each claim, a dossier of what the tree's argument holds about its support — its role and stance, the outcomes on its predictions, the controls that validate it, the rivals it rules out, what it requires and is part of (with their roles), and the claims that support, extend, confirm or refute it. It shows a reader every claim with its sentence and that dossier, and asks for a warrant level from the vocabulary the claim's role allows — a prediction's confirmed/refuted/untested, an alternative's ruled-out/open, everything else's strong/moderate/weak/contested — a one-line why that cites the dossier, and an unsupported flag where the tree's argument records nothing.

The basis is the argument the tree records, not the paper's confidence and not any reproduction, which a later checking layer assesses and may then modify. The prompt shows the dossier and the vocabulary, never the mechanical rule: scripts/warrant.py carries a rule floor the reading is scored against afterwards, and the ruling puts that floor after the judgement, not before.

--dump-prompt writes the exact request and exits, and --answer feeds a reply back through the same validation — a slug not in the tree, or a level outside the role's vocabulary, is dropped — so a model with no backend here can still answer it.

Rests on

claim-tree · step
stance · feature
prediction-outcome · question

Feeds

Nothing — this is a leaf.

How it is defined

A model answers this layer, so the prompt is the layer. It is reproduced below from the committed file, and it is a declared input — editing it makes every run that used it stale.

extract/prompts/warrant.mdthe prompt it runs undernot in the repository

The declaration names this path and the repository does not have it. An input that does not exist hashes to nothing, so it cannot make a run stale — the layer is declared to depend on something it is not in fact tracking.

extract/prompts/contract/vocabulary.mdthe prompt it runs undernot in the repository

scripts/warrant.pythe script that runs itnot in the repository

What it produces68 claims

One paper, as the worked example — Gadeke, v3. Read from runs/gadeke-2026-guilt-insula/warrant.json · 22 KB. model supplied:runs/gadeke-2026-guilt-insula/warrant.answer.v3.json

#	slug	role	stance	warrant	why	unsupported	warrant_from
1	anterior-insula-tracks-guilt-insula	prediction	asserts	confirmed	confirmed by insula-rois-responded-more-low and mass-univariate-voxel-wise-analysis-found-small	false	`["outcome","confirmed_by"]`
2	connectivity-between-guilt-responsibility-related-outcome-ph	prediction	asserts	confirmed	confirmed by functional-connectivity-between-left-anterior	false	`["outcome","confirmed_by"]`
3	neural-substrate-tracks-participant-responsibility-2	prediction	asserts	confirmed	confirmed by one-cluster-left-sts-responded	false	`["outcome","confirmed_by"]`
4	responsibility-outcomes-generates-guilt-participant	prediction	asserts	confirmed	confirmed by when-partner-received-low-lottery	false	`["outcome","confirmed_by"]`
5	responsibility-partner-outcomes-influences-participant	prediction	asserts	confirmed	confirmed by likelihood-ratio-test-showed-responsibility and partner-reward-prediction-errors-resulting	false	`["outcome","confirmed_by"]`
6	alt-agency-aversion-not-guilt	hypothesis	rejects	ruled-out	ruled out by participant-happiness-lower-when-participant	false	`["ruled_out_by"]`
7	alt-guilt-effect-driven-by-own-outcome	hypothesis	rejects	ruled-out	ruled out by guilt-effect-occurred-whether-participant	false	`["ruled_out_by"]`
8	alt-imaging-contrast-invalid	hypothesis	rejects	ruled-out	ruled out by bilateral-ventral-striatum-more-active	false	`["ruled_out_by"]`
9	alt-model-based-glm-invalid	hypothesis	rejects	ruled-out	ruled out by manipulation-check-bilateral-ventral-striatum	false	`["ruled_out_by"]`
10	alt-participants-insensitive-to-value	hypothesis	rejects	ruled-out	ruled out by participants-probability-choosing-risky-option	false	`["ruled_out_by"]`
11	alt-social-context-shifts-risk-attitude	hypothesis	rejects	ruled-out	ruled out by risk-premiums-not-differ-between	false	`["ruled_out_by"]`
12	anterior-insula-neural-substrate-guilt	hypothesis	asserts	strong	its prediction anterior-insula-tracks-guilt-insula is confirmed and it is supported by insula-rois-responded-more-low; no rival stands	false	`["predictions","supported_by"]`
13	functional-connectivity-between-guilt-responsibility-related	hypothesis	asserts	strong	its prediction connectivity-between-guilt-responsibility-related-outcome-ph is confirmed and it is supported by functional-connectivity-between-left-anterior	false	`["predictions","supported_by"]`
14	neural-substrate-tracks-participant-responsibility	hypothesis	asserts	strong	its prediction neural-substrate-tracks-participant-responsibility-2 is confirmed and it is supported by one-cluster-left-sts-responded	false	`["predictions","supported_by"]`
15	responsibility-social-choice-yields-low	hypothesis	asserts	strong	both its predictions (responsibility-outcomes-generates-guilt-participant and responsibility-partner-outcomes-influences-participant) are confirmed and it is supported by when-partner-received-low-lot…	false	`["predictions","supported_by"]`
16	authors-suggest-left-sts-region	interpretation	asserts	weak	interprets the single moderate result one-cluster-left-sts-responded, so sits below it	false	`["interprets"]`
17	behavioural-guilt-effect-larger-happiness	interpretation	asserts	moderate	interprets the strongly-warranted guilt interaction when-partner-received-low-lottery	false	`["interprets"]`
18	connectivity-between-left-anterior-insula	interpretation	asserts	weak	interprets functional-connectivity-between-left-anterior, itself only moderate	false	`["interprets"]`
19	lower-happiness-when-participant-decision-maker	interpretation	asserts	weak	interprets participant-happiness-lower-when-participant, itself only moderate	false	`["interprets"]`
20	both-studies-participants-felt-worse	synthesis	asserts	weak	rests on a single result, when-partner-received-low-lottery; a synthesis needs two or more converging claims	false	`["supported_by"]`
21	participants-showed-very-similar-risk	synthesis	asserts	moderate	draws together four converging risk-preference results, including the rival-ruling-out control risk-premiums-not-differ-between and no-significant-interaction-between-difference	false	`["supported_by"]`
22	among-computational-models-fitted-momentary	methodological	asserts	weak	an AIC ranking that requires momentary-happiness-modelled-five-computational; nothing in the tree validates or supports it	false	`["requires"]`
23	bilateral-ventral-striatum-more-active	control	asserts	moderate	a positive control that rules out alt-imaging-contrast-invalid by replicating the risky>safe ventral-striatum effect	false	`["rules_out"]`
24	decisions-social-compared-solo-condition	empirical	asserts	weak	an fMRI contrast that requires the paradigm each-trial-participants-chose-between; no control or support bears on it	false	`["requires"]`
25	difference-response-between-low-high	empirical	asserts	weak	one measure that is part of insula-rois-responded-more-low, not independently warranted	false	`["part_of"]`
26	dot-products-between-individual-neural	control	asserts	weak	the tree records nothing bearing on this convergent-validity control	true	`[]`
27	during-outcome-phase-responses-low	empirical	asserts	weak	one measure that is part of insula-rois-responded-more-low	false	`["part_of"]`
28	during-receipt-lottery-versus-safe	empirical	asserts	weak	the tree records nothing bearing on this localizer contrast	true	`[]`
29	each-trial-participants-chose-between	scope	asserts	weak	a scope statement the tree records nothing bearing on	true	`[]`
30	findings-rest-two-samples-healthy	scope	asserts	weak	a scope statement the tree records nothing bearing on	true	`[]`
31	functional-connectivity-between-left-anterior	empirical	asserts	moderate	confirms the prediction connectivity-between-guilt-responsibility-related-outcome-ph	false	`["confirms"]`
32	guilt-effect-occurred-whether-participant	control	asserts	moderate	a control that rules out alt-guilt-effect-driven-by-own-outcome	false	`["rules_out"]`
33	hold-partner-behaviour-constant-across	methodological	asserts	weak	the tree records nothing bearing on this methodological choice	true	`[]`
34	individual-grbs-dot-product-values-not	control	asserts	weak	the tree records nothing bearing on this null control	true	`[]`
35	insula-rois-responded-more-low	empirical	asserts	moderate	confirms anterior-insula-tracks-guilt-insula and is validated by dot-products-between-individual-neural, but rests on the unsupported localizer during-receipt-lottery-versus-safe, so not strong	false	`["validated_by","confirms","requires"]`
36	left-ifg-cluster-showed-opposite	empirical	asserts	weak	an uncorrected result the tree records nothing bearing on	true	`[]`
37	left-superior-temporal-sulcus-cluster	empirical	asserts	weak	the tree records nothing bearing on this result	true	`[]`
38	likelihood-ratio-test-showed-responsibility	empirical	asserts	moderate	confirms the prediction responsibility-partner-outcomes-influences-participant and is supported by several happiness results	false	`["confirms","supported_by","requires"]`
39	linear-mixed-model-containing-all	methodological	asserts	weak	the tree records nothing bearing on this methodological result	true	`[]`
40	manipulation-check-bilateral-ventral-striatum	control	asserts	moderate	a manipulation-check control that rules out alt-model-based-glm-invalid	false	`["rules_out"]`
41	mass-univariate-voxel-wise-analysis-found-small	empirical	asserts	moderate	confirms the prediction anterior-insula-tracks-guilt-insula	false	`["confirms"]`
42	mixed-effects-regressions-choices-social-condition	empirical	asserts	weak	one analysis that is part of participants-chose-risky-option-lottery	false	`["part_of"]`
43	model-based-glm-entered-best-fitting-computational	methodological	asserts	moderate	validated by the control manipulation-check-bilateral-ventral-striatum	false	`["validated_by"]`
44	model-selection-among-happiness-models	methodological	asserts	weak	the tree records nothing bearing on this methodological choice	true	`[]`
45	momentary-happiness-modelled-five-computational	methodological	asserts	moderate	validated by the control risk-aversion-parameter-not-differ-between	false	`["validated_by"]`
46	no-significant-interaction-between-difference	control	asserts	weak	the tree records nothing bearing on this null control	true	`[]`
47	one-cluster-left-sts-responded	empirical	asserts	moderate	confirms the prediction neural-substrate-tracks-participant-responsibility-2 and is supported by left-superior-temporal-sulcus-cluster	false	`["confirms","supported_by","requires"]`
48	only-precuneus-tpj-showed-positive	empirical	asserts	weak	one analysis that is part of decisions-social-compared-solo-condition	false	`["part_of"]`
49	parameter-recovery-procedure-synthetic-data-generated	methodological	asserts	weak	the tree records nothing bearing on this methodological check	true	`[]`
50	participant-happiness-lower-when-participant	empirical	asserts	moderate	an empirical result that rules out alt-agency-aversion-not-guilt	false	`["rules_out"]`
51	participant-momentary-happiness-varied-rewards	empirical	asserts	weak	the tree records nothing bearing on this result	true	`[]`
52	participant-momentary-happiness-varied-rewards-2	empirical	asserts	weak	the tree records nothing bearing on this result	true	`[]`
53	participants-chose-risky-option-lottery	empirical	asserts	weak	a choice result that requires the paradigm each-trial-participants-chose-between; no control or support bears on it	false	`["requires"]`
54	participants-own-reward-prediction-errors	empirical	asserts	weak	the tree records nothing bearing on this result	true	`[]`
55	participants-probability-choosing-risky-option	control	asserts	moderate	a control that rules out alt-participants-insensitive-to-value	false	`["rules_out"]`
56	participants-slightly-more-risk-averse	empirical	asserts	weak	the tree records nothing bearing on this result	true	`[]`
57	partner-reward-prediction-errors-resulting	empirical	asserts	moderate	confirms the prediction responsibility-partner-outcomes-influences-participant	false	`["confirms"]`
58	pre-task-icebreaker-succeeded-establishing-positive	control	asserts	weak	the tree records nothing bearing on this control	true	`[]`
59	prior-functional-connectivity-work-shown	literature-context	asserts	weak	prior-literature context the tree records nothing bearing on	true	`[]`
60	prior-literature-documents-association-between	literature-context	asserts	weak	prior-literature context the tree records nothing bearing on	true	`[]`
61	responsibility-choices-not-influence-happiness	control	asserts	weak	the tree records nothing bearing on this null control	true	`[]`
62	responsibility-model-yielded-higher-values	empirical	asserts	weak	an R2 comparison that requires momentary-happiness-modelled-five-computational; nothing validates or supports it	false	`["requires"]`
63	responsibility-redux-model-incorporating-expected	empirical	asserts	moderate	validated by the control parameter-recovery-procedure-synthetic-data-generated	false	`["validated_by","requires"]`
64	risk-aversion-parameter-not-differ-between	control	asserts	weak	the tree records nothing bearing on this control	true	`[]`
65	risk-premiums-not-differ-between	control	asserts	moderate	a null control that rules out alt-social-context-shifts-risk-attitude	false	`["rules_out"]`
66	rutledge-colleagues-established-changes-momentary	literature-context	asserts	weak	prior-literature context the tree records nothing bearing on	true	`[]`
67	study-reproduced-study-design-inside	scope	asserts	weak	a scope statement the tree records nothing bearing on	true	`[]`
68	when-partner-received-low-lottery	empirical	asserts	strong	confirms the prediction responsibility-outcomes-generates-guilt-participant and is validated by four controls including guilt-effect-occurred-whether-participant	false	`["validated_by","confirms","requires"]`

Across the corpus

9 not run · 1 stale·a paper links to its own cell, where this layer's output for it is rendered

Paper	State	Version	Last run	Output	Cell
A three-dimensional immunofluorescence atlas of the …	not run	—	—	—	json
Distinct representational properties of cues and con…	not run	—	—	—	json
Computational modelling identifies key determinants …	not run	—	—	—	json
Contributions of insula and superior temporal sulcus…warrant v3 on the completed tree	stale	v3	2026-09-13	warrant.json	json
Spatially targeted inhibitory rhythms differentially…	not run	—	—	—	json
Feedback of peripheral saccade targets to early fove…	not run	—	—	—	json
iGABASnFR2 is an improved genetically encoded protei…	not run	—	—	—	json
A deep learning pipeline for mapping in situ network…	not run	—	—	—	json
Self-association enhances early attentional selectio…	not run	—	—	—	json
Impaired excitability of fast-spiking neurons in a n…	not run	—	—	—	json

Inputs and outputs

Reads, besides its dependencies: extract/prompts/warrant.md · declared, and not in the repository — it hashes to nothing, so it cannot make a run stale
extract/prompts/contract/vocabulary.md · declared, and not in the repository — it hashes to nothing, so it cannot make a run stale
scripts/warrant.py · declared, and not in the repository — it hashes to nothing, so it cannot make a run stale
Produces: runs/{paper}/warrant.json

One per paper — the table above links each one that exists.
Views: table — rendered above, over the 68 claims in the artifact

Running it

The command comes from the declaration, so this text and what actually runs cannot diverge. pipeline.py run also runs the unmet dependencies first.

python3 scripts/pipeline.py run <paper> warrant

Underneath, that runs cd extract && python3 -m claim_graphs.cli warrant --paper {paper}.