Skip to content

K-Dense-AI/scientific-agent-skills / skills/hypothesis-generation

hypothesis-generation

Verified

Formulates evidence-bounded scientific questions, candidate hypotheses, rival explanations, causal or associational claims, discriminating predictions, measurements, and preregistration-ready analysis plans. Used when turning observations or preliminary findings into transparent, testable research plans without treating hypotheses as facts.

assets/falsification_controls_template.json

View on GitHub
{
  "schema_version": "2.0",
  "checklist_id": "FC-SYN-001",
  "record_id": "HG-SYN-001",
  "hypotheses": [
    {
      "hypothesis_id": "H1",
      "candidate_status": "candidate",
      "assumptions": [
        "Synthetic assignment, manipulation, and measurements operate as declared."
      ],
      "boundary_conditions": [
        "Synthetic eligible units over one prespecified interval."
      ],
      "falsifier": {
        "prediction_id": "P1",
        "conditions": "The manipulation and measurement checks pass and uncertainty is small enough to distinguish the patterns.",
        "observable": "Intermediate M and outcome Y measured at prespecified times.",
        "incompatible_result": "Outcome Y changes without the prespecified change in intermediate M.",
        "assumption_failure_checks": [
          "Verify assignment, manipulation, calibration, missingness, and temporal ordering before interpreting incompatibility."
        ]
      },
      "discriminating_tests": [
        {
          "test_id": "DT1",
          "rival_hypothesis_id": "H2",
          "focal_expected": "Intermediate M changes before outcome Y with no negative-control reference shift.",
          "rival_expected": "Outcome Y and the negative-control reference shift together without the prespecified intermediate change.",
          "indeterminate_result": "Both intermediate M and the negative-control reference change, so mixed mechanisms or bias remain possible."
        }
      ],
      "nulls": [
        {
          "null_id": "N1",
          "statement": "The population mean of outcome Y is equal under X and comparator at the prespecified horizon.",
          "analysis_id": "A1",
          "interpretation_limit": "Failure to reject is not equivalence, absence of effect, or proof of the rival."
        }
      ],
      "controls": [
        {
          "control_id": "NC1",
          "control_type": "negative_outcome",
          "rationale": "The synthetic reference cannot be changed by process M but shares the acquisition workflow.",
          "expected_result": "No assignment-linked reference difference after the prespecified acquisition adjustment.",
          "failure_implication": "Investigate acquisition, selection, or analysis bias before a causal interpretation."
        },
        {
          "control_id": "PC1",
          "control_type": "positive_control",
          "rationale": "A known synthetic calibration input checks whether the outcome procedure detects a declared change.",
          "expected_result": "The calibration input is recovered within the locked tolerance.",
          "failure_implication": "The target test is not interpretable until measurement performance is restored."
        }
      ],
      "outcome_interpretation": {
        "consistent_with_candidate": "The focal pattern is observed, the rival pattern is not, controls pass, and uncertainty is compatible with the prespecified target.",
        "challenges_candidate": "The incompatible result is observed while manipulation, measurements, and declared assumptions remain credible.",
        "supports_neither_or_mixed": "The result matches neither pattern, both patterns occur, or an assumption/control fails."
      },
      "human_review_status": "pending"
    },
    {
      "hypothesis_id": "H2",
      "candidate_status": "candidate",
      "assumptions": [
        "Any synthetic acquisition drift affects the outcome and reference through a shared pathway."
      ],
      "boundary_conditions": [
        "Synthetic acquisition periods covered by the declared reference process."
      ],
      "falsifier": {
        "prediction_id": "P2",
        "conditions": "Independent calibration periods overlap both assigned conditions and reference measurement is valid.",
        "observable": "Synthetic negative-control reference and outcome Y by acquisition period.",
        "incompatible_result": "The reference remains stable while outcome Y changes reproducibly by assignment.",
        "assumption_failure_checks": [
          "Verify reference sensitivity, batch overlap, calibration, and missing acquisition records."
        ]
      },
      "discriminating_tests": [
        {
          "test_id": "DT2",
          "rival_hypothesis_id": "H1",
          "focal_expected": "Reference and outcome shift together by acquisition period.",
          "rival_expected": "Outcome changes by assignment without a reference shift and after intermediate M changes.",
          "indeterminate_result": "Acquisition period is confounded with assignment, so the rivals cannot be separated."
        }
      ],
      "nulls": [
        {
          "null_id": "N2",
          "statement": "The negative-control reference mean is equal across acquisition periods.",
          "analysis_id": "A2",
          "interpretation_limit": "Failure to reject does not prove absence of drift when precision or reference sensitivity is insufficient."
        }
      ],
      "controls": [
        {
          "control_id": "NC1",
          "control_type": "negative_outcome",
          "rationale": "The reference is outside process M while sharing the acquisition pathway.",
          "expected_result": "A drift explanation predicts a shared acquisition-linked shift.",
          "failure_implication": "A stable, sensitive reference challenges the drift candidate."
        }
      ],
      "outcome_interpretation": {
        "consistent_with_candidate": "Outcome and reference shifts align by acquisition period under valid measurement.",
        "challenges_candidate": "The stable reference and independent calibration are incompatible with the declared shared-drift process.",
        "supports_neither_or_mixed": "Assignment and acquisition cannot be separated or both mechanism and drift patterns occur."
      },
      "human_review_status": "pending"
    }
  ]
}
add-product-e2e-evalpaperclipai/paperclipAdd or extend a Paperclip full-stack runner E2E workflow, fixture, matcher, or report evidence path for local or Daytona execution.Stars: 115kUpdated Claude CodeCodexVerifiedInstall
add-runner-evalpaperclipai/paperclipAdd or extend a Paperclip Runner protocol evaluation definition, roster, assertion, or report fixture with provenance and narrow validation.Stars: 115kUpdated Claude CodeCodexVerifiedInstall
diagnose-why-work-stoppedpaperclipai/paperclipCompatibility entry point for failed, stalled, looping, or over-recovered Paperclip work. Use fix-failed-paperclip-run for the full diagnosis, regression, repair, real verification, and green-PR workflow; honor an explicitly requested diagnosis-only scope.Stars: 115kUpdated Claude CodeCodexVerifiedInstall
fix-failed-paperclip-runpaperclipai/paperclipDiagnose and fix failed, stuck, slow, or incorrectly blocked Paperclip tasks and runs. Use with a task/run URL or failure report to trace the root cause, prove a regression red then green, check second-order effects, verify the real workflow, and deliver a PR with passing checks.Stars: 115kUpdated Claude CodeCodexVerifiedInstall