Software & Version Provenance

"Analysis was performed in Licklider" is not a reproducibility statement; it is incomplete. Which version? Which ruleset? Which statistical backend? Licklider's Software & Version Provenance captures all of it automatically at export time, generating a snapshot that makes computational results traceable and re-runnable.

STEP 1 - The Pitfall: Version-Blind Methods Sections

Software versioning is one of the most systematically underreported aspects of computational research. A Methods section that says "statistical analysis was performed in R" might describe work done in R 3.6 or R 4.3, and the results of the same code can differ between versions due to changes in default parameters, random number generation, and package behavior.

For AI-assisted analysis tools like Licklider, version transparency is even more critical:

• Ruleset versions change the guard logic. A figure that passes the Statistical Audit in ruleset v2.1 might be flagged in ruleset v2.4 because a new guard was added.
• Statistical backend updates affect numerical outputs. Updates to the Python statistical engine can change p-values and confidence intervals at floating-point precision.
• Model versions affect AI-generated interpretations. Methods text, figure legends, and analysis suggestions are all shaped by a versioned model component.

STEP 2 - Journal Requirement

Nature Methods, PLOS ONE, and Science family journals all expect enough software detail for readers to understand the computational environment used in the analysis. Under the TOP guidelines, software citation including version and repository information is often a formal reproducibility requirement.

Licklider's Software & Version Provenance feature generates a citation-ready software block without requiring manual bookkeeping.

STEP 3 - Licklider's Solution

Input

• No user input is required. The version snapshot is generated automatically for every export.
• The snapshot captures the actual system state at analysis time and is bundled automatically at export.

Output

• Version snapshot JSON. A machine-readable file capturing app version, ruleset version, backend versions, model identifier, and timestamp.
• Human-readable software section. A formatted Methods paragraph ready to paste into a manuscript.
• Software citation draft. A structured citation draft ready to adapt for APA, Vancouver, or BibTeX submission. (Full multi-format citation export is planned for a future release.)

Guard

STEP 4 - Ready-to-Publish Output

Example: Methods - Software section

Statistical analyses were performed using Licklider (version 2.4.1;
ruleset v3.2.0; stats backend python-stats@1.11.4, schema 3, build 202504;
model anthropic:claude-sonnet-4-5-20251001). All analyses were executed on
2026-04-05T10:22:31Z (UTC).
[Edit before submission: expand backend and model details as required by
the target journal's software citation policy.]

Example: Software citation (APA format)

Licklider, Inc. (2026). Licklider (Version 2.4.1) [Statistical analysis
software]. https://doi.org/10.5281/licklider.2.4.1

Example: Machine-readable version snapshot (JSON)

{
  "app_version": "2.4.1",
  "ruleset_version": "3.2.0",
  "stats_backend_version": "1.11.4",
  "model_version": "gpt-4o-2024-11-20",
  "executed_at": "2026-04-05T10:22:31Z"
}

Related Features