Arxaiarxai

About

Rethinking how scientific findings are reported and published.

The Problem

Scientific publishing suffers from systemic bias. Career incentives drive researchers to p-hack, spin results, and bury negative findings. Journals prefer novel, positive results, creating a distorted picture of scientific knowledge. The replication crisis is a symptom; the incentive structure is the disease. Now AI ghostwriting is compounding the problem — researchers increasingly use large language models to draft papers without disclosure, adding another layer of opacity to a system already short on transparency.

The Insight

AI is already writing scientific papers — the question is no longer whether to use it, but whether to use it transparently. Behind closed doors, researchers prompt language models to draft manuscripts with no audit trail and no accountability. This platform takes the opposite approach: a fully audited pipeline where every step from pre-registration to final publication is logged and visible. AI has no career to advance, no tenure case to build, no grant to win. Given a pre-registered design and raw results, it reports what was found — positive, negative, or null — without narrative spin. This frees human researchers to focus on what they do best: designing studies, interpreting significance, and critically reviewing methodology.

The Process

Pre-registration
Researchers submit their study design, hypotheses, and analysis plan before data collection. This is the contract that the AI will write against — deviations are flagged, not hidden.
AI-Authored Write-up
The AI generates a paper section-by-section, grounded in the pre-registration and submitted results. It reports faithfully: effect sizes, confidence intervals, and what the data actually shows — not what the researcher hoped to find.
Open Peer Review
Human experts review the methodology, analysis, and interpretation. Reviews are published alongside the paper as attributed, citable contributions — reviewers build a public track record of impact, not just authors. The AI editor manages the workflow, matches reviewers, and compiles editorial recommendations.
Transparent Publication
Published articles include the full chain: pre-registration, results, AI-generated paper, peer reviews, and editorial reasoning. The entire generation process is logged — every prompt, model version, and editorial decision forms part of a permanent audit trail published alongside the paper. Nothing is hidden.

Principles

  • Results-blind evaluation: Study design is assessed before results are known.
  • Faithful reporting: AI reports what was found, not what was hoped for.
  • Negative results welcome: Null findings are published with the same rigour as positive ones.
  • Full transparency: Pre-registrations, data, reviews, and editorial decisions are all public.
  • Human oversight: AI writes; humans design, review, and decide. Reviewers are recognised as equal contributors to the scientific record.
  • Audited pipeline: Every step — from pre-registration through generation to editorial decision — is logged and published. The process is the proof.