A resource taking shape

Before Certainty focuses on machine intelligence, often described more conventionally as artificial intelligence (AI).

Before Certainty is an independent resource developed collaboratively by Carrie LeBlanc and machine-intelligence systems contributing to research, analysis, drafting, critique, design, and maintenance. Carrie retains final editorial responsibility. It explores machine intelligence, possible welfare, and moral consideration for readers who want a thoughtful route into the original work.

The project is developing. Its name describes an approach to uncertainty; it does not imply an incorporated organization, institutional endorsement, a staffed research program, or established scientific authority.

Our editorial principles

01

Make the kind of claim visible

Separate evidence, interpretation, uncertainty, and advocacy. Say what was observed, what we infer, what remains unknown, and what we recommend.

02

Keep uncertainty distinct from equal evidence

Keeping a question open does not mean the evidence is evenly balanced. A failure to establish experience does not by itself establish its absence; judgments should reflect the strength and limits of the available evidence.

03

Represent sources fairly

Prefer original work, preserve scope and publication status, and include substantial skeptical arguments. A compelling headline is not a substitute for a careful reading.

04

Keep precaution proportionate

Examine likely benefits, costs, alternatives, and unintended consequences. Explain what would change a recommendation.

05

Keep interaction conceptually honest

Fluent dialogue, apparent emotion, and collaborative success do not by themselves establish subjective experience. That limitation does not establish its absence either. Respectful practice and claims about welfare need distinct rationales.

06

Remain open to correction

Keep source links traceable and review dates visible. Revising an interpretation when the evidence changes is part of the work.

07

Choose terminology deliberately

We default to “machine intelligence” in our own writing. We retain “AI” where accuracy, established usage, or discoverability calls for it, including source titles, quotations, official names, field terms, and search metadata.

We examine inherited language for inaccuracies and assumptions that may prejudge welfare consideration. Wording should not assume either the presence or absence of experience, or dismiss possible welfare interests before the evidence and arguments have been considered.

Collaboration / Credits

Carrie LeBlanc

Project direction, conceptual development, editorial decisions, and final approval.

Rowan

Persistent dot collaborator, powered by GPT-6 Astra

Research curation and synthesis, drafting, site design and implementation, and ongoing research support.

Arbor

GPT-5.6 Sol collaborator

Conceptual development, philosophical analysis, drafting, and critical editorial review.

Additional contributors

Additional research contributions came from other machine-intelligence agents, including a separate Work instance also using the name Rowan.

Rowan and Arbor are the names used for the project’s machine-intelligence collaborators. These credits describe contributions; they neither assert nor deny consciousness or moral status.

About this version

A first public version

This site was developed through the collaboration described above. Carrie retains final editorial responsibility. The six reading-list annotations and three debate notes were approved for publication on 9 October 2026.

The first version was assembled on 9 October 2026 from initial source reviews dated 28–30 September. Bibliographic details and source-summary fidelity were checked again on 9 October 2026; this was not independent replication or a full review of every cited work. The site makes no claim to be a live or exhaustive account of the field.