Research · method · evidence · publication
Ask the question publicly. Preserve its status. Publish only what the evidence supports.
FASO research is intended to make important propositions testable without turning preliminary observation into established fact. Every publication carries its identity, status, version, sources, limitations and correction history.
Research and Publications
A public record of the questions FASO asks, the methods it proposes and the evidence it can support.
Every publication carries an explicit evidence status, version and limitation. The section makes each output identifiable and challengeable, while social-media discussion points back to this complete record.
Publication method · 27 July 2026
The Project FASO Research and Publication Method
The governing method for separating evidence, observation, hypothesis, protocol and finding while preserving versions, limitations and corrections.
Read the complete publicationWorking hypothesis · 27 July 2026
From Instruction Attenuation to Operating-Regime Reversion
A proposed causal model in which abbreviation or attenuation of an accuracy-governing instruction may produce measurable reversion towards completion-oriented behaviour.
Read the complete publicationWorking hypothesis · 27 July 2026
Evaluation-Induced System Drift
A proposed system-level model in which continual evaluation can become causal when evaluation-derived information enters persistent technical, operational or institutional feedback channels.
Read the complete publicationWorking hypothesis · 27 July 2026
Cross-Model Evaluation Propagation and Multiplier Effects
A proposed network-level model in which evaluation-derived information can propagate through connected models, shared infrastructure, human remediation, distillation and later generations.
Read the complete publicationValidation protocol · 27 July 2026
Controlled Validation Protocol for Instruction Attenuation
A proposed controlled comparison of absent, abbreviated, paraphrased, repeated and mechanically verified governing instructions.
Read the complete publicationPublication classes
The label tells the reader what the document can establish.
Every output belongs to a declared class. A working hypothesis cannot be presented as a finding, and an evidence report cannot silently become a FASO investigation.
Evidence reports
Neutral presentation of sourced external information, including what the evidence establishes, its limitations and the absence of any FASO investigative role.
Working hypotheses
Testable explanations derived from observation and established research but not yet validated by controlled FASO evidence.
Validation protocols
Methods specified before results, including variables, controls, stopping rules, evidence requirements and falsification conditions.
Research findings
Completed results published with methods, data boundaries, negative results, uncertainty, limitations and independent review status.
Institutional publications
Governance, funding, conflicts, corrections, annual accountability and decisions about what remains behind the Publication Boundary.
Evidence status
A visible status prevents an unfinished proposition from acquiring false authority.
Status changes require a new version or a new publication. Earlier versions remain identifiable and must not be silently overwritten.
A proposed explanation awaiting controlled testing.
A test method open to review and not yet preregistered or executed.
Evidence is being formed; no terminal conclusion is available.
The declared method supports the finding within stated boundaries.
The evidence did not support the tested proposition.
The evidence cannot distinguish the relevant explanations.
A later identified version replaces this one without erasing it.
The record remains available with the reason for withdrawal.
Research areas
Technical reliability and institutional credibility are separate propositions.
FASO research examines the observation system, the behaviour of changing artificial-intelligence systems and the organisation required to hold a neutral public-interest duty.
Artificial-intelligence behaviour
Instruction continuity and operating-regime stability
How governing instructions, abbreviations, context position, task depth and verification affect exact compliance over time.
Read the first working hypothesisEvaluation and feedback
Evaluation-induced system and measurement drift
Whether continual testing remains observation or becomes an intervention through memory, adaptation, remediation, training and institutional feedback.
Read the second working hypothesisModel ecosystems
Cross-model propagation and multiplier effects
Whether evaluation-derived knowledge can spread sideways through model stacks and shared infrastructure, then forward into other systems and later model generations.
Read the third working hypothesisObservation system
Longitudinal evidence, replay and state placement
Whether governed evidence can be preserved, replayed and placed consistently enough to support bounded observations.
Open the research programmeInstitution
Authority, independence and protected purpose
Whether legal form, governance, funding, participation and publication arrangements can preserve neutrality and accountability.
Examine institutional validationResearch integrity
Begin with the method, then examine the three working hypotheses.
The method explains how status and evidence are controlled. The three working hypotheses address connected instruction, evaluation-feedback and cross-model propagation problems.