Noesis
Menu
EN
繁中简中EN

SCIENCE & LIMITS

Keep the research question, results, and limits separate.

This page records the current state of a research programme. It does not turn a design, a source, or product intent into a result.

Current status: research design only; no public results

A self-report score and a conversational estimate are different outputs

A standardized self-report result comes only from your explicit answer to every item in that inventory. If you actively ask for it, AI may organize an observation from conversation as an AI-assisted conversational estimate that you can confirm, revise, or reject one point at a time; it never becomes an inventory score automatically.

Until the G2 concurrent-validity study is completed, reviewed, and its results are released, Noesis does not say that this conversational estimate is equivalent to, replaces, or validates any standardized self-report result.

G2: the study design is complete; execution has not started

The planned study compares one participant’s explicit self-report with an optional, correctable conversational estimate. It must first complete applicable ethics, recruitment, data-protection, and preregistration steps.

At the time of this update, there has been no recruitment, participant contact, data collection, analysis, validity coefficient, or public research conclusion. This is not a claim that any product feature has passed validation.

Why a conversational impression cannot be treated as a scale

Published and preprint studies both show that an LLM’s personality inferences from text or conversation vary with the data source, interaction format, and research design. Those external studies provide context for a question; they are not validity evidence for Noesis.

For that reason, the two-lane product contract requires the basis of an estimate, your confirmation, and the provenance of self-report answers to remain separate. Future research data must also distinguish the time an event is described as occurring from the time the record was made. This public page does not ask you to enter or submit either kind of data.

High-sensitivity adult self-understanding work is not claimed as a general public feature

Any product entry point that may involve high-sensitivity adult reflection first needs a verifiable adult-only boundary, explicit and specific consent, a data-protection assessment, and applicable legal and content review. Until those gates are complete, Noesis does not describe it as a publicly available feature or collect that material on a public page.

Any future session-only guidance must let a person keep uncertainty, skip, or stop. It is not diagnosis, treatment, crisis care, or a tool for making relationship decisions for someone else.

AI-assisted visuals and low-latency delivery are a separate matter

AI-assisted editorial visuals on this site carry a visible disclosure, an editorial contact, and a machine-readable source marker. This is a Noesis release-governance practice; it does not label every abstract image as a particular legal category.

Approved static visuals are delivered through same-origin, content-versioned URLs that can be cached. Private source and release bytes remain in a controlled archive. That delivery design does not change the standards for research evidence, consent, or product validity.

Optional cross-platform context remains separate from scores

The private local-review flow can read an official ChatGPT or Claude archive you choose, or a Kimi memory summary you paste; that page does not upload the raw archive or treat its content as inventory answers or scores. Every excerpt starts excluded and requires item-by-item review of source, time, sensitivity, and third-party information.

Review optional conversation context locally

Read external research with its source data and limits

For item sources, papers, and public release records, use the Evidence Library. Open the Evidence Library