The Bond Test: A Cognitive Inversion of the Turing Test — A Protocol for Measuring Whether People Can Tell Their Own Writing from an AI's Reflection

Participants write a short text without assistance, develop the same topic with a fluent AI system, and then classify short segments — from their own earlier text, from the AI, and from another participant — as “mine” or “not mine,” rating their confidence; finally, they perform an equivalent task alone. The Bond Test, introduced as a secondary measure in version 2.0 of the Sufficient Mirror Theory (SMT) and retained in the parent theory v3.0 (§10.8), is developed here into a stand-alone protocol. It is framed as a cognitive inversion of the Turing Test in which the judge is also one of the witnesses: the subject’s own unassisted writing is the human reference against which the machine’s output is judged. The design separates self-recognition from detection of machine style through an other-human class, adds an exploratory reflected class — the system’s responses to the participant’s own prompts — and measures failure at two levels: source discrimination (d′) and the monitoring of that discrimination (meta-d′ and an authorship calibration gap). β is estimated only from exposure-phase data (firewall rule); the Parasitic Flow Score, which contains β by construction, is reported only descriptively. The document reports no data. It formulates seven exploratory hypotheses (B1–B7) with refutation conditions, including a passive-exposure arm that separates familiarity with the model’s style from loss of agency. The central prediction is that, under qualifying high-coupling conditions characterized by high fluency, people’s ability to recognize their own writing declines specifically against the AI source, and that their estimate of their own accuracy declines less than their observed accuracy. This record contains the canonical English version and the Portuguese version. Project website: https://www.sufficientmirror.com

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-12
DOI
https://doi.org/10.5281/zenodo.24064860
Primary Topic
Authorship Attribution and Profiling
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

The Bond Test: A Cognitive Inversion of the Turing Test — A Protocol for Measuring Whether People Can Tell Their Own Writing from an AI's Reflection

C. S. S. Sandalo
Zenodo (CERN European Organization for Nuclear Research)
Authorship Attribution and Profiling
preprint

The Bond Test: A Cognitive Inversion of the Turing Test — A Protocol for Measuring Whether People Can Tell Their Own Writing from an AI's Reflection

C. S. S. Sandalo
preprint en

Abstract

Participants write a short text without assistance, develop the same topic with a fluent AI system, and then classify short segments — from their own earlier text, from the AI, and from another participant — as “mine” or “not mine,” rating their confidence; finally, they perform an equivalent task alone. The Bond Test, introduced as a secondary measure in version 2.0 of the Sufficient Mirror Theory (SMT) and retained in the parent theory v3.0 (§10.8), is developed here into a stand-alone protocol. It is framed as a cognitive inversion of the Turing Test in which the judge is also one of the witnesses: the subject’s own unassisted writing is the human reference against which the machine’s output is judged. The design separates self-recognition from detection of machine style through an other-human class, adds an exploratory reflected class — the system’s responses to the participant’s own prompts — and measures failure at two levels: source discrimination (d′) and the monitoring of that discrimination (meta-d′ and an authorship calibration gap). β is estimated only from exposure-phase data (firewall rule); the Parasitic Flow Score, which contains β by construction, is reported only descriptively. The document reports no data. It formulates seven exploratory hypotheses (B1–B7) with refutation conditions, including a passive-exposure arm that separates familiarity with the model’s style from loss of agency. The central prediction is that, under qualifying high-coupling conditions characterized by high fluency, people’s ability to recognize their own writing declines specifically against the AI source, and that their estimate of their own accuracy declines less than their observed accuracy. This record contains the canonical English version and the Portuguese version. Project website: https://www.sufficientmirror.com

Zenodo (CERN European Organization for Nuclear Research)
Authorship Attribution and Profiling
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon