Audit a theory-of-mind study without turning one belief-reasoning score into “mind reading”

Reset: Re-read this authored instruction — “Ask: “In the declared population, does the named training improve novel belief-reasoning accuracy beyond an active control, and is any gain retained?”” — and its success check, then attempt only this step.

1. Possible snag: Call vignette accuracy broad social competence.

   Correction: Label unmeasured function “not tested.” within the two-study theory-of-mind audit.

Stop / get help: Stop the review if a source cannot be verified or the study identity is uncertain.

Reset: Re-read this authored instruction — “Make four columns: belief, emotion, intention and language. Mark which each test actually requires.” — and its success check, then attempt only this step.

1. Possible snag: Treat every social-cognition measure as the same theory-of-mind trait.

   Correction: Split outcomes by belief, emotion, intention and language demand.

Stop / get help: Stop the review if a source cannot be verified or the study identity is uncertain.

Reset: Re-read this authored instruction — “Record sample size, age or diagnosis, recruitment, randomisation, exclusions and attrition for each study.” — and its success check, then attempt only this step.

1. Possible snag: Use one score to diagnose or rank a person.

   Correction: Remove person-level labels and return to group-level study claims.

Stop / get help: Stop the review if a source cannot be verified or the study identity is uncertain.

Reset: Re-read this authored instruction — “Write the exact stimulus, response format, scoring rule, cueing and whether the item was trained or novel.” — and its success check, then attempt only this step.

1. Possible snag: The result from “Write the exact stimulus, response format, scoring rule, cueing and whether the item was trained or novel.” does not yet meet this declared check: Another reader could tell what counted as a correct answer.

   Correction: Return to the start of “Rebuild the task”, reduce complexity or pace, and repeat only the part needed to satisfy: “Another reader could tell what counted as a correct answer.”

Stop / get help: Stop the review if a source cannot be verified or the study identity is uncertain.

Reset: Re-read this authored instruction — “Name the active or passive control, immediate test, delayed test and any chosen functional outcome.” — and its success check, then attempt only this step.

1. Possible snag: The result from “Name the active or passive control, immediate test, delayed test and any chosen functional outcome.” does not yet meet this declared check: A within-group gain is not mistaken for an advantage over control.

   Correction: Return to the start of “Record comparators and timing”, reduce complexity or pace, and repeat only the part needed to satisfy: “A within-group gain is not mistaken for an advantage over control.”

Stop / get help: Stop the review if a source cannot be verified or the study identity is uncertain.

Reset: Re-read this authored instruction — “Check language practice, repeated items, expectancy, assessor awareness and unequal contact time.” — and its success check, then attempt only this step.

1. Possible snag: The result from “Check language practice, repeated items, expectancy, assessor awareness and unequal contact time.” does not yet meet this declared check: At least one viable non-mentalising explanation is retained when the design cannot remove it.

   Correction: Return to the start of “Test alternative explanations”, reduce complexity or pace, and repeat only the part needed to satisfy: “At least one viable non-mentalising explanation is retained when the design cannot remove it.”

Stop / get help: Stop the review if a source cannot be verified or the study identity is uncertain.

Reset: Re-read this authored instruction — “Put primary interaction, confidence interval or uncertainty beside the authors’ conclusion; note the active-control null finding where applicable.” — and its success check, then attempt only this step.

1. Possible snag: Celebrate improvement in the intervention arm alone.

   Correction: Copy the active-control estimate and its uncertainty into the table.

Stop / get help: Stop the review if a source cannot be verified or the study identity is uncertain.

Reset: Re-read this authored instruction — “Require independent replication, novel tasks, delayed retention, participant-valued function and adverse-event reporting before calling the method trainable.” — and its success check, then attempt only this step.

1. Possible snag: The result from “Require independent replication, novel tasks, delayed retention, participant-valued function and adverse-event reporting before calling the method trainable.” does not yet meet this declared check: The final paragraph states what evidence is missing and contains no home protocol.

   Correction: Return to the start of “Write the next gate”, reduce complexity or pace, and repeat only the part needed to satisfy: “The final paragraph states what evidence is missing and contains no home protocol.”

Stop / get help: Stop the review if a source cannot be verified or the study identity is uncertain.