Synthetic Memory & Source Monitoring

Scope

Consider source confusion, repeated synthetic narratives, AI-generated retrieval cues, and authenticity calibration. Keep a person’s report of recollection separate from digital evidence of exposure and from an experimentally supported causal explanation.

Questions worth testing

• Can participants distinguish a generated cue from an original source?
• How do source labels and time delays affect judgments?
• What design separates exposure effects from prior knowledge and suggestion?

Evidence and study material

With appropriate consent, controlled stimuli, versioned exposure logs, source labels, interview protocols, and preregistered measures can support a study. Private chat records are not an ethically neutral substitute for a participant research design.

Methods to begin with

• preservation
• corroboration
• confidence

Limits on interpretation

The established source-monitoring framework does not by itself prove an AI-specific effect. Self-report, small samples, leading questions, and uncontrolled prior exposure can limit interpretation. No clinical advice or individual diagnostic inference is offered.

Open research questions

Open questions include authenticity calibration over time and the preservation of source labels in AI-assisted work. Terms such as reality-model drift remain project working concepts until operational definitions and empirical tests are published.

Tools, projects, datasets, and publication

See Synthetic Imprinting for the related project and the glossary for carefully bounded working terms. No clinical tool or participant corpus is available here. Journal studies should state ethics review, consent, measures, and replication limits.

• Tools and documentation resources
• Current project status
• Available reference data and dataset limits
• Journal

How do you study source judgments responsibly?

Define the exposure, comparison, source labels, task, and consent process. Measure source judgments and confidence under those conditions, keeping participant privacy and study limitations visible. Separate errors in a task from claims about a person’s wider memory or wellbeing.

Report the measured outcome and suitable uncertainty. Project terms such as synthetic imprinting can organize a hypothesis, but they do not constitute a diagnosis or an established AI-specific causal effect.

Related fields & glossary terms

• Interaction Forensics

Apply the examination methods, inspect the evidence classes, and check relevant external standards and specifications.