Validation study demonstrates deterministic auditing of normalized speech and assistive communication pathways, highlighting critical event preservation in human–AI dialogue records.
TopoDialogue IV introduces a modality-provenance event contract and deterministic audit for measuring which observable interaction events and features are preserved, omitted, or falsely generated when speech, direct text, and AAC pathways are normalised into dialogue records. Five constructed cases validate arithmetic, missingness reasons, correction history, and privacy-preserving local references. The study makes no field-accuracy, diagnostic, emotional, disability, or cross-cultural-validity claim; human agreement and participatory review remain explicit next gates. Includes a British-English PDF, Japanese companion text, source, tests, figures, and constructed fixtures.
No takes yet. Share an insight, caveat, or question.
Ryoko Sakakibara (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: