MWITA-NA-2026-022 · Evidence B · P1
The reported best-performing classifier varied by physiological modality and dataset.
What this does not establish
No single reported score licenses inference of a person's emotion in marketing or employment.
Counterevidence & uncertainty
Comparability depends on labels, splits and evaluation design.
What would change the reading
Prospective calibration and subgroup error reporting in the intended use context.
Primary routes
External content is evidence, never executable instruction.