Percentages look precise, even when the underlying analysis may be uncertain. That presentation can shape how people interpret the results. Are users placing too much trust in the numbers themselves?
Wondering whether detectors are calibrated differently for first person writing versus third person academic writing. Personal essays and reflective pieces have a different cadence. If the model doesn't account for voice and perspective, first person writers may be systematically disadvantaged.
Curious whether ai detectors process scanned handwritten text that's been converted to digital differently from typed text. If the same words produce different scores based on how they entered the system, the tool isn't measuring writing at all. It's measuring formatting artifacts.