BERT and MentalBERT remain resilient under sentence shuffling in ENS-based mental health classification¶

Claim · All claims
Evidence · 1 study · causal r? · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person · recheck by 2029: tool and model accuracy changes quickly

Subclaims¶

q2 i? Both BERT and MentalBERT remain resilient under sentence shuffling (P-value < 0.05), a capability the authors deem critical for analyzing extended narratives. → Jinwen Tang 2026

Evidence¶

Jinwen Tang 2026¶

Jinwen Tang. (2026). NEW AI-DRIVEN TOOLS FOR ENHANCING CAMPUS WELL-BEING: A PREVENTION AND INTERVENTION APPROACH. Doctoral dissertation, University of Missouri-Columbia. https://arxiv.org/abs/2605.10804 doi:10.48550/arxiv.2605.10804

q2 · i? · causal · r?

Sentence-shuffling manipulation experiments on Expressive Narrative Stories tested order sensitivity of transformer models. The abstract reports both BERT and MentalBERT "remain resilient under sentence shuffling (P-value < 0.05)"; no effect size is printed.

"Both BERT and MentalBERT remain resilient under sentence shuffling (P-value < 0.05), a critical capability for analyzing extended narratives, particularly in ENS-based comparisons of individuals with and without self-declared mental health issues."

Discussion¶

Related Claims¶

• Conventional classifiers and MentalBERT depend heavily on explicit mental health terms in Expressive Narrative Stories, while BERT(128) maintains robust accuracy without keyword cues — related
• Text splicing into single sentences improves ML and BERT Bloom classification on the Scaria OOD dataset