Search papers, labs, and topics across Lattice.
This study investigates the interpretability of Domain-Adapted models via Prompt-based Fine-tuning (DAPF) for dementia detection, framing the task as diagnosis-related masked-token prediction. The authors found that while DAPF achieved superior performance metrics (accuracy=0.83 and macro-F1=0.83), the model's representations did not yield faithful token-level explanations, as attributions were influenced more by language artifacts than by the underlying diagnostic information. These findings highlight a critical trade-off between diagnostic accuracy and interpretability in AI-driven dementia screening tools.
DAPF models excel in dementia detection but struggle to provide trustworthy explanations, revealing a gap between performance and interpretability.
Spoken-language analysis via prompt-based domain-adaptive models is a promising direction for low-resource, non-invasive dementia screening, but such models remain internally opaque. We study the interpretability of the Domain-Adapted models via Prompt-based Fine-tuning (DAPF) framework, which casts dementia detection as diagnosis-related masked-token prediction. We interpret DAPF and strong baselines using a variety of probing and analysis techniques, finding that DAPF achieved the best overall performance (accuracy=0.83 and macro-F1=0.83) with diagnosis most recoverable from its [MASK] representation. However, this representational advantage did not extend to token-level explanation faithfulness. DAPF attributions primarily reflected language task vocabulary, discourse markers, and transcription artifacts, with perturbation tests showing weak or negative effects. This suggests that its masked-token interface determines diagnosis information without producing faithful token-level explanations.
