SEAM: Shortcut-Aware Real-Time Detection of Scripted vs. Spontaneous Speech for Interview Guardrails
arXiv:2606.06837v1 Announce Type: cross Abstract: Scripted vs spontaneous speech detection is appealing for interview guardrails, but benchmark performance can be inflated by shortcuts tied to corpus identity, channel conditions, and recording artifacts rather than speaking style itself. We present SEAM, a shortcut-aware framework for real-time scriptedness detection that combines uniform preprocessing, seam-aware sampling, non-speech augmentation, and a compact DistilHuBERT backbone. With 8s windows, the model achieves 0.971 +- 0.004 ROC-AUC on an external interview-domain evaluation set. Rem
The proliferation of advanced AI systems and the increasing use of automated interviews necessitate more robust methods for detecting subtle linguistic cues, and this research addresses a key limitation in current models.
This work directly impacts the reliability and fairness of AI-driven interview processes by improving the detection of deceptive speech patterns, crucial for applications ranging from hiring to security.
The ability to accurately differentiate scripted from spontaneous speech, even in the presence of 'shortcuts,' means AI models can better assess authenticity in verbal interactions, enhancing trust and reducing bias.
- · AI agents
- · Recruitment platforms
- · Security sectors
- · Speech analytics companies
- · Malicious actors
- · Interviewees using canned responses
Improved interview guardrails will lead to more accurate assessments of candidate sincerity and knowledge.
The enhanced detection capabilities could foster the development of more sophisticated AI interaction models, able to discern nuanced human communication.
This technology might eventually reshape human-AI interaction in critical decision-making processes, leading to new ethical considerations and regulatory frameworks for AI authenticity detection.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.LG