FROM SPEECH RECOGNITION TO ALGORITHMIC TRIAGE: HOW POST-9/11 INTELLIGENCE AUTOMATION RECONFIGURED POWER, BIAS, AND ACCOUNTABILITY

Utkarsh Srivastava, Rahul D Ray, Rishi Mohapatra

Track: tiny / short paper (up to 4 pages)

Keywords: algorithmic triage, voice AI, speech recognition, voice biometrics, surveillance infrastructures, post-9/11 security, algorithmic governance, sociotechnical audits, bias and accountability, militarization of AI, fusion centers, epistemic infrastructure, positive peace, preemptive security

TL;DR: Voice-based AI now functions as algorithmic triage in post-9/11 security systems, filtering and ranking speech upstream of human judgment while embedding bias and diffusing accountability.

Abstract:

Post-9/11 security reforms normalized large-scale automated surveillance by reframing
intelligence failure as a problem of data integration. In parallel, advances
in artificial intelligence, particularly speech recognition and voice biometrics rendered
spoken communication computable at population scale. This paper argues
that voice-based AI operates as algorithmic triage: upstream systems that probabilistically
filter, rank, and render speech intelligible prior to human judgment. We
formalize algorithmic triage as an epistemic infrastructure with identifiable stages
and failure modes, and show why voice is a uniquely powerful and dangerous
modality, entangling identity, behavior, and cultural difference in a single signal.
We further propose a voice-specific sociotechnical audit framework as a workin-
progress research agenda. We argue that algorithmic triage erodes conditions
of positive peace by normalizing perpetual suspicion and shifting accountability
away from contestable human institutions. Addressing these dynamics requires not
only ethical critique, but methodological tools for interrogating how voice AI is
embedded within security infrastructures.

Email Sharing: We authorize the sharing of all author emails with Program Chairs.

Data Release: We authorize the release of our submission and author names to the public in the event of acceptance.

Submission Number: 11