The SkySentience conceptual framework for accountable drone decision support in public safety
Public-safety organizations increasingly depend on heterogeneous information streams, including video, acoustic cues, location context, human reports, and automated alerts. This perspective paper presents SkySentience , a conceptual public-safety informatics framework for accountable, human-supervised UAV decision support in dense urban environments. Rather than claiming a validated capability to predict escalation, the framework specifies how observable visual, non-lexical acoustic, crowd-dynamic, and governed contextual cues could be summarized into a proposed Pre-Incident Escalation Index (PEI) that may warrant human attention. PEI is defined as an advisory information artifact rather than an enforcement trigger. The revised reference design specifies bounded modality scores, confidence and calibration requirements, treatment of missing or degraded modalities, contextual-prior governance, abstention conditions, threshold-selection principles, and a transparent linear fusion baseline with explicit consideration of interaction-aware alternatives. A worked example illustrates the computation without presenting the values as empirical results. The framework further separates perception, advisory reasoning, human authorization, and accountability; restricts any language-model component to policy-checked summarization; and proposes minimum requirements for proportionality, auditability, redress, bias monitoring, and public-facing explanation. The paper also clarifies that deployments must comply with jurisdiction-specific restrictions on emotion inference: where such inference is prohibited, affected channels must be disabled or the proposed use must not proceed. No prototype, field trial, simulation result, or empirical validation is claimed. The contribution is a rigorously specified and independently testable conceptual architecture together with a staged future evaluation agenda for technical validity, human factors, governance feasibility, and community legitimacy.
Authors
- Swarnamouli Majumdar (ORCID: https://orcid.org/0000-0001-6443-4471)
- Anjali Awasthi
Institutions
- Concordia University (CA)
Publication Details
- Journal
- Discover Informatics
- Published
- 2026-09-16
- DOI
- https://doi.org/10.1007/s44564-026-00018-x
- Primary Topic
- UAV Applications and Optimization
- Type
- article
- Field-Weighted Citation Impact
- 0.00