The drone sees everything. Dispatch sees nothing.

Drone footage exists. Dispatchers can't see it.

Police drones capture aerial intelligence during active incidents, but that footage lives in a separate system, operated by a separate person, on a separate screen. Dispatchers coordinate response across radio, CAD systems, and incident notes while mentally tracking priority and toggling between windows.

OVERWATCH synthesizes live drone feeds, officer positions, and incident data into one interface that matches how dispatchers actually think during incidents. I led research and design for this thesis project across seven Georgia law enforcement agencies.

Research at a Glance

Fragmented Systems, Fragmented Awareness

Users: 911 dispatchers, patrol officers, drone pilots, shift supervisors.

What's at stake: Seconds matter. A dispatcher juggling four systems during an armed-person call can miss a location update or send units to the wrong entrance. Fragmented information creates cognitive load that compounds under stress.

The fragmentation problem: Dispatchers today use CAD for call queue, a separate map for officer positions, radio for real-time updates, and have no direct access to drone feeds. Drone pilots relay what they see verbally. If the dispatcher mishears "northeast corner" as "east entrance," officers approach blind.

This isn't about adding a feature. It's about consolidating existing information into one interface that matches how dispatchers actually think during incidents.

Field Research Across Seven Agencies

30+ interviews with personnel across Atlanta PD, Georgia Tech PD, and Georgia State Patrol. 17 ride-along observations to see dispatch and field operations in real time.

Key Insights

Research Artifacts

Empathy Mapping

Synthesized officer and dispatcher perspectives into a unified empathy map, capturing what users say, think, do, and feel during high-stress incidents.

Personas

Four personas representing the spectrum of patrol officer experience, technology comfort, and operational context. These informed design decisions around information density, voice interaction, and threat assessment features.

OVERWATCH: Three Columns, One View

A three-column dashboard designed for high-stress, time-critical scenarios. Each panel serves a distinct operational purpose, and placement reflects how officers naturally scan during incidents.

Select any dashboard region to connect the full system view to its component-level decisions.

Left column

Calls for Service

A priority-sorted queue keeps active and pending incidents visible without pulling the officer away from the current call.

Operational roleScan the queue, select a call, and retain awareness of competing incidents.

Design decisionSeverity stripes, compact identifiers, and a persistent position make priority readable at a glance.

Supporting Workflows

Two supporting moments sit outside the persistent dashboard: beginning a traceable session and completing focused, multi-step tasks.

Operator identification

A single required field reduces shift-start friction while tying later actions to a named officer for audit and review.

Focused task dialogs

New calls, handoffs, and incident comparisons open in a focused layer so complex tasks do not compete with the live dashboard.

Design System

A dark, utilitarian interface system built for high-stakes dispatch environments. Every decision optimizes for rapid threat scanning, night-shift legibility, and zero ambiguity under stress. No decorative elements, no sci-fi chrome.

Key Design Decisions

Information hierarchy: The primary challenge was determining what information deserves screen real estate. Through testing with dispatch staff, I established: calls and incident (left), activity and suspect (center), map and drone (right). This matches the dispatcher's mental model: "What's happening? Who's involved? Where is everyone?"

Alert escalation: Three-tier system: Priority 1 (red) for immediate threats, Priority 2 (amber) for developing situations, Priority 3 (blue) for routine calls. The system suggests drone deployment for Priority 1 incidents, but dispatchers can override. The threat score is algorithmic but not authoritative.

Speak their language: The interface uses 10-codes (10-71, not "Armed Person Alert") because dispatchers are fluent. Translating to plain English would slow them down and feel patronizing.

Validating with Real Officers

I conducted mixed-methods evaluation combining expert-based and user-based approaches: heuristic evaluation, cognitive walkthrough, and task-based usability testing with 10 sworn patrol officers from Georgia Tech Police Department.

Methodology

Quantitative Results

Per-Officer SUS Breakdown

Individual scores ranged from 73 to 82, with officers consistently describing OVERWATCH as "the best CAD interface I've seen."

What Officers Said

NASA TLX Workload Profile

Officers reported moderate overall workload, with mental and temporal demands inherent to patrol work. OVERWATCH shifted cognitive load away from reconstructing events toward monitoring and decision-making.

Validation & Impact

This is thesis research, not a shipped product. Honest assessment of where it stands:

Prototype testing: 10 sworn officers completed task-based usability sessions. Mean SUS score of 78 indicates good usability, with officers consistently rating OVERWATCH above their current CAD tools.

Agency interest: Georgia Tech Police Department expressed interest in continued collaboration for pilot testing. The research contributes to HCI literature on emergency response interface design.

What I can't claim: Hard metrics on response time or error reduction. Those require deployment and measurement that thesis timelines don't allow.

What I Learned

• Users don't want new features. They want fewer systems.

  The instinct is to add capabilities. The insight was that consolidation creates more value than innovation. Dispatchers didn't need a new tool. They needed their existing tools to talk to each other.

• Domain fluency is non-negotiable.

  I couldn't design for dispatchers until I understood 10-codes, CAD workflows, and the rhythm of a shift. The ride-alongs taught me things interviews never could. You have to be in the room when it's happening.

• Every interaction must be immediately clear under stress.

  There's no room for "figure it out" in emergency response. This discipline now shapes how I approach all interface work: if it requires a second look, it's wrong.

What I'd Do Next

• Run A/B testing on information hierarchy to validate the three-column layout against alternatives
• Integrate with actual CAD and AVL systems to move from prototype to pilot
• Expand research to 911 call-takers, who sit upstream of dispatchers and have different information needs