AI Hostage Negotiation Simulator

https://tomreilly004-negotiatorai-app-sklmjo.streamlit.app/ - Click here to try save the hostages and descalate the situation

Under Development - currently working on turing into a 3D Experience using Unreal Engine 5

Vision

A narrative-driven hostage negotiation simulator where players use their own voice to negotiate with AI-controlled suspects. Every playthrough is unique because conversations, emotions, and outcomes evolve dynamically.

Core Gameplay Loop

Receive briefing → Assess situation → Build rapport → Gather intelligence → Influence suspect decisions → Resolve peacefully or through escalation → Review performance.

Key Features

Natural voice conversations, dynamic AI personalities, hidden emotional simulation, branching outcomes, replayable scenarios, performance analysis.

AI System

Three layers: (1) Game State stores facts and events. (2) Decision Layer updates trust, stress, anger, empathy and chooses actions. (3) Dialogue AI expresses those decisions naturally while staying in character.

Suspect Design

Each suspect has a biography, motivation, fears, goals, relationships, mental state, negotiation style, and memory of previous dialogue.

Hostages

Hostages have personalities, injuries, panic levels and can secretly provide intelligence if allowed to speak.

Police Command

Updates from SWAT, intelligence, family members, media and medical staff create changing priorities.

Scenario Ideas

Bank robbery, domestic hostage situation, school lockdown, political extremist, failed kidnapping, desperate parent, prison escape.

Progression

Unlock harder cases, new negotiation tools, psychological techniques, and increasingly complex suspects.

UI

Radio feed, conversation transcript, subtle emotional indicators, mission timer, notebook for gathered intelligence.

Audio

Speech-to-text for player, text-to-speech or recorded AI voices for suspects, ambient police radio.

Endings

Peaceful surrender, hostages released, SWAT assault, suspect suicide, suspect escape, partial success, total failure.

Mission Report

Grades rapport, empathy, listening, deception detection, hostages saved, casualties, negotiation duration and overall rank.

Technical Stack

Suitable for Unreal Engine or Unity using speech recognition, LLM-based dialogue, TTS, and a structured simulation backend.

Future Ideas

Co-op negotiations, online leaderboards, procedural scenarios, campaign mode, VR support, mod tools.