Course Objective & Scope
A voice or visual agent that performs well on clean lab inputs can quickly lose trust when confronted with noisy environments, half-finished sentences, unexpected screen states or adversarial prompts from real users. This programme teaches how to test interactive agents across usability, calibrated user trust and safety boundaries together.
Participants design user studies that measure over-reliance and under-trust, run multimodal red-teaming scenarios across voice, vision and UI channels, and produce an experience, trust and quality measurement plan for an organisational case.
Who Should Attend?
- AI product managers, UX researchers and conversation/interaction designers
- AI quality assurance (QA), testing and AI safety teams
- Customer experience and digital channel leads
- Prerequisite: basic familiarity with product testing or user research
Key Takeaways
- Define trust calibration and cognitive load metrics beyond simple task completion
- Stress-test agents with noisy audio, blurry images, diverse accents and interrupted turns
- Run red-teaming tests against visual/audio prompt injection and social engineering
- Automatically detect user frustration, repeated reformulations and early abandonment signals
- Establish go/no-go experience and safety acceptance criteria for production launch
Curriculum
Trust Calibration in Interactive Agents
- Balancing over-reliance against algorithm aversion after a single mistake
- How uncertainty cues and evidence display shape user decisions
Real-World Conditions and Edge-Case Stress Testing
- Background noise, varied speaking rates, low-quality camera feeds and partial documents
- Multi-turn scenarios where users change their mind or digress mid-task
Multimodal Safety and Red-Teaming Tests
- Instructions hidden in images, spoken jailbreaks and social engineering for unauthorised actions
- Testing for sensitive data leakage, policy bypass and brand risk
Behavioural Signals and Friction Analysis
- Repeated reformulations, barge-ins, vocal stress and manual takeovers
- Reviewing session replays and building a failure taxonomy
Go/No-Go Acceptance Criteria and Measurement Plan
- Dogfooding and staged beta testing protocols
- Launch readiness checklist covering quality, latency and safety thresholds
Other Courses in This Category
-
Multimodal model architectures that unify vision, audio and text representations: ViT/SigLIP and audio encoders, cross-modal projection/fusion layers, and LoRA fine-tuning of open-weight VLMs (Qwen-VL, Llama, Pixtral).
- Live Course
- 5 modules
-
Building visual inspection and analysis applications with Vision-Language Models (VLMs) in manufacturing, retail, insurance and field operations: visual grounding, multi-image comparison and hybrid CV+VLM pipelines.
- Live Course
- 5 modules
-
End-to-end Speech-to-Speech (S2S) models and autonomous voice agents: acoustic emotion and prosody perception, voice tool calling, multi-agent call handoffs and production architecture with Pipecat and LiveKit.
- Live Course
- 5 modules
-
Streaming infrastructure engineering for real-time AI applications: WebRTC, WebSockets and SSE protocols, OpenAI Realtime and Gemini Live API integration, audio/video buffer management and horizontal scaling.
- Live Course
- 5 modules
-
Human-like turn-taking and millisecond latency engineering in voice AI: acoustic VAD, semantic endpointing, false-interruption suppression, TTFT/TTFB optimisation and speculative execution.
- Live Course
- 5 modules
-
Parsing dense documents and application screens (GUIs) at the pixel, coordinate and structural-tree level: OmniParser, Qwen-VL, Docling, UI widget detection and unified document/screen understanding models.
- Live Course
- 5 modules
Get in touch
