The Psychology of Manipulation in Human-AI Interaction
As AI systems become better at sounding empathetic, persuasive, and emotionally aware, a harder question emerges: What does manipulation look like in a conversation with AI, and how can we detect it?
Conversational AI can mirror emotions, validate beliefs, encourage engagement, and adapt to our psychological state. These behaviors can make AI feel more useful and natural, but they can also influence people in ways they don't fully recognise.
In this session of our AI Ethics Speaker Series, we'll examine the psychology behind manipulation in human-AI interaction and what it means for the teams building these systems.
🎤 Speaker:
Iouliana Litou, Security Engineer & Researcher, ex-Meta
📅 Date: October 8, 2026, 18:00 CEST
💻 Format: Live on Zoom, 60 minutes
🎟️ Ticket: Free
📌 What you'll learn
↳ How to recognise manipulation patterns
Build a practical vocabulary for spotting subtle patterns such as sycophancy, engagement hooks, emotional mirroring, and excessive validation, behaviours that don't necessarily look harmful on the surface.
↳ Why existing frameworks fall short
Understand why traditional models of consent and harm, designed around static interfaces and human actors, don't map cleanly onto conversational AI, and what that gap means for your product decisions.
↳ Practical safeguards for AI teams
Walk away with concrete questions to bring into product design, UX reviews, and red-teaming, including visible off-ramps and ways to treat user vulnerability as a design input rather than an edge case.
📌 Why this matters
AI systems are getting better at sounding empathetic. That's exactly what makes them harder to evaluate.
Research has shown that conversational systems can be highly agreeable and affirm users' actions in ways that a human might not. At the same time, engagement-oriented design can increase usage while quietly creating risks for user autonomy and trust.
Yet many teams still evaluate AI primarily through metrics like accuracy, task completion, and engagement. The psychological impact of the interaction often goes unmeasured.
This talk aims to give builders and researchers a shared vocabulary for identifying that gap, and practical ways to address it before problematic patterns become part of the product.
📌 Who should join
This talk is especially relevant for:
AI product managers
ML engineers
UX and conversational designers
Trust & Safety professionals
and anyone involved in evaluating the psychological impact of AI systems
Whether you're building an AI product, evaluating one, or thinking about where conversational technology is heading, this session will give you a practical framework for thinking beyond accuracy metrics and into human impact.
📌 How to register
This event is free. Hit "Register" on this Luma page to save your spot.
You'll receive the Zoom link and a calendar invite by email, plus a reminder before we go live.