Teaching Through Simulation: Virtual Children for Helpline Counsellor Training
Using Cognitive Models and LLMs to Simulate Safe Counselling Practice
This event has passed
Description
This talk explores the development of AI-based simulations designed to help people practise sensitive conversations. It focuses on training children’s helpline counsellors, where trainees interact with Lilobot, a virtual child who reaches out through chat.
The talk will explain the need for this type of training, how Lilobot models a child’s beliefs, desires, and intentions using a rule-based system, and how the approach was evaluated to ensure it effectively supports learning.
The challenges faced when integrating AI and LLMs into simulation-based training will also be discussed. Particularly, balancing realistic, flexible conversations with the reliability, consistency, and control required in sensitive contexts.
Lecturer Bio
Mo(hammed) is a PhD candidate in the Interactive Intelligence Group at TU Delft, supervised by Prof. Willem-Paul Brinkman and Dr. Myrthe Tielman.
He obtained his MSc in Computer Science from TU Delft in 2020, and his Bachelor's degree in Software Engineering (cum laude) from King Saud University (KSU) in 2016. He also completed an internship at the IBM IoT Lab in Durham, North Carolina (2016).
Mohammed's research lies at the intersection of Human-Computer Interaction, Behaviour Change Technologies, Conversational Agents, Educational Systems, and Social Simulation. He focuses on developing interactive training tools that simulate social interactions to support the development of social skills.
His current project involves designing and evaluating a training system for child helpline counsellors. This includes the development of a cognitive-based chatbot that simulates a virtual child, as well as feedback and guidance mechanisms to support skill acquisition within the system.
Notes
Free lunch will be provided for attendees.