Schedule
Location
CHI 2026, Barcelona, room to be assigned, check the official program.
Format
A single 90 minute session focused on Human in the Loop Bayesian Optimization for HCI design.
Phase details
Introduction and Motivation
Why optimization helps HCI, core ideas of Bayesian Optimization, exploration and exploitation, surrogate models and acquisition functions, shared vocabulary to ground the session.
Live Demo, Foundations of BO
Step by step notebook, fit a surrogate, update an acquisition function, propose designs, Unity backend connected to Python to show real time parameter updates.
Small Group Use Case Mapping
Teams identify design parameters, objectives, constraints, and sketch a Human in the Loop workflow for their own problem.
Break
Short pause.
Panel Discussion, Multi Objective BO
Methods for noisy feedback, high dimensional spaces, fairness across user groups, reading Pareto fronts, hypervolume, and knee points.
Pipeline Sketching and Presentations
Teams draft pipelines with inputs, feedback loops, optimizer updates, stopping rules, then present for quick feedback.
Wrap Up and Next Steps
Synthesis, how to continue, where materials live.
Organizers
Pascal Jansen, Institute of Media Informatics, Ulm University, research on multi objective Bayesian Optimization in HCI, automated vehicles and mixed reality.
Mark Colley, UCL Interaction Centre, UCL, research on accessibility, mobility, and optimization for interface design.
Prerequisites
Background in HCI design or studies. Optional, a laptop with Python 3.13 or later and BoTorch v0.15.1 or later, optional Unity 6.2 for the integration demo.
Accessibility
Real time captioning requested, tagged PDFs with alt text and high contrast, short captioned demo videos, low barrier group templates and cloud notebooks.
Materials
GitHub repository with annotated notebooks, Unity project for real time parameter updates, example datasets and evaluation templates, slide decks, curated references, quick start guides, captioned demo videos, a short workshop summary.