Bayesian Methods in Interaction Design (Tutorial)

Bayesian Methods in Interaction Design

This full day course introduces a Bayesian perspective on interaction and design in HCI.

Bayesian methods offer a powerful approach for thinking about and implementing interactive systems that can deal with uncertainty and noise. This course introduces the theory and practice of computational Bayesian interaction, covering inference of intention and design of interface features.

Participants will explore interactive examples in the tutorial, which is built around hands-on Python programming with modern computational tools. This is interleaved with presentations covering theory and more in-depth examples of problems of wide interest in human-computer interaction.

Summary & objectives

The course will cover:

Bayesian optimisation: solving design problems by optimising over interface configurations with methods that can cope with noisy, slow user measurements.
Inference for interaction: a principled and robust approach to designing a transformation from input to useful action.

This course will:

demonstrate how a Bayesian approach can be used to represent interaction and design problems in a fresh way, with a rigorous mathematical basis;
extend researchers' capabilities to build robust interfaces and develop optimised designs across a wide range of contexts and in the presence of uncertainty and noise;
demonstrate how interaction can be represented as a problem of closed-loop probabilistic inference;
show how powerful Bayesian optimisation can be in the class of problems that HCI needs: optimisation for small, noisy datasets of human observations.

Pre-requisites

To get the most out of this course, participants should have working knowledge of undergraduate mathematics, particularly basic probabilty and linear algebra (even if it is a vague and rusty memory!).

Software and preparation

All of the materials will be delivered as interactive Jupyter notebooks, which interleave notes and live coding examples. All coding examples will be in Python, and familiarity with Python is strongly recommended. We also provide the materials for download and use on your own machine before and after the conference course sessions.

Citing

To cite this course, use the following BibTeX entry:

                      @inproceedings{bayesianinteraction_iui2020,
                                      author = {Per Ola Kristensson and 
                                                Antti Oulasvirta and
                                                John Williamson},
                                      title = {Bayesian methods Interaction Design},
                                      booktitle = {ACM IUI 2020 Tutorials},
                                      year = {2020}}