Cognition & Design

James Derek Lomas

Design Fellow at UC San Diego

Email: dereklomas@gmail.com

Office: Atkinson Hall 1601, La Jolla CA 92093

I am a postdoctoral Design Fellow in the Design Lab at UC San Diego, mentored by Don Norman. I have expertise in designing learning software based on cognitive science. In addition to my academic work, I am the founder of Playpower Labs, which has produced over 35 learning games that have been played by millions of children. Two of those children are mine!

I received my PhD in Human-Computer Interaction from Carnegie Mellon University. I also received an MFA in Visual Art from UC San Diego and a BA in Cognitive Science at Yale University.

Published Research

Past Projects

Scientific Contributions

Large-scale experiments for optimization and basic research

Basic and applied design experiments can be conducted within education products with a large number of users. These experiments can help drive improvements in the desired product outcomes (applied) and can also test generalizable theories of learning or motivation (basic). In a series of experiments involving over 100,000 subjects, we demonstrate various optimization methods and also produce evidence regarding the effects of difficulty, novelty and suspense on player motivation. This evidence provides the first controlled results showing that moderate levels of difficulty (probability of task failure) do not improve player motivation. In contrast, novelty and suspense are other factors of “challenge” that may account for its motivational nature. Rather than “Not too hard, not too easy”, we suggest the maxim should be “Not too hard, not too boring.”

Lomas, D., Forlizzi, J., Patel, N., Shodhan, S., Patel, K., Poonwala, N., Koedinger, K., (2017) Is Difficulty Overrated? Investigating the Effects of Choice, Novelty and Suspense on Intrinsic Motivation. ACM CHI 2017
Lomas, D., Forlizzi, J., Koedinger, K., (2013) Optimizing Challenge in an Educational Game Using Large- Scale Design Experiments. ACM CHI 2013

AI-Assisted Experimentation

Running large numbers of randomized controlled experiments in order to optimize the design of an educational game is inefficient. Instead of a random search through the design space, machine learning approaches offer ways of optimizing the experimental conditions that are tested. Using a multi-armed bandit approach, we showed how AI-assisted experimentation could more rapidly identify effective design conditions. Furthermore, these experiments reduce the likelihood that participants will be placed into low-performing design conditions.

Lomas, D., Forlizzi, J., Poonwala, N., Patel, N., Shodhan, S., Patel, K., Koedinger, K., Brunskill, E. Interaction Design as a Multi-Armed Bandit Problem. (2016) ACM CHI 2016

Google Sites

Report abuse