Research
Questions arising from my work in image analysis, model evaluation and software.
01
When do better image-processing outputs lead to more reliable measurements and useful predictions?
Cell segmentation, quantitative morphology, evaluation groups, domain differences and the relationship between upstream processing and downstream utility.
02
How should a model use context and communicate uncertainty when observable signals do not uniquely determine emotion?
Building from facial-expression work towards questions about dialogue, contextual information, explanations and carefully bounded interpretation.
03
What evidence is needed before a model result can support a decision or a software workflow?
Baseline comparisons, reproducible evaluation, failure analysis, transparent uncertainty and practical interfaces for inspecting results.
04
How can mechanistic assumptions and measured data be connected without confusing model fit with explanation?
An additional area of interest: individual-based models, sensitivity, uncertainty and multiscale reasoning.