What is it about?
The core idea behind g-distance is to assess how well a formal psychological model (a mathematically defined theory implemented as a computer algorithm to simulate human behaviour) truly captures the range of human behaviours it aims to model.
Featured Image
Why is it important?
Traditional methods often evaluate models based on how closely their behaviour matches a single, often averaged, set of real-world results – known as a "goodness-of-fit" approach. However, this overlooks the fact that both human and model behaviour can be heterogeneous (meaning there can be different subgroups of people behaving differently, and models can produce a range of outputs depending on their parameters). g-Distance addresses this by proposing that a model's adequacy should be measured by how similar the range of behaviours it can produce is to the range of behaviours observed in humans. This involves comparing the full spectrum of a model's capabilities across its "parameter space" (where different parameter settings result in different outputs).
Perspectives
I think it a novel way to look at both how we conceptualise behaviour and how we evaluate our explanations and theories of it. We are often not concerned with how many different consequences our explanations/theories entail (how many things our theory says should happen but does not). I think this requires a shift in thinking by moving from trying to get the best numerical approximation. g-Distance represents this shift from viewing model adequacy as an optimization problem (finding the best fit) to an estimation problem of the overlap between human and model behaviours
Lenard Dome
Read the Original
This page is a summary of: g-Distance: On the comparison of model and human heterogeneity., Psychological Review, April 2025, American Psychological Association (APA),
DOI: 10.1037/rev0000550.
You can read the full text:
Contributors
The following have contributed to this page
