Embedded Files
My name is Liqiang Huang, and I work in the Department of Psychology at The Chinese University of Hong Kong. You can reach me at lqhuang@cuhk.edu.hk.
My research is ultimately based on the concept of the PBS (Precision–Breadth–Simplicity) impossible trinity. So, what exactly does the PBS impossible trinity mean? Specifically, it entails that precise and simple experimental findings are inevitably fragmented, while broad and simple theoretical notions are inevitably ambiguous.
Already familiar with this? Sure—but the key point is that these reflect a fundamental limit. This is not a problem that can be solved through self-disciplined hard work, nor one that will improve over time. Instead, it calls for a method fundamentally different from the traditional theory-driven approach—what we call the Comprehensive Exploration (CE) approach.
In the CE approach, we use experimental benchmarks on a massive scale—for example, a single experiment involving 30,000 hours of data—to cover a broad range of phenomena. We then build models that are as precise as possible while remaining only moderately complex. This provides the practical best solution to the PBS impossible trinity.
Below on this page, you will find the four steps of the CE approach. Details of this approach are described in this paper and further explained on this YouTube channel. A list of published and ongoing projects using CE is available here.
I came up with this approach through a long and painful 12-year struggle. If you're interested in the story, you can take a look at this blog or watch a video.
Four steps of CE approach
Four steps of CE approach
1. Design: Stimulus-driven design
1. Design: Stimulus-driven design
Experimental psychology typically tests predefined hypotheses, which can limit discovery to anticipated possibilities. Stimulus-driven design moves beyond this constraint. In ImageNet, modeling millions of varied images can reveal object-category regularities not known in advance. Similarly, a recent CE study of visual working memory (Huang, 2025) modeled performance across randomly generated color patterns without prespecifying mechanisms, recovering all classic mechanisms examined and identifying several novel ones.
2. Data: Large-scale experimentation
2. Data: Large-scale experimentation
As in AI, this stage establishes a robust empirical foundation. For instance, Huang (2025) collected 40 million responses—33,000 hours of data—across 10,000 color patterns. This scale provides the basis for a comprehensive exploration of the phenomena of interest.
3. Analysis: Iterative model development
3. Analysis: Iterative model development
This is the most labor-intensive stage of the CE approach. The general strategy is to draw on both theoretical insights from the literature and patterns observed in the data to explore candidate model components and alternative ways of assembling them into a complete model.
4. Outcome: An interpretable computational model
4. Outcome: An interpretable computational model
A defining characteristic of this outcome is the deliberate balance between predictive accuracy and interpretive parsimony. Neural networks often achieve high predictive accuracy at the expense of interpretability, whereas traditional psychological theories are interpretable but usually lack comprehensive predictive power. CE models seek to combine these strengths. For example, the model in Huang (2025), containing only 57 parameters, outperforms a complex neural network with 30,796 parameters. The CE approach thus aims to deliver a relatively simple, interpretable theory with broad explanatory and predictive power.
Page updated
Google Sites
Report abuse