Bayesian Belief Engine
Visit →Background
This started as a theology and philosophy project. I wanted to understand how beliefs relate to each other: which claims imply others, where worldviews diverge, and why people can hold contradictions without noticing them.
Most belief tests classify people directly. Bayesian Belief Engine tries to model the structure underneath the answers.
Approach
The app updates a probabilistic model as the user answers questions. Each answer changes the likelihood of connected philosophical and theological positions, so the system learns from the pattern rather than treating each response in isolation.
The output is a belief map. It highlights alignment, tension, and positions that are strongly implied by the user's answers.
What I learned
The inference was only as good as the graph. Deciding which positions were adjacent, orthogonal, or logically dependent took longer than the interface.
The design challenge was tone. Showing someone a contradiction in their worldview is sensitive. The product had to surface tension without treating the user as a problem to fix.