Problem
An answer can hide the path that produced it
When sources, time or context conflict, merging everything into a single representation can erase important information.
EMPV / AI LabD
Research / epistemic AI
Trustworthy Reasoning
Research into AI systems that do not automatically compress conflicting evidence into one answer, but preserve history, contradictions, versions and the reasons behind conclusions.
PhasePhase 0 — Call decomposition and research governance. No novelty claim is considered validated.
Direction
Reasoning that carries its evidence
The research explores versioned graphs, epistemic branching, temporal contradiction, belief revision and planning that seeks new evidence when what is available is insufficient.
01 / The question
How should a system reason when reality is not consistent?
The aim is to treat disagreement and change as part of the problem, rather than noise to be removed before reasoning starts.
02 / The research
Evidence graphs, time and belief revision.
The work combines versioned knowledge, temporality, contextual contradictions, proof-carrying merge, abstraction and evidence-seeking planning.
03 / Why it stays in the Lab
This is a scientific direction, not a feature to sell.
Before product claims come novelty, benchmarks and falsification. Psychology and education are possible validation domains, not automatic proof of the project.
EMPV / NOTE
The project is in an early research and scientific-governance phase. No breakthrough or novelty claim is presented as already validated.