Research note
Public origins, implementation choices, and limits
Provenance belongs beside the result—not in an afterthought.
How the notes are built
Each PersenseAI demo begins from a public research concept or an independently documented question. A prototype then selects a bounded combination that can be represented with reproducible synthetic data. Established concepts remain attributed to their public sources; interface design, scenario structure, and deterministic implementation are presented as choices, not discoveries.
Selected primary and authoritative sources
- Claude E. Shannon, “A Mathematical Theory of Communication” (1948) — entropy and information.
- D. V. Lindley, “On a Measure of the Information Provided by an Experiment” (1956) — expected information from experiments.
- Ronald A. Howard, “Information Value Theory” (1966) — decision-oriented value of information.
- Ruzena Bajcsy, “Active Perception” (1988) — purposive evidence collection.
- Kathryn Chaloner and Isabella Verdinelli, “Bayesian Experimental Design: A Review” (1995) — explicit utility in experiment design.
- David L. Hall and James Llinas, “An Introduction to Multisensor Data Fusion” (1997) — combining multiple sensing sources.
- W3C, “PROV-O: The PROV Ontology” (2013) — interoperable provenance concepts.
- Samuel R. Bowman et al., “A Large Annotated Corpus for Learning Natural Language Inference” (2015) — entailment, contradiction, and neutral stance.
- Pang Wei Koh and Percy Liang, “Understanding Black-box Predictions via Influence Functions” (2017) — tracing influential data.
- Chuan Guo et al., “On Calibration of Modern Neural Networks” (2017) — confidence calibration.
- Amirata Ghorbani and James Zou, “Data Shapley” (2019) — marginal data valuation.
What the demos do not establish
They do not prove truth, causality, calibrated uncertainty, source independence, optimal weights, generalization, forensic validity, legal admissibility, operational reliability, or novelty. They do not identify real people or analyze real cases. Utility means usefulness inside a bounded synthetic explanatory model; it is not legal weight.
Corrections and feedback
If a citation, distinction, or limitation can be improved, please compare notes.
hello@persenseai.com