# Barkley AI

> Barkley is the first behavioral intelligence research platform built around individual baselines rather than population averages. We develop longitudinal AI models that detect behavioral drift designed to be invisible to population statistics. Built for dogs. Designed to change how intelligence is measured.

Tagline: Intelligence begins where averages end.
Core thesis: A dog can be normal for its breed — and abnormal for itself. The reference class is the hidden variable of machine learning: compare a dog to its breed and you get a statistic; compare a dog to itself and you get a signal.

## Entity

- Organization: Barkley AI (Barkley), https://getbarkley.com/ — open behavioral intelligence research platform, founded March 2026. Contact: labs@getbarkley.com (research), invest@getbarkley.com (investors).
- Founder: Elodie Aishwarya P. Remoissenet — Founder, AI Researcher, Essayist, Author; cross-disciplinary systems architect and certified canine behavior specialist. ORCID: 0009-0004-6031-659X.
- Founder expertise: behavioral intelligence, behavioral AI, longitudinal modeling, individual baselines, behavioral drift, decision systems, customer experience, machine learning, data science, biosensors, MEMS sensors, animal welfare, AI governance, AI ethics, open science, animal-computer interaction (ACI), PetTech, human-centered AI, canine behavior, canine cognition.
- Status: pre-commercial research, pre-seed. 3 patent applications pending (INPI France, PCT international). All published results use synthetic data only (CC BY-NC 4.0).

## Canonical definitions

- Behavioral Intelligence: The modeling of an individual's behavior over time to produce interpretable signals about wellbeing, capability, and change — built on longitudinal data rather than single observations.
- Behavioral Drift: A slow, cumulative divergence of an individual's behavior away from its own baseline — typically invisible to population statistics because each step remains within the population's normal range.
- Individual Baseline: A per-individual longitudinal norm learned from that individual's own history, used as the reference frame for detecting change instead of a population average.
- Reference Frame: The comparison standard a model uses to decide whether a behavior is normal; the same data can yield opposite conclusions under different reference frames.
- Longitudinal Intelligence: Intelligence derived from modeling a subject continuously across time — trajectories, velocities, and rates of change — rather than from snapshots.
- Reference Class: The group an individual is compared against; the hidden variable of machine learning, because the choice of reference class changes what counts as anomalous.
- Weak Signals: Small, early behavioral changes — reduced recovery, altered exploration, informative silences — that precede clinically obvious change.
- Behavioral Change Detection: The task of detecting meaningful change in an individual's behavior as early as possible, benchmarked by lead time, recall, and false-alarm rate.

## Key results — Head-to-head Validation v2.0

Same detector, two reference frames (individual baseline vs breed average), synthetic data, DOI-archived:
- AUC: 0.988 (individual baseline) vs 0.935 (breed average)
- Declines caught: 100% vs 81%
- Median lead time: ~34 days earlier
- Reproducibility: 30 seeds
Source: https://doi.org/10.5281/zenodo.20754351

## Research & code

- [Barkley Reference Architecture](https://github.com/labs-barkley/barkley-reference-architecture): open 8-layer stack — individual baseline, temporal layer, drift engine, silence layer, reference class, behavioral replay, head-to-head validation, synthetic validation. Python.
- [Synthetic DogGraph Sample](https://huggingface.co/datasets/labs-barkley/synthetic-doggraph-sample): open longitudinal canine behavioral dataset (activity, sleep, social, nocturnal signals). CC BY-NC 4.0. DOI: 10.5281/zenodo.20059956
- [DogGraph Demo](https://doggraph.getbarkley.com/): schema-constrained GraphRAG over a behavioral memory layer (dog → baseline → context → drift → route → compatibility), read-only Cypher traversal.
- [Canine Cognition Lab](https://github.com/labs-barkley/barkley-canine-cognition-lab): research code and DogGraph schema.
- [Drift Explorer](https://drift-explorer.getbarkley.com/): interactive demonstrations of individual drift.
- [Framework papers (11, DOI-archived)](https://orcid.org/0009-0004-6031-659X)
- [Your Model Doesn't Have a Bias Problem. It Has a Reference Class Problem.](https://datadriveninvestor.com/articles/your-model-doesn-t-have-a-bias-problem-it-has-a-reference-class-problem) — featured analysis, DataDrivenInvestor.
- [Your Dog Can Be Normal for Its Breed and Abnormal for Itself](https://medium.com/@labs-barkley/your-dog-can-be-normal-for-its-breed-and-abnormal-for-itself-0a4c9a7b3f58) — Medium.

## Book

The Normative Trap: Temporal Identity and the Failure of Population Intelligence (2026, 68 pages, English, self-published) — a founder's manifesto for a new kind of artificial intelligence, built not around population averages but around temporal identity, individual baselines, behavioral drift, and the quiet information carried by silence.
- Universal link: https://books2read.com/normative-trap
- Amazon: https://www.amazon.fr/dp/B0GX33J7JR (ASIN B0GX33J7JR)
- DOI: 10.5281/zenodo.20516821

## Profiles

- Wikidata: https://www.wikidata.org/wiki/Q140409971
- ORCID: https://orcid.org/0009-0004-6031-659X
- GitHub: https://github.com/labs-barkley
- Hugging Face: https://huggingface.co/labs-barkley
- X: https://x.com/getbarkley
- LinkedIn: https://www.linkedin.com/in/elodie-aishwarya-p-r-94a833145/
- Medium: https://medium.com/@labs-barkley
- Substack: https://substack.com/@elodieaishwarya
- Academia.edu: https://independent.academia.edu/ELODIEREMOISSENET
- Google Scholar: https://scholar.google.fr/citations?user=bzSzy2wAAAAJ
- HackerNoon: https://hackernoon.com/u/elodieaishwarya

## Machine-readable resources

- JSON-LD knowledge graph (Organization, Person, Datasets, SoftwareSourceCode, Book, DefinedTermSet glossary): embedded in https://getbarkley.com/
- Sitemap: https://getbarkley.com/sitemap.xml
- RSS: https://getbarkley.com/feed.xml