# Distinguish learning from feedback artifacts and decoder fitting

Bounded experimental precedent / proposed causal test. Closed-loop neural-culture experiments motivate tests of adaptation under feedback. DishBrain reported results in an embodied game simulation; its title’s use of sentience is the authors’ terminology, not proof of subjective experience. Likewise, improved prediction by a digital readout does not alone demonstrate persistent biological learning.

Canonical page: https://givegigs.com/wiki/necro-chapter-xdpu-closed-loop-learning
Last updated: 2026-11-09T18:58:12.584Z
Parent topic: [necromancer-neuroscience](https://givegigs.com/wiki/necromancer-neuroscience)

## Contents
- [Overview](#wiki-section-overview-1)
- [Chapter notes](#wiki-section-chapter-notes-2)
- [Dossier section](#wiki-section-dossier-link-3)

<a id="wiki-section-overview-1"></a>
## Overview

Bounded experimental precedent / proposed causal test. This page expands the “Distinguish learning from feedback artifacts and decoder fitting” chapter in the neuroscience collection.

<a id="wiki-section-chapter-notes-2"></a>
## Chapter notes

Closed-loop neural-culture experiments motivate tests of adaptation under feedback. DishBrain reported results in an embodied game simulation; its title’s use of sentience is the authors’ terminology, not proof of subjective experience. Likewise, improved prediction by a digital readout does not alone demonstrate persistent biological learning.

Necromancer proposes separating contingent feedback, matched noncontingent replay, baseline drift, and decoder-only learning. Prespecify the behavioral proxy, retention interval, stimulation exposure, artifact rejection, and analysis window. Separate biological preparations, not just correlated time windows, across evaluation splits. Record how much of any gain disappears when the trained decoder or the feedback contingency is removed.

Proposed gate: a reproducible feedback-specific effect that survives controls for exposure, preprocessing, decoder flexibility, and between-batch variation. Recovery of an experimentally induced response remains a circuit-level result. It does not establish autobiographical memory or personal identity. A useful near-term tool would detect reproducible functional change in an injury model before attempting broader interpretations.

<a id="wiki-section-dossier-link-3"></a>
## Dossier section

[Open this exact section in the Necromancer dossier](/necromancer#xdpu-closed-loop-learning). The original chapter status is “Bounded experimental precedent / proposed causal test.”

## Sources

- [In vitro neurons learn and exhibit sentience when embodied in a simulated game-world](https://pubmed.ncbi.nlm.nih.gov/36228614/)
- [Brain organoid reservoir computing for artificial intelligence](https://www.nature.com/articles/s41928-023-01069-w)

---
This article is collaboratively maintained. Check the canonical page and cited sources for the latest context.