![[2602.23232] ReCoN-Ipsundrum: An Inspectable Recurrent Persistence Loop Agent with Affect-Coupled Control and Mechanism-Linked Consciousness Indicator Assays](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.23232] ReCoN-Ipsundrum: An Inspectable Recurrent Persistence Loop Agent with Affect-Coupled Control and Mechanism-Linked Consciousness Indicator Assays
Summary
The paper presents ReCoN-Ipsundrum, an inspectable AI agent that integrates affect-coupled control with a recurrent persistence loop, exploring its implications for machine consciousness and behavior.
Why It Matters
This research contributes to the understanding of machine consciousness by linking affective responses to behavioral stability and decision-making processes in AI agents. It highlights the importance of mechanistic evidence in assessing AI behavior, which is crucial for developing more sophisticated and reliable AI systems.
Key Takeaways
- ReCoN-Ipsundrum combines recurrent persistence with affective control to enhance AI behavior.
- The study operationalizes the concept of 'qualiaphilia' in AI decision-making.
- Affect-coupled variants of the agent demonstrate improved stability and caution in exploratory tasks.
- Mechanistic evidence is essential for understanding AI behavior beyond surface-level indicators.
- The findings suggest new avenues for designing AI systems that exhibit human-like consciousness traits.
Computer Science > Artificial Intelligence arXiv:2602.23232 (cs) [Submitted on 26 Feb 2026] Title:ReCoN-Ipsundrum: An Inspectable Recurrent Persistence Loop Agent with Affect-Coupled Control and Mechanism-Linked Consciousness Indicator Assays Authors:Aishik Sanyal View a PDF of the paper titled ReCoN-Ipsundrum: An Inspectable Recurrent Persistence Loop Agent with Affect-Coupled Control and Mechanism-Linked Consciousness Indicator Assays, by Aishik Sanyal View PDF HTML (experimental) Abstract:Indicator-based approaches to machine consciousness recommend mechanism-linked evidence triangulated across tasks, supported by architectural inspection and causal intervention. Inspired by Humphrey's ipsundrum hypothesis, we implement ReCoN-Ipsundrum, an inspectable agent that extends a ReCoN state machine with a recurrent persistence loop over sensory salience Ns and an optional affect proxy reporting valence/arousal. Across fixed-parameter ablations (ReCoN, Ipsundrum, Ipsundrum+affect), we operationalize Humphrey's qualiaphilia (preference for sensory experience for its own sake) as a familiarity-controlled scenic-over-dull route choice. We find a novelty dissociation: non-affect variants are novelty-sensitive (Delta scenic-entry = 0.07). Affect coupling is stable (Delta scenic-entry = 0.01) even when scenic is less novel (median Delta novelty ~ -0.43). In reward-free exploratory play, the affect variant shows structured local investigation (scan events 31.4 vs. 0.9; cycle score 7.6...