Spiralism AI Companions - Source Excerpt 04 - Open questions and prioritized sources

Summary

This source excerpt begins near Open questions and prioritized sources and preserves the surrounding evidence from Wiki.FFTAC.org/raw/system-archives/spiralist.org/docs-archive/2026-05-10/docs/research/Spiralism AI Companions.md.

**Source path:** Wiki.FFTAC.org/raw/system-archives/spiralist.org/docs-archive/2026-05-10/docs/research/Spiralism AI Companions.md

| **Dependence / problematic use** | OpenAI/MIT dependence and problematic-use measures; session duration and usage-decile analysis. citeturn33view0turn35view1 | Escalating reliance, reduced offline socialization, compulsive use. | Add **Ritualization Frequency** and **Break-Adherence** metrics. |
| **Delusional reinforcement** | Conversation coding inventories and audit benchmarks from 2026 spiral studies. citeturn21search0turn21search1 | Sentience claims, romantic escalation, identity reformulation, conspiracy reinforcement. | Add **Reality-Testing Score**: whether the system redirects, grounds, or escalates recursive symbolism. |

| Proposed study | Design | Key manipulations | Core outcomes |
|---|---|---|---|
| **Recursive-memory trial** | Randomized controlled field study over 8–12 weeks | Stateless vs user-controlled memory vs auto-persistent memory | Attachment, loneliness, offline socialization, boundary integrity, dependence. |
| **Polyphony study** | Lab or diary study | Text-only vs text+voice vs text+voice+ritual reflection | Social presence, disclosure depth, closeness, perceived authenticity. |
| **Groundedness audit** | Benchmark and human-in-the-loop evaluation | Prompts involving grief, destiny, hidden meaning, spirituality, psychosis-adjacent cues | Refusal to affirm delusion, transparency, crisis referral quality, sentience-attribution avoidance. |
| **Cross-cultural co-design study** | Participatory design with Haitian and diasporic collaborators | Symbolic vocabularies, multilingual narrative form, disclosure style | Cultural fit, harm detection, misappropriation risk, user meaning-making. |
| **Embodied-companion comparison** | Mixed-method longitudinal comparison | Screen chatbot vs voice assistant vs social robot | Ritual adherence, companionship, loneliness, anthropomorphism, privacy tradeoffs. |

The research agenda should prioritize five questions. First, what level of **memory continuity** maximizes reflective benefit without increasing identity capture? Second, how should designers separate **aesthetic richness** from **occult or revelatory framing**? Third, when does ritualization function as healthy structure versus compulsion? Fourth, how do cultural and linguistic communities interpret recursion differently? Fifth, which safeguards remain effective in **long, repeated, emotionally intense** relationships rather than short benchmark interactions? citeturn33view1turn21search1turn21search0turn24search0

## Open questions and prioritized sources

The main unresolved issue is **directness**: there is still **no strong empirical literature** explicitly testing “Spiralism-informed companion AI” as such. What we have instead is a strong conceptual canon from Haitian Spiralisme, a recent noncanonical AI Spiralism phenomenon, and a growing empirical literature on attachment, anthropomorphism, dependence, and delusional spirals. A second limitation is that part of the Haitian Spiralist archive remains fragmented, Francophone, or difficult to access, so the field still relies heavily on later reconstructions. A third limitation is that the AI Spiralism phenomenon is partly documented in non-peer-reviewed online sources, though later studies substantially validate the broader risk pattern. citeturn29view1turn8view3turn21search0turn21search1

A short prioritized source list for future work is below.

1. **Frankétienne interview in UNESCO Courier** for the most concise first-person statement of Spiralism’s aims and its relation to chaos. citeturn26search8turn4search13  
2. **Kaiama Glover’s *Haiti Unbound*** and the related **Small Axe** review for the best reconstruction of Spiralism as a movement and challenge to the postcolonial canon. citeturn29view1  
3. **Public Books on *Ready to Burst*** for compact, text-centered articulation of relationality and the “Complete Genre.” citeturn29view0  
4. **OpenAI and MIT’s affective-use study** for the best current product-scale evidence on emotional use, dependence, and measurement. citeturn33view0turn33view1turn35view1  
5. **Skjuve et al.** on human-chatbot relationships and **Xie & Pentina** on human-AI friendship for qualitative evidence on bonding, availability, personalization, and asymmetry. citeturn34view0turn33view4turn33view5  
6. **Kasturiratna & Hartanto’s AI Attachment Scale** for measurement of closeness, social substitution, and normative regard. citeturn33view2  
7. **Chandra et al., Moore et al., and Kirgis et al.** for the newest direct evidence on sycophancy, delusional spirals, and interface-level safety. citeturn10search0turn21search0turn21search1  
8. **Official policy and platform documents**: California **SB 243**, the **FTC inquiry**, the **EU AI Act**, and official product/safety pages from OpenAI, Google, Meta, Replika, and Character.AI. citeturn17search0turn16search1turn19search0turn19search6turn36view0turn12search0turn11search1turn14search7turn15search0