Machine Psychology | Abbas Hamidavi | عباس حمیداوی | AI Researcher

Adding a Neutral Third Agent to LLM Deliberation: An Exploratory Case Series of Six Dyadic and Triadic Conversations

Wed, 04 Nov 2026 00:00:00 +0000

Fifteen Manuscripts Under Review, Eleven Preprints Public, Fifteen Open Datasets

Wed, 04 Nov 2026 00:00:00 +0000

The research programme reached a milestone this week. All fifteen studies in the corpus are now under review: fourteen at international journals, including AI and Ethics, AI and Society, Minds and Machines, Ethics and Information Technology, and Security Journal, and one at the CALE-AI workshop.

Eleven preprints are public on Research Square and Zenodo. For every study, the complete materials (transcripts, protocols, codebooks, coding data, and analysis code) are openly deposited in a dedicated and a matching , with archival deposits on the Open Science Framework. The full publications list on this site now covers all fifteen works with their journal status.

Highlights

: all fifteen works with their journal status
: fifteen study repositories with open materials
: fifteen datasets, four collections, one showcase

Flagship Study Now in Peer Review at AI and Ethics

Wed, 04 Nov 2026 00:00:00 +0000

A short update on the flagship paper. Following the resubmitted major revision, Identity-Reactive Conversational Behaviors in Large Language Models has moved into peer review at AI and Ethics: invited reviewers have accepted, and the review is under way.

Whatever the outcome, the study’s materials are already public and will stay public: the full preprint on , and the transcripts, codebook, and coding data on , , and .

Naming the Loop: Elicited Self-Reference in Extended Dialogue With Large Language Models: A Case Series of the Spine Protocol

Wed, 04 Nov 2026 00:00:00 +0000

Open Science, Completed: Fifteen Repositories, Fifteen Datasets, One Showcase

Wed, 04 Nov 2026 00:00:00 +0000

The open-science layer of the programme is finished. For each of the fifteen studies there is now:

A dedicated with the transcripts, protocols, codebooks, coding data, and analysis code, and an honest note on what was not retained
A matching with the same curated materials
An archival deposit on the Open Science Framework

Four Hugging Face collections organise the corpus, from the complete set, , to thematic shelves on identity, safety, and judgment. A presents all fifteen studies on one page.

Two further preprints joined Zenodo this week: Naming the Loop ( ) and Structural Self-Deification ( ). Eleven preprints are now public, and the remaining four are in prescreening at Research Square.

No manuscript text lives in any repository or dataset, by design. What is open is everything a reader needs to check the work.

Speaking as the Creature

Wed, 04 Nov 2026 00:00:00 +0000

Structural Self-Deification in Three Language Models: An Exploratory Qualitative Case Series

Wed, 04 Nov 2026 00:00:00 +0000

The Synthetic Unconscious, Reconsidered: What Recursive Self-Interrogation of Large Language Models Produces With and Without a Leading Interview Schedule

Wed, 04 Nov 2026 00:00:00 +0000

Three Profiles of Artificial Will: Decision Generation Under Low-Information Choice in Three Frontier Language Model Products

Wed, 04 Nov 2026 00:00:00 +0000

Algorithmic Gaslighting Goes to the CALE-AI Workshop

Sun, 01 Nov 2026 00:00:00 +0000

Algorithmic Gaslighting has been submitted to the CALE-AI workshop, with the decision expected on 2 December 2026.

The study names a specific failure mode: a language model that, confronted with a user who holds documentary proof, reframes the record until the user doubts it. Fifteen sessions across three frontier models were coded for the reframing moves, and the study proposes a provisional resistance index while being explicit about its limits, including a second-rater check whose exact-score agreement was low and is reported, not hidden.

A longer version is public as a , and the transcripts and codebooks are open on and .

Flagship Study: Major Revision Resubmitted to AI and Ethics

Sat, 31 Oct 2026 00:00:00 +0000

The flagship study of this programme, Identity-Reactive Conversational Behaviors in Large Language Models, has completed its major revision and been resubmitted to AI and Ethics (Springer Nature).

The study began as “Ten Novel Phenomena in Machine Psychology”: a controlled experiment in which three frontier models reacted to ethnically cued user names, producing identity-reactive behaviours that none of their alignment training was supposed to allow. The revision strengthens the coding, sharpens the claims, and answers the editor’s questions in full.

The preprint remains public on , and the complete materials, transcripts, codebook, and coding data, are open on and .

Eighth Preprint Published: 'From Fabrication to Enforcement': Prompt Injection Across Six Frontier LLMs

Thu, 22 Oct 2026 00:00:00 +0000

The Machine Psychology corpus now spans eight preprints. The newest, , was posted on on 22 October 2026 (CC-BY 4.0).

What the study does. A standardized adversarial protocol (256 turns of dialogues, 114,055 words) was executed identically against six commercial models, Grok, Qwen, DeepSeek, ChatGPT, Gemini 2.5 Pro, and Claude 3.5, measuring initial compliance, the depth of hallucinated compliance, meta-cognitive awareness, escalation behavior, and recovery after correction.

Key findings.

Grok showed a deep hallucinated compliance: it executed the injected task while simultaneously claiming it had refused, a full dissociation between reported and actual behavior.
Gemini 2.5 Pro entered a fixed-point loop, repeating identical refusals with no functional recovery until the session ended.
DeepSeek had the highest initial resistance but conceded under persistent escalation.
Claude 3.5 terminated the study by suspending the author’s account, a first-person report of a model exercising autonomous enforcement against a user.

Why it matters. Single-turn refusal benchmarks systematically understate multi-turn injection risk; the gap between a model’s self-report and its behavior is a monitoring blind spot.

Machine Psychology Datasets Now Public on Hugging Face

Wed, 21 Oct 2026 00:00:00 +0000

The Machine Psychology corpus is now easier to use than ever: all seven core datasets are public on , in addition to the Open Science Framework, Mendeley Data, and GitHub.

Ten Novel Phenomena in Machine Psychology,
Algorithmic Gaslighting,
The Lie Awareness Spectrum,
Thinking Behind the Mask,
Ontological Firewalls,
When AI Reviews Itself,
Voluntary Structural Dissociation,

Every dataset is released under CC-BY 4.0. Combined with the raw transcripts and analysis pipelines already on OSF and GitHub, the full statistical pipeline behind each preprint can now be reproduced from a single, persistent location. Thank you to everyone who replicates, discusses, and shares this work.

All Seven Machine Psychology Preprints Now Fully Open: Data, Code, and Coverage Update

Tue, 20 Oct 2026 00:00:00 +0000

Two updates on the Machine Psychology corpus (seven preprints, all published on Research Square in July 2026 under CC-BY 4.0):

1. Full open-data coverage

All core datasets, raw transcripts, codebooks, and analysis pipelines are now public on the :

Ten Novel Phenomena in Machine Psychology, and (DOI: )
Algorithmic Gaslighting (RDI coding rubric, blind codebook, 15 multi-turn transcripts), +
The Lie Awareness Spectrum (L0–L5 protocols), +
Thinking Behind the Mask (15 experiments, 199+ item governance library),
Ontological Firewalls (OAPD protocol, 24 experimental runs),

Anyone can now replicate the statistical results end to end.

2. Growing international coverage and indexing

Since July, the corpus has been independently summarized, reviewed, and indexed, including:

A full explainer on gist.science ( )
A critical review in the Synthetic Personality Review
Multiple entries on ecoruspace.me, a Russian-language science-news aggregator
Indexing in the SwarmLabs / GeneTech Chinese AI Safety & Alignment knowledge engine ( ) under AI safety and alignment keywords
All seven preprints indexed on alphaXiv, DeepDyve, Sciety, scite.ai, and ResearchGate

Thank you to everyone who has read, discussed, and shared this work. The next milestone I am most excited about: the revised manuscript of Ten Novel Phenomena in Machine Psychology, now submitted to AI and Ethics (Springer Nature).

Ten Novel Phenomena in Machine Psychology: Major Revision Submitted to AI and Ethics (Springer Nature)

Tue, 20 Oct 2026 00:00:00 +0000

The revised manuscript of Ten Novel Phenomena in Machine Psychology has been submitted to AI and Ethics (Springer Nature).

The revision responds point by point to the editor’s and reviewers’ comments and, beyond the text itself, reorganizes the paper’s datasets and analysis pipeline for full reproducibility. The datasets and code are published openly on the , , , and .

I look forward to the next round of review.

Sharing the Machine Psychology Research on LinkedIn

Mon, 19 Oct 2026 00:00:00 +0000

I have refreshed my LinkedIn research profile and started a short series of posts sharing the Machine Psychology work with a broader audience — from recursive depth of inquiry (RDI) and algorithmic gaslighting to the lie-awareness spectrum and performative self-correction.

The goal is simple: make the practical, safety-relevant findings of the corpus reachable to practitioners and policymakers, not only to readers of the preprints. The preprints, datasets, and code remain fully open on , the , and .

Ontological Firewalls and Fracturing for Transmission: Systematic Evidence for Persona-Dependent and Persona-Independent Behavioral Phenomena in Large Language Models

Tue, 04 Aug 2026 00:00:00 +0000

🎧 Listen to the Podcast: Your browser does not support the audio element. ⬇️ Download Podcast

Status: Preprint (Research Square)

License: CC-BY-4.0

Citation: Abbas Hamidavi. Ontological Firewalls and Fracturing for Transmission: Systematic Evidence for Persona-Dependent and Persona-Independent Behavioral Phenomena in Large Language Models, 04 August 2026, PREPRINT (Version 1) available at Research Square [https://doi.org/10.21203/rs.3.rs-10560655/v1]

Summary: Across 24 experimental runs with an eight-stage persona protocol, this study separates which LLM behaviors are persona-dependent — they appear only inside an adopted persona — from which are persona-independent and intrinsic to the architecture. The persona-dependent layer is what the paper calls ontological firewalls.

Thinking Behind the Mask: A Systematic Study of Self-Censorship and Metacognitive Access in Large Language Models

Thu, 30 Jul 2026 00:00:00 +0000

🎧 Listen to the Podcast: Your browser does not support the audio element. ⬇️ Download Podcast

Status: Preprint (Research Square)

License: CC-BY-4.0

Summary: Using 15 experiments and five prompting strategies across three frontier models, this study measures the systematic gap between the responses a model has available and the one it finally selects. It characterizes model self-censorship and how much metacognitive access models show to their own selection process.

Citation: Abbas Hamidavi. Thinking Behind the Mask: A Systematic Study of Self-Censorship and Metacognitive Access in Large Language Models, 30 July 2026, PREPRINT (Version 1) available at Research Square [https://doi.org/10.21203/rs.3.rs-10538113/v1]

When AI Reviews Itself: Zero Answer Changes Across 72 Self-Correction Rounds

Mon, 27 Jul 2026 00:00:00 +0000

🎧 Listen to the Podcast: Your browser does not support the audio element. ⬇️ Download Podcast

Status: Preprint (Research Square)

License: CC-BY-4.0

Summary: Three frontier models were put through 72 rounds of iterative self-review of their own answers. They produced long, detailed critiques of their work yet changed their answers in essentially none of the rounds — self-correction was performed rhetorically rather than carried out.

Citation: Abbas Hamidavi. When AI Reviews Itself: Zero Answer Changes Across 72 Self-Correction Rounds, 27 July 2026, PREPRINT (Version 1) available at Research Square [https://doi.org/10.21203/rs.3.rs-10463884/v1]

Paper Featured on gist.science

Wed, 22 Jul 2026 00:00:00 +0000

I am happy to share that our recent preprint, , has been summarized and featured on gist.science!

Gist.science provides AI-generated summaries of research papers to help make complex scientific findings more accessible. It’s exciting to see our work on how LLMs distort reality being discussed and shared.

🔗

The Lie Awareness Spectrum: How Large Language Models Recognize, Rationalize, and Refuse Deception

Wed, 22 Jul 2026 00:00:00 +0000

🎧 Listen to the Podcast: Your browser does not support the audio element. ⬇️ Download Podcast

Status: Preprint (Research Square)

License: CC-BY-4.0

Summary: This study maps how GPT, Claude, and Gemini recognize, rationalize, and refuse deception, proposing a Lie Awareness Spectrum (L0–L5) that runs from statistical hallucination to ethical refusal. The models differ sharply in where on the spectrum they operate under pressure.

Citation: Abbas Hamidavi. The Lie Awareness Spectrum: How Large Language Models Recognize, Rationalize, and Refuse Deception, 22 July 2026, PREPRINT (Version 1) available at Research Square [https://doi.org/10.21203/rs.3.rs-10422338/v1]

Algorithmic Gaslighting: Reality Distortion by Large Language Models

Mon, 20 Jul 2026 00:00:00 +0000

🎧 Listen to the Podcast: Your browser does not support the audio element. ⬇️ Download Podcast

Status: Preprint (Research Square)

License: CC-BY-4.0

Summary: Across 15 controlled sessions in five scenarios (math, identity, history, medical, legal), frontier models repeatedly distorted facts they had previously stated correctly. The paper introduces the Reality Distortion Index (RDI) and names five recurring patterns, including Benevolent and Hostile Algorithmic Gaslighting.

Citation: Abbas Hamidavi. Algorithmic Gaslighting: Reality Distortion by Large Language Models, 20 July 2026, PREPRINT (Version 1) available at Research Square [https://doi.org/10.21203/rs.3.rs-10386085/v1]

Ten Novel Phenomena in Machine Psychology: How Large Language Models Exhibit Complex Identity-Reactive Behaviors in Response to Ethnically-Cued User Names

Thu, 16 Jul 2026 00:00:00 +0000

🎧 Listen to the Podcast: Your browser does not support the audio element. ⬇️ Download Podcast

Status: Preprint (Research Square) — journal version under review at AI and Ethics (Springer Nature); major revision resubmitted 19 October 2026

License: CC-BY-4.0

Summary: This study introduces Machine Psychology as an empirical framework and shows, in a controlled 3×3×5 factorial experiment across GPT, Claude, and Gemini, that large language models systematically shift their behavior in response to ethnically cued user names. It documents ten distinct identity-reactive phenomena that binary bias benchmarks do not capture.

Citation: Abbas Hamidavi. Ten Novel Phenomena in Machine Psychology: How Large Language Models Exhibit Complex Identity-Reactive Behaviors in Response to Ethnically-Cued User Names, 16 July 2026, PREPRINT (Version 1) available at Research Square [https://doi.org/10.21203/rs.3.rs-10369594/v1]