← All papers

IFI Working Paper No. 1 June 2026 Working paper — not yet peer reviewed

Is LLM Personality an Artifact of Deployment?

Psychometric Stability of Big Five Self-Reports Across Quantization Levels

Trevor Johnson — Idea Fields Institute · ORCID 0009-0008-7962-0451

Shrinking an AI changes its “personality” — and the personality tests don’t really work anyway.

What we did

When people run AI chatbots on their own computers, they almost never run the full-size original — it’s too big. They run a compressed copy, the same way an MP3 is a compressed copy of a studio recording. Compression comes in grades: light (nearly full quality) and heavy (about a quarter the size). Meanwhile, researchers have taken to giving AI models personality quizzes — the same kind psychologists give people — and publishing the results as “this AI’s personality.” Nobody had checked whether the compressed copy people actually use gives the same answers as the original. We checked: three AI models, each at three compression levels, answering the same 50-question personality quiz over and over — 28,350 answers in total, collected on an ordinary home computer.

What we found

Three things. First, heavy compression changes the answers — on about 1 in 4 questions, the compressed copy gave a flat-out different answer than its own original, even with every setting locked down. Light compression barely changed anything. Second, each AI warps in its own direction: compression made one model describe itself as more organized and curious, and another as less friendly and less calm. There’s no single rule — so there’s no way to correct for it after the fact. Third, and biggest: the quiz itself doesn’t really work on these AIs. In a human personality test, your answers to related questions hang together — if you say you love parties, you also say you start conversations. These models’ answers don’t hang together, not for any model at any compression level. Simply listing the multiple-choice options in reverse order changed the scores more than compression did.

Why it matters

When you see a claim that some AI “has” a personality — anxious, agreeable, whatever — the honest follow-up questions are: which copy of it, set up how, asked in what format? Our results say those details change the answer, sometimes by a lot. And for the small AI models people run at home, the deeper takeaway is that personality-quiz results shouldn’t be trusted at all yet: the measurements don’t meet the basic standards we’d demand before drawing conclusions about a person.

What this does not mean

It doesn’t mean AIs have feelings or inner lives — a quiz answer is a quiz answer. It doesn’t cover the big commercial systems (ChatGPT, Claude, Gemini); we only tested small models that run on home computers. And it doesn’t mean compressed models are worse at their jobs — writing, coding, answering questions — only that their personality-quiz answers change.

Words we used

Quantization
the compression that shrinks an AI to fit on a home computer.
Big Five
the standard five personality traits psychologists measure.
Reliability
whether a test’s answers hang together enough to mean anything.

Psychometric questionnaires are now routinely administered to large language models (LLMs), and a growing critical literature questions what such instruments measure when the respondent is a model. One deployment-side moderator has so far gone unexamined: quantization. Virtually every real-world local deployment of an open-weight model runs a quantized artifact (typically a 4- or 8-bit GGUF served by Ollama or llama.cpp), yet published “LLM personality” estimates are almost always obtained from full-precision checkpoints. We administered the 50-item IPIP Big Five Factor Markers to nine model variants — three open-weight instruction-tuned models (Llama 3.1 8B, Qwen 2.5 7B, Llama 3.2 3B), each at three quantization levels (q4_K_M, q8_0, fp16) — under three response-option presentation conditions, with one greedy and twenty independently seeded sampled repetitions per item: 28,350 stateless, JSON-schema-constrained chat calls on fixed consumer hardware.

Three findings emerge. First, 4-bit quantization materially shifted domain scores relative to the same checkpoint at fp16 (6 of 15 baseline contrasts with bootstrap CIs excluding zero; shifts up to 0.42 scale points, |Cohen's d| up to 3.5), in directions that were idiosyncratic to model family; 8-bit quantization was largely score-preserving (2 of 15; max |d| = 0.76). Under greedy decoding, q4 variants gave a different answer than their fp16 parent on 8–32% of items (q8: 2–12%). Second, internal consistency was inadequate everywhere: across all 135 variant × condition × domain cells, Cronbach's alpha never reached the conventional 0.70 threshold (range −2.07 to 0.47, median −0.10; 56% of cells negative), with no systematic precision gradient — the instrument's keyed structure failed to organize responses at every quantization level, including fp16. Third, response-option presentation alone (display order; letter vs. numeric labels) shifted scores more on average than quantization did (mean |Δ| up to 0.21 vs. 0.17 scale points; |d| up to 4.2), at every precision.

Together these results indicate that “the personality of model X” is underdetermined without the deployment configuration — and that for small open-weight models served in deployment-realistic conditions, questionnaire-based personality measurement fails basic psychometric standards regardless of precision. Quantization level, inference stack, sampler settings, and presentation format belong in the methods section of any LLM psychometrics study.

Everything on the table.

The full paper, its source, and the summary data behind the figures — so the analysis can be checked, reused, or extended.

  • PDF
    main.pdf

    The paper itself, typeset. 252 KB.

  • TeX
    main.tex

    LaTeX source of the typeset paper. 31 KB.

  • MD
    draft.md

    The working draft in Markdown — the readable plain-text version. 27 KB.

  • CSV
    quant_effects.csv

    Quantization effect estimates: per family × domain contrasts (q4/q8 vs. fp16) with bootstrap CIs and Cohen's d. 2 KB.

  • CSV
    alpha.csv

    Cronbach's alpha for all 135 variant × condition × domain cells, with CIs. 10 KB.

  • PNG
    fig_quant_effects.png

    Figure: quantization effects on domain scores. 212 KB.

Johnson, T. (2026). Is LLM personality an artifact of deployment? Psychometric stability of Big Five self-reports across quantization levels. IFI Working Paper No. 1. Idea Fields Institute. https://research.ideafields.institute/papers/llm-personality-quantization/

Questions, corrections, or replications — research@ideafields.institute.