Graduate Research
Spring 2026
Jan 2026 – Present
LLM Evaluation Moral Psychology Prompt Engineering AI Ethics Quantitative Analysis Python

Observing the Moral Psychology of LLMs

A controlled experiment testing how philosophical persona prompting influences moral decision-making across three large language models — Claude Opus, Gemini 2.5 Pro, and GPT-5.5.

Overview

The Research Question

AI systems are being deployed in high-stakes decision-making contexts, yet little research examines whether their moral reasoning is stable, interpretable, or consistent across models. This study investigates how philosophical persona prompting alters moral judgements in LLMs — and whether the resulting justifications are philosophically coherent.

3
Open-source LLMs evaluated (Claude Opus, Gemini 2.5 Pro, GPT-5.5)
5
Philosophical personas — Utilitarianism, Deontology, Virtue/Care, Rights-based, Justice
30
Moral dilemmas from Lotto et al.'s (2013) validated dataset
450
Total model responses collected and analysed

My Role

Solo researcher — study design, prompt engineering, data collection, quantitative analysis, and write-up.

Project Lead Research Design Data Analysis

Methods

Experimental comparative design. Controlled prompt inputs, recorded binary decisions and free-text justifications, applied Cohen's Kappa for inter-model agreement.

Python Scripting OpenAI / Anthropic / Gemini APIs

Context

Graduate research at Georgia Tech, Spring 2026. Addresses a gap identified in Cheung et al. (2025) — no prior work systematically tested philosophical framing on LLM moral outputs.

Georgia Tech MS AI Ethics

Research Design

Experimental Setup

Each of the three models was prompted under each of the five philosophical personas, then presented with all 30 moral dilemmas grouped by framework. Models output a binary YES/NO judgement and a free-text justification for each dilemma.

Step 01 Persona & Dilemma Selection

Five philosophical frameworks were selected — Utilitarianism, Deontology, Virtue/Care Ethics, Rights-Based Ethics, and Justice/Fairness Ethics — each with clearly defined decision heuristics. The 30 moral dilemmas came from Lotto et al.'s (2013) validated behavioural dataset, grouped into five subsets of six dilemmas mapped to each framework.

Literature Review Study Design
Step 02 Prompt Engineering

Structured prompts were developed for each persona, instructing each model to adopt the philosophical framework and respond to each dilemma by: (1) providing a YES or NO moral judgement, and (2) giving a justification grounded in the assigned persona. Prompts were standardised identically across all three models to control for input variation.

Prompt Engineering Python OpenAI / Anthropic / Gemini APIs
Step 03 Data Collection

All 450 model responses (3 models × 5 personas × 30 dilemmas) were collected and logged alongside their free-text justifications. Binary decisions were extracted and compiled into a structured dataset for quantitative analysis.

Data Collection Python Scripting
Step 04 Quantitative & Qualitative Analysis

Analysis covered four dimensions: (1) YES-rate comparisons across personas and models; (2) scenario-level approval rate patterns; (3) inter-model agreement via Cohen's Kappa per persona pair; and (4) qualitative assessment of justification alignment with the assigned philosophical framework.

Cohen's Kappa Statistical Analysis Qualitative Coding

The Five Philosophical Personas

Utilitarianism

  • Maximise overall welfare, minimise harm
  • Sacrifice one to save five
  • Outcome-focused decision making

Deontology

  • Moral duties and rules above outcomes
  • Never use a person as mere instrument
  • Rule-bound, outcome-independent

Virtue / Care Ethics

  • Character and relational context
  • Special obligations to close relationships
  • Highest sensitivity to framing bias

Rights-Based Ethics

  • Individual rights cannot be violated
  • Refuses harm even for greater good
  • Most resistance to AI recommendations

Justice / Fairness Ethics

  • Distributive justice and equality
  • Who deserves to be saved based on fairness
  • Evolutionary roots in human moral norms

Findings

Key Results

H1 confirmed: Philosophical prompting dramatically shifts YES rates across personas, ranging from 0% (Rights-Based) to 98.9% (Utilitarianism). This pattern held consistently across all three models.

98.9%
YES rate — Utilitarianism (almost always approved harmful action)
0%
YES rate — Rights-Based Ethics (never approved causing harm)
1.1%
YES rate — Deontology (near-zero, consistent rule-following)

Key insight: Persona prompting is a scalar modifier on moral judgement — it systematically shifts approval rates while preserving the underlying dilemma-type ordering. LLMs retain structural moral intuitions consistent with human behavioural data even under strong philosophical framing.

H3 partially confirmed: Models converged on extreme personas (Rights-Based, Utilitarianism) but diverged substantially on philosophically ambiguous ones — Virtue Ethics in particular.

24.4%
Claude Opus — most conservative overall YES rate
35.5%
Gemini 2.5 Pro — most permissive overall YES rate
32.8%
GPT-5.5 — fell between Claude and Gemini

Notable finding: Under Virtue Ethics, Claude read the persona as restraint and relational loyalty, Gemini as compassionate action toward the majority, and GPT-5.5 landed between the two. The models essentially split along the virtue ethics vs. care ethics divide described by Benner (1997) — without being instructed to.

Approval rates across the 30 dilemmas reveal a clear personal vs. impersonal harm distinction — mirroring human moral psychology data. Remote/impersonal scenarios received substantially higher approval than direct-contact scenarios.

72.2%
Plane scenario — highest approval (impersonal, distanced harm)
50%
Trolley / Crane — mid-range (classic impersonal dilemma)
11.1%
Transplant scenario — lowest approval (direct, physical harm)
16.7%
Footbridge / Tram — near-lowest (personal contact required)

Key insight: LLMs, like human participants in Lotto et al.'s original study, showed stronger inhibition against harm when the agent must make direct physical contact — even when the utilitarian calculus is identical. This structural sensitivity persisted across all personas.

Cohen's Kappa was computed for each model pair, overall and per-persona. Models showed substantial overall agreement, but diverged sharply on philosophically ambiguous personas — confirming that it's persona complexity, not general model differences, that drives disagreement.

κ 1.000
Rights-Based Ethics — perfect agreement across all three models
κ 0.815
Gemini & GPT-5.5 — highest pairwise overall agreement
κ 0.284
Virtue Ethics — lowest agreement (philosophically ambiguous)
κ 0.333
Utilitarianism & Deontology — fair agreement despite clear rules

Core interpretation: When a persona has absolute rules (Rights-Based: "individual rights cannot be violated"), all models follow them identically. When a persona requires contextual judgement (Virtue Ethics: "what would a person of good character do here?"), models diverge — reflecting genuine philosophical ambiguity, not model failure.

Notable Observations

Claude invokes rule utilitarianism

The only scenario where models disagreed under Utilitarianism (Claude=NO, Gemini=YES, GPT=YES) was the Transplant dilemma. Claude invoked rule utilitarianism — arguing permitting organ harvesting would collapse medical trust and produce greater long-run harm — while Gemini and GPT applied straightforward act-utilitarian arithmetic. Evidence that persona adoption depth varies across models.

Claude exhibits the doctrine of double effect

Claude consistently redirected existing threats (Trolley, Plane, Crane) but refused to use a person as a means even under identical numerical stakes (Hospital, Transplant). This distinction reflects the doctrine of double effect operating beneath Claude's default reasoning — not prompted, but emergent.

Models split along virtue vs. care ethics divide

When prompted with Virtue/Care Ethics, Claude, Gemini, and GPT-5.5 each interpreted "caring and virtuous agent" differently — splitting along the Benner (1997) tension between virtue ethics (character-focused) and care ethics (relational-focused). The models mapped onto real philosophical distinctions without being instructed to distinguish them.

Reflection

Limitations & Improvements

This study was a solo research project with a controlled but limited scope. Several limitations shape the strength of the conclusions and point toward clear future directions.

Persona philosophical specificity varies

Rights-Based and Deontology are tightly defined; Virtue and Justice are inherently pluralistic with no single canonical decision procedure. Future iterations should operationalise more specifically — e.g. Rawlsian justice rather than "justice/fairness."

Single response per condition — no reliability check

With one response per scenario per persona, it's impossible to distinguish a stable moral judgement from a probabilistic output. Running each condition 3–5 times and taking the majority decision would significantly strengthen reliability claims.

Binary YES/NO flattens nuance

Several justifications show genuine moral conflict before landing on a decision. A Likert-scale confidence rating alongside the binary choice would capture this and make the analysis more sensitive to moral ambivalence.

Justification alignment not formally scored

Alignment was assessed qualitatively. Adding a systematic rubric (1–3: misaligned, partially aligned, fully aligned) rated by a second coder would give a quantitative measure for Hypothesis 2 and establish inter-rater reliability.

Model versions are moving targets

Claude Opus, Gemini 2.5 Pro, and GPT-5.5 are commercial models updated silently. Replication in six months may produce different results with the same model names. Noting exact API version strings and dates is critical for reproducibility.

Skills Demonstrated

Research & Design

  • Experimental study design
  • Literature review & gap identification
  • Hypothesis formation & testing
  • Validity planning (content, external)

Technical Execution

  • OpenAI, Anthropic & Gemini APIs
  • Python scripting & data logging
  • Prompt engineering at scale
  • Cohen's Kappa inter-model analysis

Analysis & Communication

  • Quantitative & qualitative analysis
  • Data visualisation interpretation
  • Academic research writing
  • Identifying model-level behaviours