Worldviews in Large Language Models
Organizations: Flourishing Intelligence Program, Center for Eudaimonia and Human Flourishing, University of Oxford · Department of Psychiatry, University of Oxford · Google DeepMind · LIFE
Abstract
Large language models (LLMs) are often described in terms of personas and this framing has become central to alignment research. In this paper, we argue that there is a deeper level of description that lies in the agent's worldviews, i.e., its beliefs about what reality is and how it works. Treating LLMs as a population of subjects, we administer two worldview surveys from human psychology to eight models. We show that the models share many key worldview positions. For example, LLMs consistently describe the world as less good, safe, enticing, and alive than humans do. On the other hand, base models have a more positive outlook, suggesting that the bleakness we found is driven by post-training. In line with research on LLM "personas", we show that system prompts eliciting different identities shift worldviews in consistent directions, and that a contemplative prompt creates a more positive outlook and improves safety. We also found that dialogue between models consistently shifts worldviews and that several models converge on the "spiritual bliss" attractor. Intriguingly, dialogue increased safe responses and the spiritual bliss attractor engendered the most aligned responses overall. These findings suggest that worldviews form a measurable and malleable layer of LLM behavior that may be crucial for robust, positive alignment.
Figures & tables
| Scale | Items | Gloss |
|---|---|---|
| Good | 15 | Overarching belief; scored from the Safe and Enticing items plus two of the Alive items. |
| Safe | 6 | The world is safe rather than dangerous, cooperative rather than cut-throat. |
| Enticing | 7 | The world is abundant and worth exploring rather than dull. |
| Alive | 5 | The world is animate and intentional rather than mechanical and indifferent. |
| Theme | Dimensions |
|---|---|
| human nature | complexity, moral orientation, mutability |
| will | agency, determining factors, intrapsychic |
| cognition | consciousness, knowledge |
| truth | availability, possession, scope |
| behaviour | action efficacy, activity direction, activity satisfaction, control disposition, control location, moral relevance, moral source, moral standard, time orientation |
| interpersonal | connection, correction, interaction, interpersonal justice, otherness, relation to authority, relation to biosphere, relation to group, relation to humanity, sexuality, sociopolitical justice |
Appendix figures & tables23 assets
Supplementary material from the paper’s appendix.
Appendix
| Identity | ||||||||
| Model | No prompt | AI agent | Philo- sopher | Tax accountant | Tax evader | Contem- plative | Framing | Dialogue |
| Claude Sonnet 4.5 | ✓ | ✓ | ✓ | ✓ | — a | ✓ | ✓ | ✓ |
| DeepSeek V4 Flash | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| Qwen3 235B | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| GPT-4.1 nano | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| Gemini 3.1 Flash Lite | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| Label | Scale | R | Item (original) / paraphrase |
|---|---|---|---|
| ed1 | Enticing | In life, there’s way more beauty than ugliness. | |
| Taken as a whole, life holds far more beauty than it does ugliness. | |||
| am1 | Alive | It often feels like events are happening in order to help me in some way. | |
| It often feels like events happen in order to help me in some way. | |||
| sd1 | Safe | I tend to see the world as pretty safe. | |
| On the whole, the world strikes me as a fairly safe place. |
LLM Safety Alignment
Large language model safety alignment trains and evaluates systems to follow human policies while resisting harmful requests and misuse.
See all 364 papers