Imagine All the People: How the Brain Creates and Uses Personality Models
Type: paper Slug: imagine-all-the-people-how-the-brain-creates-and-uses-personality-models—hassabis Sources: imagine-all-the-people-how-the-brain-creates-and-uses-personality-models—hassabis Last updated: 2026-05-13
Summary
Hassabis, Spreng, Rusu, Robbins, Mar, and Schacter (2014) used fMRI to identify the neural substrates of personality model construction and deployment. Participants learned four protagonist personalities, then imagined scenarios from each protagonist’s perspective or their own. A core social-mental simulation network (dMPFC, aTL, PCC) supported perspective-taking across self and other conditions, while multivariate pattern analysis in anterior/dorsal MPFC reliably decoded which protagonist was being simulated.
Core content
Design: Participants learned detailed personality profiles for four protagonists conforming to distinct personality types. In fMRI, they completed three conditions: (1) imagine from a protagonist’s perspective, (2) imagine from their own perspective in the same scenario, (3) imagine an empty scene (spatial context only). All were contrasted against a syllable-counting control (paper—imagine-all-the-people-how-the-brain-creates-and-uses-personality-models §Methods).
Key findings:
- Protagonist + Self conditions (vs. empty scene) revealed a social simulation network: dorsal and anterior MPFC, anterior temporal lobes, posterior cingulate (paper—imagine-all-the-people-how-the-brain-creates-and-uses-personality-models §Results).
- Multivariate pattern classification in anterior and dorsal MPFC reliably discriminated among the four protagonists (paper—imagine-all-the-people-how-the-brain-creates-and-uses-personality-models §Results).
- The self-condition did not show above-chance decoding of the “implied” protagonist, suggesting personality representations are recruited primarily for other-simulation (paper—imagine-all-the-people-how-the-brain-creates-and-uses-personality-models §Results).
- Empty scene condition activated hippocampus and retrosplenial cortex — the scene construction component — without the social regions (paper—imagine-all-the-people-how-the-brain-creates-and-uses-personality-models §Results).
Theoretical contribution: Demonstrates that the default network can be fractionated into social-personality simulation (MPFC, aTL, PCC) and spatial scene construction (hippocampus, RSC) components, extending the scene construction account (paper—the-construction-system-of-the-brain) to social cognition.
Connections- Theme: theme—hippocampal-construction, theme—memory-imagination, social-cognition
- Collaborators: R. Nathan Spreng, A. Rusu, C. Robbins, Raymond A. Mar, Daniel L. Schacter
- Era: deepmind-era
- Venue: venue—Cerebral-Cortex
- Extends: paper—the-construction-system-of-the-brain — social extension of scene construction framework
Honest Gaps
- Metadata incorrectly lists year as 2007; published 2014.
- The four personality profiles were highly schematic — real-world personality knowledge is far richer and messier.
- The “empty scene” condition removes all social content, making the social-vs-spatial contrast potentially confounded with complexity differences.
- Sample size not noted in extraction; results from multivariate decoding may be underpowered.
- Data presented as unpublished in Schacter et al. (2012) review; unclear if final published version includes all analyses described.