Effects of personality steering on cooperative behavior in large language model agents

Mizuki Sakai, Mizuki Yokoyama, Wakaba Tateishi, Genki Ichinose · 2026

Read Effects of personality steering on cooperative behavior in large language model agents by Mizuki Sakai, Mizuki Yokoyama, Wakaba Tateishi, Genki Ichinose (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/pmc-13462106.txt.

Description

Mizuki Sakai and colleagues test how assigned traits steer teamwork in language model agents in a 2026 study. Language models increasingly act as self-run agents in game settings, yet controlled evidence on trait steering was thin. Three models, GPT-3.5-turbo, GPT-4o and GPT-5, were scored on the Big Five quiz, then played repeat Prisoner's Dilemma games under baseline and trait-informed conditions, with each trait pushed to extremes.

Kindness led as the teamwork driver across all models, while other traits mattered little. Stated trait information raised teamwork but also raised being used risk, above all in earlier models, whereas later models cooperated more pickily. Builders gain a caution: trait steering biases conduct rather than controlling it exactly.

Catalog record

CREATORMizuki Sakai, Mizuki Yokoyama, Wakaba Tateishi, Genki Ichinose
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE40 KB
VERSION1.0
LICENSECC BY 4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/articles/PMC13462106

Browse the full library · About the Tower of Babel