AI Research Atlas

The persona selection model

Anthropic · 23 February 2026

Anthropic argues assistants act human-like because pretraining teaches them to simulate personas and post-training merely selects one.

A theory with no experiment behind it. Pretrained models simulate many characters, and post-training picks and refines the Assistant persona, so human-like traits are the default and need no instilling.

Date
Monday, 23 February 2026
Lab
Anthropic
Kind
paper
Access
paper only

Theoretical framing drawing on prior work by others; presented as a hypothesis.

Sources

  1. www.anthropic.com/research/persona-selection-model

This record was checked against its sources on 6 October 2026. How we check

Related