Saved in:
Bibliographic Details
Main Authors: Underwood, Ted, Nelson, Laura K., Wilkens, Matthew
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2505.00030
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908344231395328
author Underwood, Ted
Nelson, Laura K.
Wilkens, Matthew
author_facet Underwood, Ted
Nelson, Laura K.
Wilkens, Matthew
contents Before researchers can use language models to simulate the past, they need to understand the risk of anachronism. We find that prompting a contemporary model with examples of period prose does not produce output consistent with period style. Fine-tuning produces results that are stylistically convincing enough to fool an automated judge, but human evaluators can still distinguish fine-tuned model outputs from authentic historical text. We tentatively conclude that pretraining on period prose may be required in order to reliably simulate historical perspectives for social research.
format Preprint
id arxiv_https___arxiv_org_abs_2505_00030
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Can Language Models Represent the Past without Anachronism?
Underwood, Ted
Nelson, Laura K.
Wilkens, Matthew
Computation and Language
Before researchers can use language models to simulate the past, they need to understand the risk of anachronism. We find that prompting a contemporary model with examples of period prose does not produce output consistent with period style. Fine-tuning produces results that are stylistically convincing enough to fool an automated judge, but human evaluators can still distinguish fine-tuned model outputs from authentic historical text. We tentatively conclude that pretraining on period prose may be required in order to reliably simulate historical perspectives for social research.
title Can Language Models Represent the Past without Anachronism?
topic Computation and Language
url https://arxiv.org/abs/2505.00030