Humans vs. machines: can enaction theory help?
Abstract
This study seeks to contribute to ongoing research on the formal differences between textual corpora produced by humans and those generated by generative artificial intelligences (GenAI). To this end, we draw on enactive theory and apply it to linguistics, conceiving the asymmetrical structural coupling between an organism and its environment as the basis of cognition. The relations between an actor and the environment are non-linear and irregular: within the perception–action loop, we selectively attend to elements of the environment that are meaningful for action. We propose a novel approach by using Sentence Length Variance (SLV) as an indicator of this asymmetry, or internal irregularity. We also experimented with prompting the machine to integrate greater variance into its texts. We found significant differences between human corpora and AI corpora in terms of SLV, as well as notable divergences among certain GenAI systems. When prompted for higher SLV, only one GenAI produced texts comparable to human writing. We advance several explanations for these findings, which we regard as promising avenues for future research. We conclude by outlining the theoretical implications of our findings, arguing that an enactive linguistics offers a fruitful framework for studying textual differences between human and AI-generated discourse.
Related
- An evolutionary origin of collective decision making in humans and machines
- Pig vocalizations contain shared acoustic structure for humans and machines, but limited correspondence with presumed affective valence
- Biohybrid toggle switches: leveraging bistability in muscle-actuated machines
- Where single-cell foundation models help: a benchmark of annotation, perturbation and regulatory inference against simple baselines
- Plant NLRs get by with a little help from their friends
BibTeX
@article{pallanti2026humans,
title = {{Humans vs. machines: can enaction theory help?}},
author = {Luca Pallanti and Rawad Chaker},
journal = {AI-Linguistica. Linguistic Studies on AI-Generated Texts and Discourses},
year = {2026},
month = jun,
doi = {10.62408/ai-ling.v4i1.50},
url = {https://hal.science/hal-05789034v1},
note = {CC BY-NC-ND 4.0},
}
RIS
TY - JOUR
TI - Humans vs. machines: can enaction theory help?
AU - Luca Pallanti
AU - Rawad Chaker
PY - 2026
DA - 2026/06/17
JO - AI-Linguistica. Linguistic Studies on AI-Generated Texts and Discourses
DO - 10.62408/ai-ling.v4i1.50
UR - https://hal.science/hal-05789034v1
AB - This study seeks to contribute to ongoing research on the formal differences between textual corpora produced by humans and those generated by generative artificial intelligences (GenAI). To this end, we draw on enactive theory and apply it to linguistics, conceiving the asymmetrical structural coupling between an organism and its environment as the basis of cognition. The relations between an actor and the environment are non-linear and irregular: within the perception–action loop, we selectively attend to elements of the environment that are meaningful for action. We propose a novel approach by using Sentence Length Variance (SLV) as an indicator of this asymmetry, or internal irregularity. We also experimented with prompting the machine to integrate greater variance into its texts. We found significant differences between human corpora and AI corpora in terms of SLV, as well as notable divergences among certain GenAI systems. When prompted for higher SLV, only one GenAI produced texts comparable to human writing. We advance several explanations for these findings, which we regard as promising avenues for future research. We conclude by outlining the theoretical implications of our findings, arguing that an enactive linguistics offers a fruitful framework for studying textual differences between human and AI-generated discourse.
SN - 2943-0070
SN - 2943-0070
ER -