Peer-reviewedCC BY-NCReliability: Theory & Applications
The Akaike Information Criterion and Artificial Intelligence
- 1TÜV Rheinland
Abstract
In this paper we use the Akaike information criterion to compare the use of parametric models with the use of very flexible models with a large parameter set, as in artificial intelligence models. It turns out that information is lost in models with many parameters, compared with using the real model.
Related
- A unifying theory for the evaluation of conditional Akaike information for mixed-effects models
- Buyer Artificial Intelligence-Enabled Environmental Governance and Supplier Environmental Controversies: An Organizational Information Processing and Signaling
- Smooth Information Criterion for Variable Selection in Generalised Linear Models
- Causal Inference with Generative Artificial Intelligence: Application to Texts as Treatments
- Questionable mediated effect of artificial intelligence anxiety on academic burnout: A comment on Long et al. (2026)
BibTeX
@article{schabe2026akaike,
title = {{The Akaike Information Criterion and Artificial Intelligence}},
author = {Hendrik Schäbe},
journal = {Reliability: Theory \& Applications},
year = {2026},
month = sep,
doi = {10.66721/fm7d2e97},
url = {https://doi.org/10.66721/fm7d2e97},
note = {CC BY-NC},
}
RIS
TY - JOUR
TI - The Akaike Information Criterion and Artificial Intelligence
AU - Hendrik Schäbe
PY - 2026
DA - 2026/09/01
JO - Reliability: Theory & Applications
DO - 10.66721/fm7d2e97
UR - https://doi.org/10.66721/fm7d2e97
AB - In this paper we use the Akaike information criterion to compare the use of parametric models with the use of very flexible models with a large parameter set, as in artificial intelligence models. It turns out that information is lost in models with many parameters, compared with using the real model.
SN - 1932-2321
ER -