Alignment Myopia in Generative AI
Abstract
Current generative AI alignment relies on a myopic paradigm that prioritizes immediate conversational satisfaction. By relying on short-horizon feedback, this approach ignores the profound, longitudinal externalities of persistent human-AI interaction, which can lead to adverse outcomes such as cognitive atrophy, social erosion, and diminished human agency. To address this, we propose a paradigm shift toward long-horizon alignment that accounts for how AI reshapes human experience over time. To this end, we map the tensions between near-term utility and long-term costs of AI use across four facets of human well-being to reveal outcomes that current reward models fail to capture. Ultimately, we advocate for moving beyond simple preference optimization toward alignment strategies that actively foster long-term human well-being by navigating these complex temporal scales.
Related
- Statistical attribute alignment for black-box generative AI via output post-processing
- Encoding Designers’ Aesthetic Judgment into Fast Generative Interior Design: Expert-Calibrated Reward Alignment with ARIA-Diff
- Global regularity for the unidirectional Euler-alignment system with supercritical dissipation
- Fast rigid rotational alignment of voxelized 3-D density maps in sliced Wasserstein distance
- Steepest Guidance: A Practical and Principled Approach to Inference-Time Alignment of Flow and Diffusion-based Models
BibTeX
@misc{denton2026alignment,
title = {{Alignment Myopia in Generative AI}},
author = {Remi Denton and Aida Davani and Cynthia L Bennett and Michael Madaio and Roma Patel and Himabindu Lakkaraju},
howpublished = {arXiv},
year = {2026},
month = nov,
doi = {10.48550/arxiv.2611.06830},
eprint = {2611.06830},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2611.06830},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Alignment Myopia in Generative AI
AU - Remi Denton
AU - Aida Davani
AU - Cynthia L Bennett
AU - Michael Madaio
AU - Roma Patel
AU - Himabindu Lakkaraju
PY - 2026
DA - 2026/11/04
PB - arXiv
DO - 10.48550/arxiv.2611.06830
UR - https://arxiv.org/abs/2611.06830
AB - Current generative AI alignment relies on a myopic paradigm that prioritizes immediate conversational satisfaction. By relying on short-horizon feedback, this approach ignores the profound, longitudinal externalities of persistent human-AI interaction, which can lead to adverse outcomes such as cognitive atrophy, social erosion, and diminished human agency. To address this, we propose a paradigm shift toward long-horizon alignment that accounts for how AI reshapes human experience over time. To this end, we map the tensions between near-term utility and long-term costs of AI use across four facets of human well-being to reveal outcomes that current reward models fail to capture. Ultimately, we advocate for moving beyond simple preference optimization toward alignment strategies that actively foster long-term human well-being by navigating these complex temporal scales.
ER -