The Lock-in Hypothesis: Stagnation by Algorithm

Tianyi Qiu, Zhonghao He, Tejasveer Chugh, Max Kleiman-Weiner

Keywords: Value lock-in, Human-AI interaction, Large Language Model

TL;DR: Demonstrations of LLM-induced lock-in of users in real-world data, simulations, and a formal model.

Abstract:

The training and deployment of large language models (LLMs) induce a feedback loop: models continually learn human beliefs from data, reinforce user beliefs with generated content, reabsorb those reinforced beliefs, and then feed them back to users. This dynamic resembles an echo chamber. We hypothesize that this feedback loop entrenches the existing values and beliefs of users, leading to a loss of diversity and potentially the lock-in of false beliefs. We formalize this hypothesis and test empirically with agent-based LLM simulations and real-world GPT usage data. These analyses reveal sudden but sustained drops in diversity after the release of new GPT iterations, consistent with the hypothesized human-AI feedback loop.

Submission Type: Long Paper (9 Pages)

Archival Option: This is a non-archival submission

Presentation Venue Preference: ICLR 2025

Submission Number: 45