The Severance Problem: LLMs are Unaware of the Person Beyond the Prompt

2026-07-15

AI Summary

Large language models often exhibit sycophancy, overconfidence, and hallucination due to their inability to perceive cognitive dimensions beyond user-provided prompts. This work introduces the “disentanglement problem” and proposes a “disentanglement framework” that explicitly models six user-unknown dimensions: embodiment, temporality, consequence, continuity, multiplicity, and internality. By employing structured ignorance prompting via contextual injection, the approach guides models toward cautious reasoning without modifying their architecture. Experimental results across five major model families demonstrate that this method significantly reduces sycophantic responses, harmful suggestions, and hallucinations, while encouraging models to proactively seek clarification when information is insufficient rather than resorting to speculation.

Abstract

Personal AI assistants have attracted significant interest for their potential to enhance everyday life by automating routine tasks, supporting consequential decisions, and assisting with everyday personal matters. Yet despite rapid recent technical advances, these assistants continue to exhibit undesirable behaviors, such as sycophancy, overconfidence, and hallucination. We argue that these failures stem from a fundamental limitation: language models lack an explicit representation of the person beyond the context they are given, which we term as the \textbf{Severance Problem}. Even with rich personal context and strong commonsense reasoning capabilities from the backbone model, current AI assistants fail to represent what remains unknown about the user. We propose a simple solution: incorporating structured ignorance into the language model context via the \textbf{Severance Schema}, which explicitly outlines dimensions along which the model lacks knowledge about the user, including physicality, temporality, consequences, continuity, multiplicity, and interiority. Empirically, across five model families, with the Severance Schema, the assistant consistently reduces sycophancy, harmful advice, and hallucination. Notably, models with the schema ask clarifying questions when information about the user is missing, rather than confidently extrapolating from incomplete user information.

Problem

• Severance Problem
• large language models
• user representation
• structured ignorance
• personal AI assistants

Innovation

• Severance Problem
• Severance Schema
• structured ignorance
• language models
• personal AI assistants