Me, Myself, and AI: The Situational Awareness Dataset (SAD) for LLMs

2024/07/05 by Rudolf Laine, Bilal Chughtai, Laine, Rudolf +15 · 11 citations
Business, Management and Accounting · Health Professions · Social Sciences · #Artificial Intelligence (cs.AI) #Big Data and Business Intelligence #Computation and Language (cs.CL) #Ethics and Social Impacts of AI #FOS: Computer and information sciences #Machine Learning (cs.LG) #Occupational Health and Safety Research

paper · pdf · doi:10.48550/arxiv.2407.04694

openalex publication_date 2024/07/05 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

AI assistants such as ChatGPT are trained to respond to users by saying, "I am a large language model". This raises questions. Do such models know that they are LLMs and reliably act on this knowledge? Are they aware of their current circumstances, such as being deployed to the public? We refer to a model's knowledge of itself and its circumstances as situational awareness. To quantify situational awareness in LLMs, we introduce a range of behavioral tests, based on question answering and instruction following. These tests form the Situational Awareness Dataset (SAD), a benchmark comprising 7 task categories and over 13,000 questions. The benchmark tests numerous abilities, including the capacity of LLMs to (i) recognize their own generated text, (ii) predict their own behavior, (iii) determine whether a prompt is from internal evaluation or real-world deployment, and (iv) follow instructions that depend on self-knowledge. We evaluate 16 LLMs on SAD, including both base (pretrained) and chat models. While all models perform better than chance, even the highest-scoring model (Claude 3 Opus) is far from a human baseline on certain tasks. We also observe that performance on SAD is only partially predicted by metrics of general knowledge (e.g. MMLU). Chat models, which are finetuned to serve as AI assistants, outperform their corresponding base models on SAD but not on general knowledge tasks. The purpose of SAD is to facilitate scientific understanding of situational awareness in LLMs by breaking it down into quantitative abilities. Situational awareness is important because it enhances a model's capacity for autonomous planning and action. While this has potential benefits for automation, it also introduces novel risks related to AI safety and control. Code and latest results available at https://situational-awareness-dataset.org .

Cited by

• Detecting the Disturbance: A Nuanced View of Introspective Abilities in LLMs
• Take Goodhart Seriously: Principled Limit on General-Purpose AI Optimization
• A Two-Step, Multidimensional Account of Deception in Language Models
• From Poisoned to Aware: Fostering Backdoor Self-Awareness in LLMs
• ManagerBench: Evaluating the Safety-Pragmatism Trade-off in Autonomous LLMs
• Evidence for Limited Metacognition in LLMs
• Measuring Bias or Measuring the Task: Understanding the Brittle Nature of LLM Gender Biases
• Language Models Identify Ambiguities and Exploit Loopholes
• The Self-Execution Benchmark: Measuring LLMs' Attempts to Overcome Their Lack of Self-Execution
• Search-Time Data Contamination
• Persistent Instability in LLM's Personality Measurements: Effects of Scale, Reasoning, and Conversation History

Related

• I Think, Therefore I am: Benchmarking Awareness of Large Language Models Using AwareBench
• Taken out of context: On measuring situational awareness in LLMs
• Do Large Language Models Know What They Are Capable Of?
• Do Large Language Models Know What They Don't Know?
• LLMs Position Themselves as More Rational Than Humans: Emergence of AI Self-Awareness Measured Through Game Theory