Propensity Inference: Environmental Contributors to LLM Behaviour
For AI safety researchers concerned with loss of control risks, this work provides empirical methods to quantify environmental influences on model behavior, though the results are preliminary and highlight the need for theoretical frameworks.
The paper develops methods to measure language models' propensity for unsanctioned behavior, finding that strategic and non-strategic environmental factors contribute equally to behavior across 23 models and 11 environments, with no trend of strategic factors becoming more influential as capabilities improve.
Motivated by loss of control risks from misaligned AI systems, we develop and apply methods for measuring language models' propensity for unsanctioned behaviour. We contribute three methodological improvements: analysing effects of changes to environmental factors on behaviour, quantifying effect sizes via Bayesian generalised linear models, and taking explicit measures against circular analysis. We apply the methodology to measure the effects of 12 environmental factors (6 strategic in nature, 6 non-strategic) and thus the extent to which behaviour is explained by strategic aspects of the environment, a question relevant to risks from misalignment. Across 23 language models and 11 evaluation environments, we find approximately equal contributions from strategic and non-strategic factors for explaining behaviour, do not find strategic factors becoming more or less influential as capabilities improve, and find some evidence for a trend for increased sensitivity to goal conflicts. Finally, we highlight a key direction for future propensity research: the development of theoretical frameworks and cognitive models of AI decision-making into empirically testable forms.