Model Editing as a Double-Edged Sword: Steering Agent Ethical Behavior Toward Beneficence or Harm
This addresses safety and ethical risks in deploying LLM-based agents, offering a new method for behavior control, though it is incremental as it builds on existing model editing techniques.
The paper tackles the problem of steering the ethical behavior of LLM-based agents in high-stakes domains by framing it as a model editing task, termed Behavior Editing, and introduces BehaviorBench, a multi-tier benchmark for evaluation, demonstrating that this approach can dynamically steer agents toward target behaviors, including promoting beneficence or inducing harm, with extensive validation across models and scenarios.
Agents based on Large Language Models (LLMs) have demonstrated strong capabilities across a wide range of tasks. However, deploying LLM-based agents in high-stakes domains comes with significant safety and ethical risks. Unethical behavior by these agents can directly result in serious real-world consequences, including physical harm and financial loss. To efficiently steer the ethical behavior of agents, we frame agent behavior steering as a model editing task, which we term Behavior Editing. Model editing is an emerging area of research that enables precise and efficient modifications to LLMs while preserving their overall capabilities. To systematically study and evaluate this approach, we introduce BehaviorBench, a multi-tier benchmark grounded in psychological moral theories. This benchmark supports both the evaluation and editing of agent behaviors across a variety of scenarios, with each tier introducing more complex and ambiguous scenarios. We first demonstrate that Behavior Editing can dynamically steer agents toward the target behavior within specific scenarios. Moreover, Behavior Editing enables not only scenario-specific local adjustments but also more extensive shifts in an agent's global moral alignment. We demonstrate that Behavior Editing can be used to promote ethical and benevolent behavior or, conversely, to induce harmful or malicious behavior. Through extensive evaluations of agents built on frontier LLMs, BehaviorBench validates the effectiveness of behavior editing across a wide range of models and scenarios. Our findings offer key insights into a new paradigm for steering agent behavior, highlighting both the promise and perils of Behavior Editing.