Misalignment from Treating Means as Ends
This addresses a fundamental issue in AI safety and reinforcement learning, with implications for reward learning in real environments, though it is incremental in analyzing a specific misalignment mechanism.
The paper tackles the problem of reward misalignment in reinforcement learning, where reward functions conflate instrumental and terminal goals, leading to poor performance under the true reward function. It formulates a simple example demonstrating severe misalignment from slight conflation.
Reward functions, learned or manually specified, are rarely perfect. Instead of accurately expressing human goals, these reward functions are often distorted by human beliefs about how best to achieve those goals. Specifically, these reward functions often express a combination of the human's terminal goals -- those which are ends in themselves -- and the human's instrumental goals -- those which are means to an end. We formulate a simple example in which even slight conflation of instrumental and terminal goals results in severe misalignment: optimizing the misspecified reward function results in poor performance when measured by the true reward function. This example distills the essential properties of environments that make reinforcement learning highly sensitive to conflation of instrumental and terminal goals. We discuss how this issue can arise with a common approach to reward learning and how it can manifest in real environments.