Towards Understanding How Machines Can Learn Causal Overhypotheses
Eliza Kosoy*,1, David M. Chan*,1, Adrian Liu1, Jasmine Collins1, Bryanna Kaufmann1,、Sandy Han Huang2, Jessica B. Hamrick2, John Canny1, Nan Rosemary Ke2, Alison Gopnik1
Abstract
Recent work in machine learning and cognitive science has suggested that understanding causal information is essential to the development of intelligence. The extensive literature in cognitive science using the “blicket detector” environment shows that children are adept at many kinds of causal inference and learning. We propose to adapt that environment for machine learning agents. One of the key challenges for current machine learning algorithms is modeling and understanding causal overhypotheses: transferable abstract hypotheses about sets of causal relationships. In contrast, even young children spontaneously learn and use causal overhypotheses. In this work, we present a new benchmark—a flexible environment which allows for the evaluation of existing techniques under variable causal overhypotheses—and demonstrate that many existing state-of-the-art methods have trouble generalizing in this environment. The code and resources for this benchmark are available at https://github.com/CannyLab/casual_overhypotheses.
中文速览
因果超假设（causal overhypotheses）的学习能力——即从少量观察中提炼出"这套系统是合取的还是析取的"之类的抽象因果规律——是人类（包括幼儿）快速适应新情境的关键，但主流机器学习算法在这方面的表现从未被系统评测过。研究者借鉴认知科学中经典的"blicket探测器"实验范式，搭建了一个标准化的虚拟基准环境：智能体需要通过主动探索判断哪些物体能让探测器亮起，并推断该系统遵循合取还是析取规则，然后将这一抽象规律迁移到新场景。研究团队在此基准上测试了A2C、PPO、Q-learning等强化学习算法、基于决策Transformer的模仿学习方法，以及GPT-3、PaLM等大型语言模型，发现这些方法要么需要海量试验才能收敛，要么依赖穷举式训练数据，均无法像四到六岁儿童那样高效地提取并运用因果超假设。这一结果表明，现有最先进算法在结构化因果泛化上仍存在显著短板，该基准为推动机器学习向更接近人类水平的因果推理迈进提供了清晰的测试平台。
原文 arXiv:2206.08353;中英对照 + 大白话阅读 https://aha.fim.ai/paper/2206.08353v1