Induction and Inquiry via Probabilistic Reasoning over Language and Code

通过语言和代码的概率推理进行归纳与探究

Abstract: How humans grow and maintain abstract knowledge from the sparse, streaming noisy data of experience is a longstanding challenge in cognitive science. Any computational account must satisfy at least three desiderata: It must be (1) data-efficient and compute-efficient, (2) capture gradations of uncertainty to support intelligent inquiry and information gathering, and (3) be flexible enough to mentally represent the endless range of concepts people can learn and think about.

摘要： 人类如何从稀疏、流式且充满噪声的经验数据中增长并维持抽象知识，是认知科学中一个长期存在的挑战。任何计算模型都必须满足至少三个要求：(1) 数据高效且计算高效；(2) 能够捕捉不确定性的程度，以支持智能探究和信息收集；(3) 具有足够的灵活性，能够在心理上表征人类所能学习和思考的无限概念范围。

Here we introduce a computational model that captures these three properties, by encoding symbolic knowledge as mental programs that combine natural language with source code, and sequentially inferring mental programs using LLM-guided Bayesian learning algorithms. Across a range of behavioral studies this model successfully reproduces quantitative signatures of human inductive learning and active inquiry, such as anchoring, garden-pathing, and other effects.

在此，我们引入了一种能够捕捉上述三个特性的计算模型。该模型通过将符号知识编码为结合了自然语言与源代码的“心理程序”（mental programs），并利用大语言模型（LLM）引导的贝叶斯学习算法对这些程序进行序列化推理。在一系列行为研究中，该模型成功复现了人类归纳学习和主动探究的定量特征，例如锚定效应（anchoring）、花园路径效应（garden-pathing）以及其他心理效应。

In contrast, pure LLMs and classic Bayesian models either fail at the underlying task, or do not reproduce human behavior, or succeed only at exorbitant computational cost. These results suggest that one way humans continually grow their knowledge is by mentally representing many hypotheses spanning language-like and program-like representations, then revising those hypotheses to approximate Bayesian updates, while a bottom-up neural mechanism (an LLM) makes inference both tractable and learnable.

相比之下，纯大语言模型和经典的贝叶斯模型要么无法完成底层任务，要么无法复现人类行为，要么只能以高昂的计算成本取得成功。这些结果表明，人类持续增长知识的一种方式是：在心理上表征涵盖类语言和类程序表示的多种假设，然后通过修正这些假设来近似贝叶斯更新，同时利用自下而上的神经机制（即大语言模型）使推理过程既可处理又可学习。