论文信息 ¶

论文原文 ¶

关键词 ¶

摘要 ¶

有效的好奇心驱动的学习需要认识到，检验假设的证据价值取决于正在考虑的其他假设。我们是否直观地表示了假设的可辨别性？在这里，我们向孩子们展示了对盒子里东西的替代假设，然后摇晃盒子（或者让孩子们自己摇晃），这样他们就能听到里面东西的声音。我们发现，孩子们能够将他们听到的证据与他们没有听到但可能在其他假设下听到的想象证据进行比较。儿童（N=160；平均值：5岁零4个月）更喜欢更容易的辨别（实验1-3），而在更难辨别的情况下探索时间更长（实验4-7）。 在16个对比中，儿童的探索时间定量地跟踪了从闻所未闻的替代方案中听到的证据的可辨别性。研究结果与儿童具有“直觉心理物理学”的观点一致：当假设难以区分时，儿童表现出自己的感知能力，探索时间更长。

实验 ¶

实验一共进行了7轮。 1、 实验者告诉孩子，在这个盒子里有闪亮铅笔或者一个小枕头，在另外一个盒子里有闪亮铅笔或者普通铅笔，闪亮铅笔是孩子们的目标物，然后实验者摇盒子，让孩子通过声音判断来选择他们更愿意打开哪个盒子。孩子们选择了小枕头那组。 2、 实验二是一个箱子里是大象或者小象，另外一个箱子里是小猪或者小象。孩子们选择了大象那组。 3、 实验三是一个有2个球或者8个球，一个有8个球或者6个球。事实上，都是有8个球，孩子们选择了更容易区分的一组。 实验结果： 实验1-3的结果表明，4岁和5岁的儿童代表了感知证据的相对可辨别性。至关重要的是，儿童的选择并不是由他们观察到的证据（不同选择之间是相同的）引导的，而是由其与闻所未闻的替代方案的对比来引导的，这与儿童可以模拟新的身体互动和将产生的感知数据的想法一致3。儿童表现自己进行这些感知辨别的能力与幼儿元认知监测的新证据一致，也表明，至少在简单的强迫选择环境中，儿童可以行使元认知控制来进行有效决策。 英文原文： The results of Experiments 1–3 suggest that 4- and 5-year-old children represent the relative discriminability of perceptual evidence. Critically, children’s choices were guided not by the evidence they observed (which was identical between choices) but by its contrast with the unheard alternatives, consistent with the idea that children can simulate novel physical interactions and the perceptual data that will result3. Children’s ability to represent their own ability to make these perceptual discriminations is consistent with emerging evidence for metacognitive monitoring in young children and also suggests that, at least in simple, forced-choice contexts, children can exercise metacognitive control for effective decision-making. 4、 实验4—7，是以球的数量不同为自变量，来观察孩子们摇的时间。

实验结果： 我们的叙述依赖于心理模拟，我们在实验4-7中的定量结果在这些心理模拟中使用理想化的感知辨别能力模型分析了儿童的探索行为。然而，孩子们可能依赖于一些更简单的认知机制或启发式方法，或者资源受限的近似方法来实现这一理想。实验4-7的一个自然选择是，孩子们只考虑了每对弹珠在语言和图形上呈现的数量的简单对比，而根本没有考虑到他们在摇晃盒子时获得的丰富感知数据，也没有考虑到通过对盒子摇晃的心理模拟可能听到的声音。基于每对弹珠数量的绝对差异和（负）比，我们评估了两个这样的启发式模型，它们避免了心理模拟可能带来的计算负担。这两个模型在数值上都表现良好（见SI，附加启发式模型），因此在实验4-7中，儿童确实有可能依赖这种机制。 英文原文： Our account relies on mental simulation, and our quantitative results in Experiments 4–7 analyzed children’s exploratory behavior using idealized models of perceptual discriminability in these mental simulations. However, it is possible that children might have relied on some simpler cognitive mechanism or heuristic, or a resource-constrained approximation to this ideal. One natural alternative to consider for Experiments 4–7 is that children took into account only a simple contrast in the linguistically and graphically presented number of marbles in each pair, without attending at all to the rich perceptual data they obtained in shaking the box or imagining possible sounds they might hear via mental simulations of box shaking. We evaluated two such heuristic models that avoid the computational burden that might accompany mental simulation, based on the absolute difference and the (negative) ratio of the numbers of marbles in each pair. Both of these models perform well numerically (see SI, Additional Heuristic Models), and so it is indeed possible that children rely on such a mechanism in Experiments 4–7.

论文结论： 总的来说，这七个实验的结果表明，至少在熟悉的简单任务领域，孩子们可以模拟身体互动和由此产生的感知数据。此外，儿童可以表现出他们自己的能力，在替代假设下进行所需的感知辨别，以将观察到的数据与模拟的、未观察到的信息进行比较。儿童以支持有效决策和探索的方式代表了不同歧视问题的相对困难：他们更喜欢更容易的问题，在更难的问题下探索得更多。儿童的探索性游戏与感知辨别问题的难度之间的精确、定量关系表明，从儿童早期开始，人类学习者直观地计算辨别替代假设的证据价值，并利用这种不确定性来合理校准他们的探索。 英文原文： Collectively, the results of these seven experiments suggest that, at least in familiar domains with simple tasks, children can simulate physical interactions and the perceptual data that will result. Furthermore, children can represent their own ability to make the perceptual discriminations needed to compare observed data with simulated, unobserved data under alternative hypotheses. Children represent the relative difficulty of different discrimination problems in ways that support effective decision-making and exploration: they prefer easier problems and explore more given harder ones. The precise, quantitative relationship between children’s exploratory play and the difficulty of perceptual discrimination problems suggests that starting in early childhood, human learners intuitively compute the value of evidence for discriminating alternative hypotheses, and use this sense of uncertainty to rationally calibrate their exploration