Computational approaches to cognition: the origins of social and causal reasoning in children and primates

想象一下，看着一个同行者走近售票机，在机器的一侧仔细地摩擦一枚一磅的硬币，然后在取票之前将其插入插槽。当轮到你的时候，你会怎么做？插入硬币似乎很重要，但也许你也会先摩擦它-毕竟，那家伙一定有理由这样做。从很小的时候起，孩子们就能很快地模仿别人的动作，而且从很少的反馈中，他们就能屏蔽掉不必要的动作，专注于有效的动作。但在这样的环境中，他们通常会按顺序复制所有的动作，即使是那些看起来不太重要的动作。黑猩猩不会这样做：当然，这两个物种都是高度社会化的生物，它们的因果学习发生在丰富的社会背景下。因此，一个自然的问题是社会互动如何通知和影响学习。但令人惊讶的是，我们对行为者的意图和观察者的先验物理知识如何影响儿童的因果判断知之甚少，更不知道我们的灵长类亲属如何利用这两种信息来源。我们建议尝试解开不同灵长类物种的学习者，包括我们自己，如何权衡这些信息来源时，他们决定复制什么。我们将首先使用一个计算模型，试图预测学习者应该如何从不同类型的证据中学习。然后，我们将进行一项研究，研究不同物种的学习者实际上做了什么。在这项研究中，一些参与者会看到两个有目的的行动之前的结果（如在男子摩擦，然后插入硬币）。其他人会看到同样的两个动作，但演示者似乎是无意中产生结果的（例如，忽略票，继续探索环境）。第三组将看到一个教学演示（想象这个人抓住你的眼睛，在摩擦硬币之前指着它：你更有可能模仿他吗？）。通过这种比较，我们将改变机械可扩展性。对于一半的学习者来说，第一个动作特别不可能具有因果力量（想象这个人摩擦一枚硬币，然后插入另一枚硬币）。我们将测量演示后儿童、猿和猴子产生这两种动作的倾向。如果他们整合了一些先前的物理知识，他们就不太可能复制两个动作，如果一个动作在机械上是不可信的。然而，如果他们整合了社会知识，如果行为者是故意做的，他们应该更有可能复制这两个动作，这可能会导致他们超越任何先前的偏见。后续研究将关注相关问题，例如不同的灵长类物种如何学会合作和协调联合行为。（帮助别人移动一个沉重的沙发。两个人不仅有把沙发拿出来的共同总体目标，他们还有中间目标，比如避免撞到墙上或卡在门口，每个人在实现这些目标方面都有自己的角色。事实上，尽管所有灵长类动物的婴儿似乎都是通过观察来学习的，但人类的孩子却有所不同，这使得他们能够以惊人的速度获得在他们的文化中发展起来的大量人工制品的工作知识，从蜡笔到iPad。与此同时，其他灵长类动物，如黑猩猩，尽管经过了数小时的密切观察，但仍需要数年时间才能掌握一些技能，如使用锤石敲坚果。了解不同之处可能会给我们提供有用的见解，让我们了解孩子们特别适合学习什么样的东西。因此，这项研究将为儿童如何从教师那里学习提供重要的见解，无论是在课堂上正式还是在世界上非正式。

Imagine watching as a fellow traveler approaches a ticket machine, rubs a pound coin carefully on the machine's side, and then inserts it into a slot before retrieving a ticket. When it's your turn, what do you do? Inserting a coin seems important, but perhaps you would rub it first too - after all, that guy must have had a reason to do so. From an early age, children are quick to copy another's action, and from very little feedback they can screen off unnecessary actions and focus on what works. But in a context like this they usually copy all the actions in a sequence, even those that seem unlikely to matter. Chimpanzees don't do this: if they can see that an action is pointless, they don't copy it. Of course, both species are intensely social beings, and their causal learning takes place in a rich social context. A natural question is therefore how social interaction informs and influences learning. But surprisingly, we know very little about how an actor's intentions and the observer's prior physical knowledge feed into children's causal judgments, and still less about how, if at all, our primate relatives take advantage of either source of information. We propose to try and disentangle how learners of different primate species, including our own, weigh these sources of information when they decide what to copy. We will first use a computational model that tries to predict how learners should learn from different kinds of evidence. We will then conduct a study examining what learners of different species actually do. In the study, some participants will see two purposeful actions precede an outcome (as in the man rubbing and then inserting the coin). Others will witness the same two actions, but the demonstrator will appear to produce the outcome accidentally (e.g. ignoring the ticket, and continuing to explore the environment). A third group will see a pedagogical demonstration (imagine the man catching your eye and pointing at the coin before rubbing it: would you be more likely to copy him?). Cutting across this comparison, we will vary mechanical plausibility. For half of the learners the first action will be particularly unlikely to have causal power (imagine the man rubbing one coin and then inserting a different one). We will measure the tendency of children, apes and monkeys to produce the two actions after the demonstrations. If they integrate some prior physical knowledge, they will be less likely to copy both actions if one is mechanically implausible. However, if they integrate social knowledge, they should be more likely to copy both actions if the actor did them on purpose, and this may lead them to override any previous bias. Follow up studies will look at related questions, such as how different primate species learn to cooperate and coordinate joint behaviors. (Think of helping someone move a heavy sofa. Both people not only have the shared overall goal of getting the sofa out, they have intermediate goals like avoiding bumping into the walls or getting stuck in doorways, and each person has a separate role in achieving those goals). The fact is that although all primate infants seem to learn from observation, something about human children is different, and allows them to acquire a working knowledge of the multitude of artifacts developed in their culture, from crayons to ipads, with breathtaking speed. Meanwhile other primates such as chimpanzees spend several years getting to grips with just a few skills, such as the use of a hammer stone to crack nuts, despite many hours of close observation. Understanding what differs may give us helpful insights into what sorts of things children are uniquely adapted to learn. The research will therefore provide important insights into how children learn from teachers, both formally in the classroom and informally in the world.