Tacit sensitivity to structure and judgements of style in everyday behaviour

通过偶然接触个别例子，人们对日常经验中的结构性规律变得敏感。因此，我们对他人的个人风格以及艺术家和作曲家个人作品的风格产生了敏感性，对某一特定写作类型的感觉，最喜欢的电视连续剧的可能结局和适当的顺序，左边车道上的司机可能做什么而右边车道上的司机不会做什么，以及与银行出纳员的正常互动是什么样的。通常，这种敏感性是毫不费力地发展起来的，我们没有任何意图去学习这些规律，甚至没有意识到有任何这样的规律需要学习。然而，普通经验和实验室研究都表明，这种隐性敏感性影响着我们在几乎每一项任务中的表现和期望。许多非人类的动物，如黑猩猩、老鼠、鸽子，甚至鱼，都能表现出对艺术风格、音乐甚至笔迹的类似敏感性。本研究探讨了这种隐性、结构性知识的来源，以及它如何影响我们在分类、识别和回忆方面的表现，以及它如何影响我们和其他动物的偏见和偏好。先前的研究表明，作为这些影响来源的知识不是以规则或原型等高级抽象形式存在的，而是潜伏在我们的记忆中，分布在我们对特定经历的记忆中。这些记忆在对类似当前事件的反应中被集体吸收，以提供一种熟悉或“正确”的感觉，或者在我们对事件的反应中提供一种流动性，这种感觉与我们在类似情况下所经历的结构一致。这种检索还提供了检测事件何时与结构不一致的能力，即使我们无法清楚地表达我们是如何“知道”它的。这项提议的研究探索了这种记忆和检索过程的模型，该模型将允许人类和动物进行这种内隐学习和使用结构。该模型以人工神经网络的形式应用于单个事件的低级表示，例如图像的像素图或数字化声音。

People become sensitive to the structural regularities in everyday experiences through incidental exposure to individual examples. Thus, we develop sensitivity to other people's personal styles and the styles in the works of individual artists and composers, a sense of what does and does not belong to a particular genre of writing, the likely conclusion to and appropriate sequences of a favorite televison series, what drivers in then left lane are likely to do that drivers in the right are not, and what a normal interaction with a teller at our bank is like. Often this sensitivity develops effortlessly and without any intention on our part to learn these regularities, and even without any awareness that there are any such regularities to learn. Yet, both common experience and laboratory research alike demonstrate that this tacit sensitivity influences our performance and expectations in virtually every task we undertake. Many non-human animals, such as chimpanzees, rats, pigeons, and even fish can demonstrate similar sensitivities to style in art, music, and even handwriting. The proposed research explores the source of this tacit, structural knowledge and how it comes to influence our performance in classification, recognition, and recall, and how it influences our and other animal's biases and preferences. Previous research suggests that the knowledge that is the source of these influences is not in the form of high-level abstractions such as rules or prototypes, but rather remains latent in and distributed over our memory for specific episodes of experience. These memories are recruited collectively in response to a similar current event to provide a feeling of familiarity or "rightness" or a fluidity to our responding to events consistent with the structure of those we've experienced in similar situations. This retrieval also provides for the ability to detect when an event is inconsistent with structure, even if we are unable to articulate how we "know" it. The proposed research explores a model of such memory and retrieval processes that would allow for this implicit learning and use of structure in both people and animals. The model is in the form of artificial neural network applied to low-level representations of the individual events, such as pixel-maps of images or digitised sounds.