启发式搜索给神经网络_神经科学如何支持UX启发式

启发式搜索给神经网络

重点 (Top highlight)

Interaction and UX designers have long known and used heuristics to guide the creation of a user-friendly interface. We know empirically that these principles work, and they make “common sense”. These heuristics themselves aim to make interface “intuitive”, but common sense and intuitive are concepts hard to define precisely.

交互和UX设计人员早就知道并使用启发式技术来指导用户友好界面的创建。 我们凭经验知道这些原则行之有效，并且它们具有“常识”。 这些启发式方法本身旨在使界面“直观”，但是常识和直观是很难精确定义的概念。

Research in neuroscience has proposed a new interpretation for how the brain works and how decisions are made, namely the Bayesian brain. When comparing the known heuristics and this new interpretation model, we find numerous similarities. Understanding the Bayesian brain should help designers go beyond vague concepts of “common sense” and “intuitive” to understand why their heuristics work or not.

神经科学研究已经提出了一种关于大脑如何运作以及如何做出决策的新解释，即贝叶斯大脑。 当比较已知的启发式方法和这种新的解释模型时，我们发现了许多相似之处。 理解贝叶斯大脑应有助于设计师超越模糊的“常识”和“直觉”概念，以了解其启发式方法为何起作用或不起作用。

我们的大脑是统计学家 (Our brain is a statistician)

The Bayesian brain is a way to describe how our perception works. We receive millions of stimuli each second from all our sensory nerves. Most of these signals, as numerous as they are, are still ambiguous. When we receive a certain amount/shape of light in our retina, multiple objects outside could produce this same pattern. However, most of the time, we don’t consider all the possible objects, we identify instantaneously the right object.

贝叶斯大脑是描述我们的感知如何运作的一种方式。 我们每秒从我们所有的感觉神经接收数百万个刺激。 这些信号中的大多数(尽管数量众多)仍然是模棱两可的。 当我们在视网膜中接收到一定数量/形状的光时，外面的多个物体可能会产生相同的图案。 但是，在大多数情况下，我们不会考虑所有可能的对象，而是会立即识别出正确的对象。

The Bayesian brain theory explains how, behind the scenes, the brain operates: it uses probability and equations similar to Bayes theorem to give us only one a single answer. Even though we are aware of a single object, our brain has processed all the possibilities and determined the most likely one, using Bayes theorem, and presents this most likely possibility as the only one. Several experiences show the power of prediction of this theory, and how the physical neuron network can actually perform these kinds of calculations.

贝叶斯脑理论解释了大脑在幕后如何运作：它使用类似于贝叶斯定理的概率和方程式，只给我们一个答案。 即使我们知道单个对象，我们的大脑也已经处理了所有可能性，并使用贝叶斯定理确定了最可能的可能性，并将这种最可能的可能性表示为唯一的可能性。 几项经验证明了该理论的预测能力，以及物理神经元网络如何实际执行此类计算。

The Bayesian brain theory makes a distinction between the unconscious, statistical treatment of stimuli, which includes many alternatives with their own probability and our awareness of a situation, which is (should be) unique. Bayes theorem uses two factors to calculate the posterior probability / make an inference: the known probability of a given situation, and previous knowledge. The decision in this model is based on the resulting inference and a function of gain/loss.

贝叶斯大脑理论区分了无意识的，对刺激的统计处理(包括许多具有自己可能性的替代方法)和我们对情况的认识(这是(应该)独特的)。 贝叶斯定理使用两个因素来计算后验概率/进行推论：给定情况的已知概率和先前知识。 该模型中的决策基于结果推断和损益函数。

为什么我们的大脑使用概率对UX设计人员如此重要？ (Why is the fact that our brains use probability so important for UX designers?)

First, this emphasizes the fundamental ambiguity of all types of signals, including the interfaces we design. Optical illusions work by leveraging this signal ambiguity and the processing done by the brain to present to us the most likely solution(s) based on our previous experience. They show us how our perception is a brain construct rather than an exact reflection of reality. Optical illusions are only extreme cases of how we constantly transform our sensations into perceptions. Thus, designers need to reconcile themselves with the inherent ambiguity of all constructs (while trying to minimize it) and cease to blame “dumb” users for their misunderstanding.

首先，这强调了所有信号类型(包括我们设计的接口)的基本歧义。 视错觉通过利用信号的模糊性和大脑进行的处理来发挥作用，从而根据我们以前的经验为我们提供最可能的解决方案。 它们向我们展示了我们的感知是大脑的构造，而不是现实的精确反映。 错觉只是我们不断将感觉转变为感知的极端情况。 因此，设计人员需要使自己与所有构造的固有歧义保持一致(同时尝试将其最小化)，并不再因误解而责怪“笨拙”的用户。

Second, Bayes theorem gives us clues on how to minimize this fundamental ambiguity, and make the right interpretation stand out among all the possible ones. Luckily, these clues map exactly to the already known heuristics, but now we have an idea why they work so well.

其次，贝叶斯定理为我们提供了有关如何最大程度地减少这种基本歧义以及使正确的解释在所有可能的解释中脱颖而出的线索。 幸运的是，这些线索正好映射到了已知的启发式方法，但是现在我们有了一个想法，为什么它们如此有效。

Third, we can now define more clearly what has long been elusive: the concept of “intuitive” itself and understand that there might not be any kind of universal intuitive system, but that what we consider as “intuitive” is mostly previously learned rules, so well internalized that we consider them as “natural”.

第三，我们现在可以更清楚地定义长期以来难以捉摸的东西：“直觉”本身的概念，并了解可能没有任何一种通用的直觉系统，但是我们认为“直觉”主要是以前学过的规则，内部化程度很高，我们认为它们是“自然的”。

减少歧义 (Minimizing ambiguity)

The Bayesian theorem produces accurate inferences based on two factors: the known probability of a given situation and the prior probability of the alternative that is considered (based on our previous experience).

贝叶斯定理基于两个因素产生准确的推论：给定情况的已知概率和所考虑的替代方法的先验概率(基于我们的先前经验)。

As a designer, we need to use each of these aspects, so they are all aligned in the same direction:

作为设计师，我们需要使用以下所有方面，因此它们都朝着相同的方向排列：

- The given situation is the stimuli: the visual/audio/text signals must all convey the same message: green checkmark with a high pitch sound: positive/red cross with a loud low pitch sound = negative.

-给定的情况是刺激：视觉/音频/文本信号都必须传达相同的消息：高音高的绿色复选标记：高音低音的正/红叉=负。

Of course, not all situations are so simple, but even things that simply can be messed up. Some card payment terminals emit a low pitch sound while displaying a “payment accepted” message, resulting in confusion in the client, and the need for the cashier to reassure them that all went well. This might seem minor, but I bet the few seconds wasted everywhere this type of terminal is used add up to significant productivity loss.

当然，并非所有情况都如此简单，但即使是简单的事情也可能被弄乱。 一些卡支付终端会在显示“已接受付款”消息时发出低调的声音，从而导致客户产生混乱，并且需要收银员向他们保证一切进展顺利。 这看似微不足道，但我敢打赌，使用这种类型的终端到处都浪费了几秒钟的时间，这会大大降低生产率。

- Prior probability represents what we call in design mental models. They are the results of the successful inferences made by the user in the past, often associated with a causality system. We learned to associate green with something valid or positive and red with errors or wrong way signal.

-先验概率代表我们在设计思维模型中所说的。 它们是用户过去进行的成功推断的结果，通常与因果关系系统相关联。 我们学会了将绿色与有效或肯定的事物关联起来，将红色与错误或错误的信号关联起来。

These should be a well-known quantity for designers: through competitive research, we know other existing systems and interfaces; through user research, we know which tools our users manipulate, what their mental models are, etc. The design should leverage this known quantity to make the new interface or system familiar, reinforcing previously learned patterns.

对于设计师来说，这些应该是众所周知的数量：通过竞争研究，我们知道其他现有的系统和接口； 通过用户研究，我们知道用户使用了哪些工具，他们的心理模型是什么，等等。设计应利用这一已知数量来使新界面或系统熟悉，从而增强以前学习的模式。

Then, the user must make a decision. According to the decision-making theory using Bayesian interpretation, we make our decision based on the inference made using Bayes theorem and a gain/loss function.

然后，用户必须做出决定。 根据使用贝叶斯解释的决策理论，我们基于使用贝叶斯定理和损益函数得出的推论做出决策。

Perception-Action loop according to Bayesian perspective — Source: Ernst & Bulthoff, 2004; quoted in S. Dehaene, 2012

根据贝叶斯观点的知觉-行动循环-来源：Ernst＆Bulthoff，2004； 引用于S.Dehaene，2012年

This maps perfectly onto other UX heuristics: clearly state the outcome of an action and keep the user informed on the system status. Having clear information on the system status and outcome will help the user make better, quicker decisions.

这完美地映射到其他UX启发式方法：清楚地说明操作的结果，并使用户了解系统状态。 具有有关系统状态和结果的清晰信息将帮助用户做出更好，更快的决策。

直觉是什么意思？ (What does intuitive mean?)

When we act seamlessly, without even thinking, is when we say that things are “intuitive”. Based on what we said previously, we have a better grasp of what makes things/interfaces feel intuitive or not.

当我们什至没有思考就无缝地行动时，就是说事情是“直觉的”。 根据我们之前所说的，我们可以更好地理解使事物/界面感觉直观与否的原因。

We make quick, almost imperceptible decisions when all the mentioned factors align perfectly: signals, mental model, know status and outcome, because our Bayesian brain has calculated a very high probability for the right alternative, and very low probability for all other possibilities.

当所有上述因素完美契合时，我们会做出快速，几乎不可察觉的决策：信号，心理模型，状态和结局，因为我们的贝叶斯大脑计算出正确选择的可能性很高，而其他所有可能性的可能性却很小。

Something is not intuitive when the multiple alternatives processed by our brain cease to operate at an unconscious level and make their way into our awareness. When does that happen? When several alternatives have similar probabilities. Then our well-trained Bayesian brain lets us become aware of several possibilities, the most likely ones, creating confusion.

当大脑处理的多种选择停止在无意识的水平运行并进入我们的意识时，有些事情就不直观了。 什么时候发生？ 当几个备选方案具有相似的概率时。 然后，我们训练有素的贝叶斯大脑使我们意识到几种可能性，最有可能的情况是造成混乱。

Thus, for things to be intuitive, there needs to be one clear winner among all the alternatives. However, as the probability is not only a function of the stimuli, but also of previous knowledge, there is no way to create an interface that would be universally intuitive. We shouldn’t say that things or interfaces ‘are’ intuitive by themselves. We can only say that they ‘feel’ intuitive for this category of users. Each person has her own previous experience, that might influence differently her perception of a given interface.

因此，为了使事情直观，所有替代方案中都必须有一个明确的赢家。 但是，由于概率不仅是刺激的函数，而且还是以前的知识的函数，因此无法创建通用直观的界面。 我们不应该说事物或界面本身就是“直观的”。 我们只能说他们对这类用户“感觉”直观。 每个人都有自己的先前经验，这可能会不同地影响她对给定界面的看法。

As an example, there is no answer to the question whether Apple or Microsoft is more intuitive: users of each system have learned patterns of usage, which became “intuitive” to them. Making the switch to a system with different patterns becomes very costly. If you eliminate the notion of a naturally intuitive system and replace it with learned patterns, then there is no intrinsic superiority of a system compared to another.

例如，没有答案是苹果还是微软更直观：每个系统的用户都已经学会了使用模式，这对他们来说是“直观的”。 切换到具有不同模式的系统变得非常昂贵。 如果您消除了自然直观系统的概念并将其替换为学习过的模式，那么与其他系统相比，该系统就没有内在的优势。

As a designer, this means that you need to know the audience you’re designing for intimately enough that you know their mental models. This is typically achieved in one of two ways:

作为设计师，这意味着您需要充分了解要设计的受众，以了解他们的思维模式。 通常通过以下两种方式之一来实现：

- Designers who are so immersed in a given “culture” that they share most patterns with their users. These designers will use the same patterns in their design, without even thinking about it. They might be great designers in a specific context, but they won’t perform so well in another, because they not only ignore the patterns of this new context, they don’t have an explicit understanding/conceptualization of what make things “intuitive” for some people and them and not others (these are the designers who say that users are dumb).

-如此沉迷于给定“文化”以至于与用户共享大多数模式的设计师。 这些设计师将在设计中使用相同的模式，甚至无需考虑它。 他们可能在特定的环境中是出色的设计师，但在其他环境中表现不佳，因为他们不仅忽略了这种新环境的模式，而且对使事物“直观”的内容没有明确的理解/概念化。对于某些人和他们而不是其他人(这些设计师说用户很愚蠢)。

- Designers who research and identify the mental models of their expected users. Of course, they also have their own learned patterns and are not completely immune to the illusion of things being intuitive, but they reflect on their own learned habits, acknowledge their preferred patterns, so they can more easily distance themselves and learn other ways of doing things. They don’t trust their “intuition” so much and prefer hard evidence.

-研究和确定预期用户心理模型的设计师。 当然，他们也有自己的学习模式，并不能完全避免直观事物的幻觉，但是他们会反思自己的学习习惯，承认自己的偏好模式，因此他们可以更轻松地与自己保持距离并学习其他做事方式东西。 他们不太相信自己的“直觉”，而是喜欢有力的证据。

For more on the Bayesian brain, see:

有关贝叶斯大脑的更多信息，请参见：

• a series of lecture by Stanilas Dehaene, College de France, 2012 (in French)

  Stanilas Dehaene的系列讲座，法国学院，2012年(法语)

• Amy Perfors, Joshua B. Tenenbaum, Thomas L. Griffiths, Fei Xu, A tutorial introduction to Bayesian models of cognitive development , Cognition, Volume 120, Issue 3, September 2011, Pages 302–321.

  Amy Perfors，Joshua B. Tenenbaum，Thomas L. Griffiths，徐飞， 认知发展的贝叶斯模型教程简介 ，《认知》，第120卷，第3期，2011年9月，第302-321页。

• Marc O.Ernst, Heinrich H.Bülthoff, Merging the senses into a robust percept, Trends in Cognitive Science, Volume 8, Issue 4, April 2004, Pages 162–169.

  马克·厄恩斯特(Marc O.Ernst)，海因里希·H·伯索夫(Heinrich H.Bulthoff)，《 将感官融合为一个强有力的感知》 ，《认知科学趋势》，第8卷，第4期，2004年4月，第162-169页。

翻译自: https://uxdesign.cc/how-neuroscience-supports-ux-heuristics-76d94d977d90

启发式搜索给神经网络