Module 03 · Impression Formation印象形成
How Do We Combine Traits Into an Overall Impression?特质整合与整体印象
Overview 概述
Up to this point, we’ve discussed how people can represent their knowledge about others. That is, they can use categories to try to get a large amount of (perhaps inaccurate) information about others, or they can put more effort into determining a person’s individual traits. However, knowing that someone is smart, sarcastic, and annoying is still a distance away from knowing whether we like that person or not. In this section, we’re going to focus on information integration theory—a prescriptive theory that addresses how we *should *mathematically (i.e., rationally) combine the disparate sources of information we have about a person into an overall impression.
到这一点为止，我们讨论了人们如何表达他们对别人的知识。也就是说，他们可以使用类别来尝试获取大量（可能不准确）的信息关于别人，或者他们可以投入更多努力来确定一个人的个别特征。然而，知道某人聪明、讽刺和讨厌，仍然与我们是否喜欢这个人有距离。在本节中，我们将关注信息整合理论——这是一种规定性理论，它解决了我们应该如何数学上（即理性地）将我们关于一个人的不同信息来源组合成一个整体印象的问题。
Initial Research 初步研究
In order to start modeling how people combined traits into an overall impression, social psychologist Norman H. Anderson, the creator of information integration theory, asked 100 college students to judge how desirable they found traits on a scale from 0 to 6 (Anderson, 1968). For example, you might say that the trait of intelligence could be scored a 6 (very positive) and the trait of annoying should be scored a 3 (slightly negative). Anderson had people do this rating task over and over again until he had average ratings of 555 trait adjectives—a huge amount!
为了开始模拟人们如何将特质组合成一个整体印象，社会心理学家诺曼·H·安德森，信息整合理论的创造者，让 100 名大学生对特质在 0 到 6 的量表上他们认为有多吸引人的进行评分（安德森，1968）。例如，你可能会说，智能这一特质可以评分为 6（非常积极），而令人讨厌这一特质应该评分为 3（略微消极）。安德森让人们反复进行这项评分任务，直到他得到了 555 个特质形容词的平均评分——这是一个巨大的数量！
Anderson realized how rare it is to know only a single fact about a person. So, he started trying to discover how people combined these traits into an overall impression. Because he had desirability ratings for so many traits, he started using them in combinations to see if he could understand how the scores of the individual traits predicted the overall evaluation of a person when they were combined. For example, in Anderson’s list, the trait of being honest was rated as the second most desirable, with an average rating of 5.55 out of 6, whereas the trait of being persistent was rated as the 232nd most desirable trait, with an average rating of 3.47 out of 6. How can we use these ratings to figure out how we rate someone who is honest *and *persistent?
安德森意识到仅了解一个人的一个事实是多么罕见。因此，他开始尝试发现人们如何将这些特质组合成一个整体印象。因为他有很多特质的吸引力评分，所以他开始将它们组合起来，看看他是否能够理解单个特质的评分如何预测当它们组合在一起时对一个人的整体评价。例如，在安德森的列表中，诚实这一特质被评为第二受欢迎，平均评分为 6 分中的 5.55 分，而坚持不懈这一特质被评为第 232 个最受欢迎的特质，平均评分为 6 分中的 3.47 分。我们如何利用这些评分来了解一个既诚实又坚持不懈的人的评分呢？
Mathematical Models of Impression Formation 数学印象形成模型
There are two potential ways that Anderson considered to model how people form impressions: they can add all of the trait ratings together (called an additive model), or they can average them (called an averaging model) to form an overall impression. While this sounds like a trivial difference (doesn’t averaging involve adding all the ratings together anyway?), it has an important implication.
有两种潜在的方法，安德森考虑用来模拟人们形成印象的过程：他们可以将所有特质评分相加（称为加法模型），或者将它们平均（称为平均模型）以形成一个总体印象。虽然这听起来像是一个微不足道的差异（难道平均不涉及将所有评分相加吗？），但它有着重要的含义。
To simplify this explanation, let’s assume that the ratings of traits are on a –3 to +3 scale, with negative traits (e.g., evil) tending to have scores toward –3, neutral traits (e.g., average height) tending to be close to 0, and positive traits (e.g., generous) tending to have scores toward +3. Now, let’s assume we know two moderately positive traits about a person (e.g., they’re friendly and funny), each of which scores a +2 on the scale. Now, let’s imagine we include a third trait in our knowledge of this person, one that is slightly positive (e.g., they’re punctual). The additive and averaging models make different predictions about what effect including this third trait will have.
为了简化这个解释，让我们假设特质的评级在-3 到+3 的范围内，负向特质（例如，邪恶）倾向于得分接近-3，中性特质（例如，平均身高）倾向于接近 0，而正向特质（例如，慷慨）倾向于得分接近+3。现在，假设我们了解这个人的两个适度正向特质（例如，他们友好且幽默），每个特质在量表上得分为+2。现在，让我们想象我们在这个人对这个人的了解中增加第三个特质，这个特质略微正向（例如，他们守时）。加法和平均模型对包括这个第三个特质将产生什么影响的预测不同。
The additive model suggests that including another positive trait, regardless of how positive it is, will lead to our having a more positive evaluation of the person. This makes intuitive sense—why wouldn’t we have a more positive impression of a person if we just learned a new positive fact about them? The averaging model suggests that our impression of the person would actually get more negative. Specifically, if we have two traits that average a +2 on our scale and then average in another +1, that would bring our overall evaluation of the person closer to 0. That is, while our rating would still be positive, overall, it would become more negative (or, at least, less positive) than it was before.
加法模型表明，无论另一个积极特质有多么积极，包括它都会导致我们对这个人的评价更加积极。这从直觉上来说是合理的——如果我们只是了解到关于他们的一个新积极事实，为什么我们不会对这个人有更积极的印象呢？平均模型表明，我们对这个人的印象实际上会变得更加负面。具体来说，如果我们有两个特质在我们的量表上平均为+2，然后平均加入另一个+1，这将使我们对这个人的总体评价更接近 0。也就是说，尽管我们的评分总体上仍然是正面的，但与之前相比，它将变得更加负面（或者至少不那么积极）。
Which of these models does research suggest does a better job of modeling people’s actual thoughts, feelings, and behaviors? It turns out that, most frequently, adding slightly positive traits to a description of a person makes the rating of them more moderate, suggesting that the averaging model does a better job of modeling impression formation. Note that this same moderating influence of adding less extreme traits also works for negative traits. Learning that someone is cruel will lead you to have a very negative impression of them, but learning that they’re cruel and have a bad taste in fashion will make your impression of the person less negative. It’s not necessarily rational, but the averaging model seems to reflect how people’s minds work.
这些模型中，研究建议哪一个在模拟人们的实际想法、情感和行为方面做得更好？结果发现，在描述一个人时，加入略微积极的特质，通常会使得对其的评价更加适中，这表明平均模型在模拟印象形成方面做得更好。请注意，加入不那么极端的特质所产生的这种调节影响也适用于负面特质。了解到某人很残忍，会让你对他们产生非常负面的印象，但了解到他们既残忍又对时尚品味不佳，会让你对他们的印象不那么负面。这并不一定是理性的，但平均模型似乎反映了人们思维的工作方式。