In his 1955 address to the National Academy of Sciences, Richard Feynman delineated three key ways in which he saw science as having value (Feynman, 1955). One of these ways was the simple “intellectual enjoyment which some people get from reading and learning and thinking” (p. 13). For many scientists, there is intrinsic value in simply coming to understand how things work. They feel a certain joy when aspects of the world that previously seemed completely mysterious or idiosyncratic become less so. And this is true regardless of how the knowledge is eventually put to use. Yet it is inarguably the case that those eventual uses represent the greatest long-term value of science to our broader society. As Feynman said in discussing this second way that science has value, science is important because it “enables us to do all kinds of things and to make all kinds of things” (p. 13). In other words, increasing scientific understanding of a domain provides for the increasing possibility that we can apply some degree of control in the domain. Science offers the promise that we can manipulate, and thus potentially master, our circumstances. This core notion certainly permeates the behavioral sciences. Throughout the literature, one consistently sees manifestations of the idea that if we come to truly understand the mechanics by which human abilities, skills, knowledge, and other life outcomes emerge, then we might be able to purposefully intervene so as to alter those outcomes for the better. And although we are absolutely (very, very) far from mastering our circumstances in this domain, there are at least many reasons to be hopeful that such goals will eventually be within our reach. Such reasons for optimism include, for example, promising and ever-growing bodies of research on behavioral interventions meant to increase mental health and well-being (Creswell, 2017; Davidson & Dahl, 2018), interventions meant to decrease bias and prejudicial actions (Lemmer & Wager, 2015; Paluck & Green, 2009), interventions meant to increase cognitive and perceptual functioning (Au et al., 2015; Bediou et al., 2018; Deveau, Jaeggi, Zordan, Phung, & Seitz, 2014), and interventions in the educational sphere, such as those to promote reading abilities (Bus & van IJzendoorn, 1999; Kim & Quinn, 2013). Yet in considering previous work, as well as in evaluating the potential of future work, it is critical to recognize that in most cases of human behavior, truly “doing good” necessitates that the effects of interventions meet at least two key criteria: (a) The impact of the given intervention needs to generalize reasonably broadly and (b) the impact of the given intervention needs to be enduring. If the impact of an intervention is exceedingly narrow, or if the positive impact lasts for only a short period of time, this will obviously reduce the real-world good that will be realized from the intervention. It is therefore somewhat unfortunate that the field of human learning has consistently run into significant obstacles on both key fronts—generalization and persistence.

中文翻译：

---

为现实世界做好事的干预措施：泛化和持久性

在 1955 年对美国国家科学院的演讲中，理查德·费曼描述了他认为科学具有价值的三种关键方式（费曼，1955 年）。其中一种方式是简单的“一些人从阅读、学习和思考中获得的智力享受”（第 13 页）。对于许多科学家来说，简单地了解事物的运作方式具有内在价值。当以前看起来完全神秘或独特的世界的各个方面变得不那么如此时，他们会感到某种喜悦。无论最终如何使用知识，这都是正确的。然而，毫无疑问，这些最终用途代表了科学对我们更广泛社会的最大长期价值。正如费曼在讨论科学具有价值的第二种方式时所说，科学很重要，因为它“使我们能够做各种各样的事情并制造出各种各样的事情”（第 13 页）。换句话说，对一个领域的科学理解的增加为我们在该领域应用某种程度的控制提供了越来越大的可能性。科学提供了我们可以操纵并因此可能掌握我们的环境的承诺。这个核心概念肯定渗透到行为科学中。在整个文献中，人们始终看到这样一种观点的表现，即如果我们真正理解人类能力、技能、知识和其他生活结果出现的机制，那么我们可能能够有目的地进行干预，以改变这些结果更好。虽然我们绝对（非常，非常）远未掌握我们在这个领域的情况，至少有许多理由希望这些目标最终会在我们的范围内。乐观的原因包括，例如，关于旨在增加心理健康和福祉的行为干预的有希望且不断增长的研究机构（Creswell，2017；戴维森和达尔，2018），旨在减少偏见和偏见行为的干预（ Lemmer & Wager, 2015; Paluck & Green, 2009)，旨在提高认知和知觉功能的干预措施 (Au et al., 2015; Bediou et al., 2018; Deveau, Jaeggi, Zordan, Phung, & Seitz, 2014)，以及教育领域的干预措施，例如提高阅读能力的措施（Bus & van IJzendoorn, 1999; Kim & Quinn, 2013）。然而，在考虑以前的工作以及评估未来工作的潜力时，重要的是要认识到，在大多数人类行为案例中，真正“做好事”需要干预措施的效果至少满足两个关键标准：(a) 给定干预措施的影响需要合理广泛地概括和 (b)特定干预措施的影响需要持久。如果干预的影响非常狭窄，或者积极影响只持续很短的时间，这显然会降低干预所带来的现实利益。因此，有点不幸的是，人类学习领域一直在两个关键方面——泛化和持久性——遇到重大障碍。(a) 特定干预措施的影响需要合理广泛地概括，并且 (b) 特定干预措施的影响需要持久。如果干预的影响非常狭窄，或者积极影响只持续很短的时间，这显然会降低干预所带来的现实利益。因此，有点不幸的是，人类学习领域一直在两个关键方面——泛化和持久性——遇到重大障碍。(a) 特定干预措施的影响需要合理广泛地概括，并且 (b) 特定干预措施的影响需要持久。如果干预的影响非常狭窄，或者积极影响只持续很短的时间，这显然会降低干预所带来的现实利益。因此，有点不幸的是，人类学习领域一直在两个关键方面——泛化和持久性——遇到重大障碍。