A framework and toolkit for understanding impulsive action

它的重要性是什么？从经济学到健康和福祉，从精神病学到社会障碍，冲动是理解许多领域个人行为的关键因素。它与许多问题行为有关，如酗酒、肥胖、多动症、酗酒或暴饮暴食、滥用药物、赌博、侵略、暴力或性虐待以及危险的性行为。在理解冲动的不同方面——以及如何最好地衡量它们——方面取得的进展将立即应用于改变这些行为的尝试。我们关注的焦点。测量冲动性的任务分为两类：一类是测量控制或抑制反应的表现，另一类是测量对风险或即时回报的偏好。我们专注于前一类：反应控制。需要什么？响应控制任务越来越受欢迎，因为它们被认为是挖掘各种问题行为背后的基本机制（如上所述）。然而，进展是低效的，因为我们缺乏一个清晰的框架来理解为什么人们在这些任务中存在差异，以及这些差异对现实世界的行为实际上意味着什么。基本的结论是，人们抑制习惯性反应的能力是不同的，但这没有解释为什么人们的表现在不同的测试这种能力的方法之间没有很好的相关性。随着越来越多的研究使用这些方法，研究资金被浪费了，因为在理解上没有相应的进步。我们如何解决这个问题？1. 支撑。部分问题在于，这些任务的一些基本工作尚未完成。大多数任务都是从群体平均水平的研究中转移过来的，在这种研究中，个体差异被视为无关紧要。因此，我们甚至不知道哪些任务最能反映稳定的特征（以及需要多少次试验才能做到这一点）。我们也不知道，对于泛化反应控制来说，重要的是人们多快掌握任务，还是他们在熟练完成任务后的表现如何。2. 应用计算建模提供一个新的框架。部分问题在于，这些任务评估的是我们真正想要测量的东西的替代品——导致冲动行为的潜在认知机制。至少有两个与冲动相关的独立因素会影响这些任务的表现——一个人有多谨慎，以及他们有多善于抑制错误的反应（或选择正确的反应）。当这些因素相互作用时，结果变得难以解释，并且在这种情况下，使用跨不同任务的相关模式来提取潜在变量的方法已被证明是不够的。相反，我们将使用行动决策的计算模型，它充分利用了行为数据的丰富性，可以帮助我们理解复杂和违反直觉的结果是如何发生的。3. 通知干预的工具。为了成功地帮助有各种有害冲动的人（无论是在成瘾、暴力、债务、学业表现或许多其他领域的背景下），研究人员、专业心理学家、经济学家、临床医生等，需要将最好的科学知识提炼成可以在没有专业知识（例如，计算模型）的情况下轻松管理的工具。我们将把我们的发现转化为这样的工具：基于模型的冲动性工具包（ModBIT）。

What is the importance? Impulsivity is a key factor in understanding individual behaviour in many spheres, from economics to health and wellbeing, from psychiatry to social disorder. It is associated with many problem behaviours, such as alcoholism, obesity, ADHD, binge drinking or eating, drug misuse, gambling, aggression, violent or sexual abuse and risky sex. Advances in understanding the different aspects of impulsivity - and how best to measure them - will have immediate application for attempts to modify these behaviours. Our focus. Tasks that measure impulsivity divide into two types: those measuring performance in the control or inhibition of responses, and those measuring preferences for risk or reward immediacy. We concentrate on the former category: response control. What is the need? Response control tasks are increasingly popular, since they are thought to tap basic mechanisms underlying a variety of problematic behaviours (as mentioned above). However, progress is inefficient because we lack a clear framework for understanding why people differ in these tasks, and what such differences actually mean for real-world behaviour. The basic conclusion is that people differ in their ability to inhibit habitual responses, but this leaves unexplained why people's performance does not correlate very well across different ways of testing this ability. As increasing numbers of studies use these measures, research money is wasted because there is no commensurate advance in understanding.How will we solve this problem? 1. Underpinning. Part of the problem is that some essential work on these tasks has not been done. Most of the tasks have been transferred from studies of group averages, where individual differences are treated as irrelevant. Thus we do not even know which tasks best reflect stable traits (and how many trials are needed to do so). Nor do we know whether the important thing for generalisable response control is how quickly people catch on to the task, or how well they perform once the task is well practised. 2. Applying computational modelling to provide a new framework. Part of the problem is that the tasks assess surrogates of what we really want to measure - the underlying cognitive mechanisms that contribute to impulsive behaviour. There are at least two separate factors related to impulsivity that influence performance on these tasks - how cautious someone is and how good they are at inhibiting the wrong response (or selecting the right one). When these factors interact, the results become difficult to interpret, and the approach of using patterns of correlation across different tasks to extract underlying variables has proved insufficient in this case. We will instead use computational models of action decision, which utilise the full richness of the behavioural data, and can help us to understand us how complex and counterintuitive results occur. 3. Tools to inform intervention. In order to successfully help people with different kinds of detrimental impulsivity (whether in the context of addiction, violence, debt, school performance or many other areas), researchers, professional psychologists, economists, clinicians, etc., need tools that distil the best scientific knowledge into packages that can be easily administered without specialist knowledge (e.g., of computational modelling). We will translate our findings into such tools: the Model-Based Impulsivity Toolkit (ModBIT).