Choices of Evidence. Tacit philosophical assumptions in the debates within the Campbell Collaboration

基于证据的政策现在在英国和美国被强制执行，并且越来越多地在欧洲，在国际，国家和地方层面。坎贝尔协作是一个独立的、国际性的、非营利性组织，以医学领域的科克伦协作为蓝本，旨在为“决策者提供基于证据的信息，使他们能够就干预措施在社会、行为和教育领域的影响做出明智的决定”（www.campbellcollaboration.org）。然而，在坎贝尔合作组织内部及其批评者之间，关于在以证据为基础的政策和实践中，什么是可接受的证据，以及如何判断其作为支持特定假设的力度，存在着重大争论。本项目聚焦于儿童社会福利和教育的案例研究，旨在通过分析隐含的哲学假设来澄清和帮助解决这些争议。鉴于知识基础的水平和主题的复杂性，大多数争议似乎源于对应该提供什么证据和可以提供什么证据的不同要求。我们将考察一些公认的但有争议的区别，我们推测这些区别会使辩论更加混乱。通常，对手似乎无法理解对方的立场，因为他们对社会世界的本质或科学方法隐含地采用了不同的哲学假设。我们将讨论六个关键问题。第一，定量和定性数据：两者之间的区别被广泛接受，前者通常被认为比后者更有效和可靠。但在社会福利和教育领域，从业者的推理是基于关于人们行为的部分和模糊的数据。定量数据和定性数据究竟在哪些方面不同，在哪些方面前者据称更可靠。一个相关的问题是客观和主观之间的区别，这在实践中成为一般知识和地方知识之间的区别，即方法论/专门知识和更以判断为基础的办法之间的区别。在不牺牲本地知识的情况下，能否获得研究方法的一般知识和专业知识的优势？第三，社会福利领域提出了以下问题：谁来决定什么是社会干预的解决方案，以及如何处理/裁决相互竞争的诉求。在一个涉及社会中最弱势群体的学科领域，这个问题特别重要。第四个问题是，明确的方法层次是否会无意中导致问题的层次，从而不可避免地优先考虑那些可以由排名较高的方法研究的问题。第五个问题说明了“确凿”证据方法（从肯定的结果中推断出结论）和“担保”证据方法（由被认为是令人满意的、可能相关的证据支持的结论）之间的区别。确凿的证据需要如此多的假设和条件，以至于我们担心，它的应用变得非常有限。如何处理适用范围和结论确定性之间的权衡？最后一个问题是关于封闭式和开放式研究方法的区别。前者，如随机对照试验，有内置的设备来检查应用该方法所需的假设。后者，如Bayes-net，需要求助于其他设备。最好的证据是否需要通过封闭的方法提供，或者更大的范围是否可以接受？这些分析的结果将为阐明符合实际研究要求和需要的科学哲学提供基础。我们的目标是提出一种证据理论，它在理论上是合理的，在现实中是可执行的，从而有助于坎贝尔项目的成功发展。

Evidence-based policy is now mandated throughout the UK and the USA, and increasingly in Europe, at international, national and local levels. Modelled on the Cochrane Collaboration in medicine, the Campbell Collaboration is an independent, international, non-profit organization which has been set up to provide 'decision-makers with evidence-based information to empower them to make well-informed decisions about the effects of interventions in the social, behavioural and educational arenas' (www.campbellcollaboration.org). There are, however, major debates, both within the Campbell Collaboration and with its critics, about what counts as acceptable evidence in evidence-based policy and practice and how its strength as support for a particular hypothesis can be judged.This project, focusing on the cases studies of child social welfare and education, aims to clarify and help resolve these controversies by analysing the philosophical assumptions that tacitly inform them. Most of the controversies appear to originate from different demands put on what evidence should deliver and what it can deliver, given the level of the knowledge base and the complexity of the subject matter. We shall examine a number of well-established and yet disputable distinctions that we conjecture add confusion to the debate. It often seems that opponents are failing to understand each other's positions because they are implicitly adopting different philosophical assumptions about the nature of the social world or of scientific method.We shall address six key issues. First, quantitative and qualitative data: a distinction between the two is widely accepted with the former being generally considered more valid and reliable than the latter. But in social welfare and education, practitioners reason on the basis of partial and ambiguous data about people's behaviour. In what ways precisely do quantitative and qualitative data differ and in what ways is the former purportedly more reliable. A linked issue is the distinction between objective and subjective that, in practice, becomes a distinction between general and local knowledge, namely between methodology/expertise and more judgment-based approaches . Can the advantages of the general knowledge and expertise in research methods be gained without sacrificing local knowledge?Third, the area of social welfare raises questions about who decides what counts as a solution in social interventions and how competing claims should be handled/adjudicated. In a subject area that deals with the least powerful groups in society, this issue is of special importance.A fourth issue is whether a clear hierarchy of methods inadvertently leads to a hierarchy of questions, inevitably prioritising those questions that can be studied by the higher ranked methods. The fifth question addresses the difference between 'clinching' methods of evidence (their conclusions are deduced from a positive result) and ''vouching' ones (conclusions upheld by evidence deemed to be satisfactory, possibly relevant etc. Clinching evidence requires so many assumptions and conditions that its application becomes, we fear, very limited. How should the trade-off between scope of application and certainty of conclusion be managed?The final issue questions the difference between closed and open research methods. The former, such as RCTs, have built-in devices for checking the assumptions that are needed to apply the method. The latter, such as Bayes-net,require appeal to other devices. Need best evidence be supplied by closed methods or can a wider range be acceptable?The outcome of these analyses will provide the basis for articulating a philosophy of science that meets the requirements and needs of practical research. We will aim at suggesting a theory of evidence that is both theoretically sound and realistically enforceable and thus contributes to the successful development of the Campbell project.