Selective citation and the shaping of scientific knowledge: Citation network analysis and the diet-heart debate

点击翻译按钮获取中文摘要

My research is concerned with developing an understanding of how new scientific knowledge is disseminated amongst scientists, and how distortions introduced by dissemination impact on the construction and development of that knowledge. Scientists contribute to a particular evidence-base by publishing the results of their experiments, but in writing their papers they also use evidence produced by others. However, they are often selective in the evidence that they cite, and can differ in their interpretations of that evidence. Amongst the questions that I ask are: how objective are scientists when they cite evidence? Why do some findings and interpretations become widely known while others fade into obscurity? And how does the use and interpretation of past studies shape what is eventually taken to be scientific knowledge?My approach to these questions involves:- Systematic literature searches to capture all papers relevant to a specific scientific debate.- Mapping the citations between these papers, and analysing the network to understand the dynamics of evidence use. - Claim-specific network analysis to analyse the interpretation of specific papers, to evaluate their impact on scientific opinion, and to detect bias in evidence selection. - Main path analysis to identify studies that are involved in the mainstream of research - well-cited papers and their references to one another.This provides a framework by which to evaluate the dissemination of particular studies and findings and their impact on scientific development, including the prevalence of citation distortions and their impact on scientific opinion.I have applied this approach to understand the development of the conjecture that dietary fats are causally related to coronary heart disease in the scientific literature from 1908-1984. The main outcomes were (i) Scientists tended to cite literature recently published and which addressed similar research questions to the citing paper. (ii) A power-law approximated the citation distribution i.e., a relatively few core papers were cited disproportionately often. (iii) Findings that supported the link were cited more often than unsupportive findings, and their impact was amplified by the factors that underlie the power law distribution. (iv) One-sided citation bias, the act of selectively citing only evidence that supports a particular conclusion, was prevalent in reviews that examined the effectiveness of dietary treatment in those with heart disease. This bias impacted published conclusions; 82% of reviews supportive of such treatment cited only the single clinical trial supporting this conclusion while ignoring the three unsupportive trials in the literature. While I applied this approach to a specific case-study, its design and methods will be applicable to other cases. During this Fellowship I will publish my analysis of the diet-heart debate, and disseminate my findings to diverse audiences of scientists in different academic communities, to policy makers interested in evidence-based policy, and to the general public. I will use my expertise in quantitative science studies and citation bias to contribute to ongoing efforts to improve research integrity by contributing to the 'Responsible Metrics' group at Edinburgh University. I will spend 3-weeks visiting Leiden University, where I will present my research and form professional ties with Europe's leading centre for scientometrics and research evaluation. I will also engage the policy-making community through discussions with researchers at the Scottish Parliament and House of Commons in a session introducing researchers to the problem of citation bias and methods for its detection. Finally, I will begin a new research project focussing not on a historical debate but on a currently unfolding controversy, about the notion that the hormone "oxytocin" is specifically involved in social behaviour, with potential benefits for the treatment of autism.