Abstract: Due to the limitations of citation-based evaluation methods, as well as the complex citing motivations, scholars are inspired to explore new quantitative evaluation methods. In recent years, the rise of Altmetrics provides new measures to carry out quantitative academic impact evaluation. Currently, a massive number of studies have been conducted on related research scopes, which can be further categorized into three. The first category of the existent studies introduces this emerging domain by organizing, analyzing and clarifying the basic concepts and development stages of Altmetrics. Meanwhile, some scholars carried out comparison researches on the various Altmetric platforms about their purposes, data sources, metric indicators, evaluation strategies, output, and other additional functions, which can be seen as the second category. The third type of researches focus on Altmetric indicators′ application in article evaluation, and some of these studies indicate that the tradition bibliometric indicators should be utilized with the Altmetric indicators to increase the credibility of the results. Overall, the current Altmetrics studies show us a new path to carry out academic impact evaluation by applying the tradition bibliometric indicators and the Altmetrics indicators together.
Although fruitful results have been achieved in the Altmetrics study, further research is still needed to clarify some critical points. First of all, the contents and limitations of some Altmetric indicators need to be further clarified to ensure they will be utilized in a consensus and proper way. For instance, it is still unclear that to what extend the indicator “number of viewers online” represents the academic impact of an article, or in other word, how should we comprehend this Altmetric indicator? The second point needs further clarification is the relationship between various Altmetric data sources. As we know, multiple Altmetrics websites and platforms exist currently, but the disparities of their data sources, data collecting methods, indicators′ definition and update frequencies have not yet been clearly compared and introduced. The chaos caused by above unclarity affected the efficacy of Altmetric indicators in their application. Therefore, this research aims to propose an innovation framework to relief issues exist in the current Altmetrics-based academic impact evaluation caused by the above two points.
In accordance with the research purpose above, firstly we need to clarify the concepts of evidence chain. Evidence chain is a long-standing term in the domain of judicial criminal science, which is formed by more than two pieces of evidence (or evidence cluster) proofing each other mutually. The pieces of evidence or evidence cluster on an evidence chain are linked with each other via evidence connectors, which are usually forensic authentications or weapons in judicial investigations, and they also can be widely-accepted logic, social law and scientific knowledge in the general reasoning process. Most of the current research on evidence chain exists in the field of the judicial criminal investigation, although this concept also has been employed in the other research fields, the research about its potential application in the field of evaluation has not yet appeared.
Furthermore, causality is the core to understand the evidence chain, since it combines solo evidence into a whole and enables the evidence chain proof an assumption. In practice, the causal connection of an evidence chain should be understood from multiple levels: the true relationship between causality exists only in the real domain that human experience cannot directly perceive, but the specific cause and effect are in the actual domain and the empirical domain is related through a causal chain that is morphologically diverse and can be recognized by human experience. Therefore, the causal chain is a specific form of a specific effect of causality, which includes a series of specific causes of the results and logical relationships between the causes. In an open system, the number of causal chains is inexhaustible, even the cause of a single causal chain is infinitely separable. Therefore, in practice, people often simplify and construct a causal chain with considering factors such as cognitive purpose, cognitive ability, and cognitive habits.
Based on a specific causal chain, an evidence chain can be formed by the following steps. The first step is to verify and confirm the causal chain based on existing natural or social science theories. When the above theory has not yet appeared, we can also use cognitive tools such as Soft System Methodology (SSM) to construct a common logical-based causal chain. The second step is to organise evidence around the causal chain, and commonly, the evidence can be the physical, social, and data traces of each element on the causal chain. In the third step, the pieces of evidence (or evidence clusters) should be linked to form an evidence chain. In this step, we need to examine the relationship between different pieces of evidence (or evidence clusters) to ensure that there are no conflicts or irrelevances between pieces of evidence (or evidence clusters).
Because of the mutual reinforcement of nonlinear evidence, an evidence chain produces a higher degree of credibility than the simple sum of individual evidence. Therefore, in this study, the evaluation of the academic impact based on the evidence chain can be transformed into an estimate of the overall probability of the evidence chain by the method of Bayesian probability.
To demonstrate the process above, an empirical study is addressed in this article, which tries to compare the actual academic impact of two journals (Journal O and Journal M) in the operation research field. The first step is constructing the causal chain of the evaluation based on the existed citation motivation theory put forward by Weinstock and the elements of “Publish-Disseminate -Understand-Cite” constitutes the chain. In the second step, the Altmetric data and bibliometric data supporting each element on the causal chain was collected to form the evidence chain. To improve the accuracy of the chain, we also combed the data sources of different Altmetrics platform and selected Altmetrics indicators with high credibility. By calculating the data of the evidence chain, we concluded that journal M′s academic impact evidence chain has higher reliability than journal O, although the former journal′s five-year impact factor is higher than journal O, and this result also matches with the subject judgments generated from experts in the OR field.
It can be pointed out according to the theoretical and empirical research in this article that the evidence chain-based academic impact evaluation framework can overcome the shortcomings of the traditional bibliometric method by integrating more evaluation information sources and reasoning in a more accurate way.

Key words: Altmetrics, academic impact, impact evaluation, evidence chain