Literature reviews and critical analyses take time. In order to write high quality research you need to engage with the literature, read the references being cited, read the citations to the paper and make sure you understand the field as a whole. Collecting and reading all relevant references and citations takes a long time. Scite AI helps you spending less time on research and writing while simultaneously boosting the quality of your work by presenting what is said about a paper right away.
Scite AI helps with some of the major aspects of performing literature reviews and critical analysis including: critically engaging with publications, understanding how a publication and its results have been cited, and finding relevant literature on the topic in question. Finally, for many researchers, the purpose of understanding previous work is not only to position your current research but also to identify and clearly explain the gaps in research that your work will address. Scite AI helps you find research gaps.
By using Scite AI to understand literature, it will be easier for you to contribute to a more in-depth and broad discussion about differences in results and arguments and identify complex ideas in the literature to surface them in your writing.
This guide will help you learn:
- How to use Scite AI to find relevant journals, authors, literature, and specific results or claims
- How to use Scite AI to critically engage with a publication and understanding how a publication has been talked about by other authors and adding nuance to the papers you are citing
- Find the gaps in the literature and identify how research can be improved
If you are writing a critical analysis and you just want to understand at a glance how experts are criticising or supporting a publications arguments skip to "Critically Engaging with a publication and its results"
If you are writing a literature review and you just need to dig into gaps in the literature, debates about results, and understanding the current state of the art skip to "Identifying Gaps in the Literature"
Finding Relevant Works
One of the hardest aspects of literature review is finding articles relevant to your topic of interest. Even if you do find relevant articles, how do you know if they are high quality studies? This section will cover how you can use Scite AI to discover high quality and relevant research.
Scite search: sorting, citation filters and aggregations
The first starting point for many Scite AI users is our Advanced Search which allows you to search and filter the metadata of publications such as authors, titles, abstract, journals, etc. One of the unique features of Scite AI is that it classifies citations by how they are talking about a paper, whether they are supporting, contrasting, or merely mentioning. These citations are collected together in a badge and we allow people to sort or filter their search results by the numbers indicated on the badge.
A basic literature search for a topic in Scite AI would then be to search for a topic and sort the results by supporting or contrasting citations.
For example one might want to find the most supported studies on "Social Anxiety Disorder" or the most contrasted studies on "Social Anxiety Disorder".
You may notice that many studies have a high number of both contrasting and supporting citations! This is common since a publication is made up of a number of results, and a highly impactful paper usually has a number of studies both supporting some elements of it and finding contrasting arguments with some of it's claims. These studies would be great to collect for later when you are critically engaging with literature as the various citations will help you understand interesting factors about a study. If you are looking for papers who have only received supporting or contrasting citations you can use the citation filters to control how many of one citation type will appear in a search result.
Finally, another great use of Scite AI for discovering literature is to use the aggregation filters to understand who the primary authors are in the field. By discovering the primary authors in a field, you will be able to understand who to pay attention to in your literature review and whose publications to look at for further investigation.
A more comprehensive guide on using Scite search is available here.
Collecting works in a dashboard and staying up to date on new citations and publications
Once you have found a few articles of interest, there are a few ways to keep track of them within Scite. The most powerful is to add publications to a dashboard. By adding publications to a dashboard you are able to save them for later use, get alerts when a group of publications receive new citations, and evaluate Scite data on publications in one place. If you have a highly relevant search you can also create a dashboard from a search result and stay up to date when that publication has new searches or when publications in the search get new citations.
For more information on how to create dashboards and saved searches, click here
Citation Chaining
Once you have found some publications, it is time to look at their Scite reports. A Scite report contains a set of citation statements that comprise of what is being said about a publication and references, which are what a publication is saying about other papers. Citation Chaining is a powerful method of discovering relevant literature by looking at who has cited a work (forward chaining) or who has been cited by a work (backward chaining). It is called chaining because this process can be repeated over and over until no more citations or references can be explored. Additionally, we can look at what papers are being cited together (co-cites) by looking at side chaining.
In this section we will present how you can use these concepts to collect chains of scientific argument and debate to include in your literature review or critical analysis.
Forward chaining
Looking at how a publication has been cited is one of the primary values of Scite, and it helps us see what other relevant papers there are in a field by looking at who is talking about the publication I am looking at. However, we cannot take what is being said by a citation statement at face value. How can we understand the quality of the citing publication and how can we trust it's support or criticism? We can do this by looking at the badge of a citing paper. If the badge of a cited paper indicates that the paper has not received support or has been contrasted (as in example 1), we should be critical of the statement's being made. If the badge of a cited paper indicates high support, we should still be critical of those statements and read though them but this could be a signal that the work is high quality. Additionally, we should click the badge (forward chain as in Example 2) and look at what that criticism or contrasting result was.
We can repeatedly look at citations of citations to collect chains of scientific argument and debate and start to understand the current state of the art in the research field of interest. We can also collect these papers to read further on literature that is more recent.
Backward chaining
Backward chaining is similar to forward chaining but instead of citations, we are looking at the quality of references. This is important because a study might cite a key reference to justify a method, result, or argumentation and we want to understand if that reference is justified. In the worst case, a publication might be referencing something that has been retracted, withdrawn or has since been highly disputed indicating that there might be some problems with the foundation the research is built on. You can surface these problems in your literature review or talk about how a publication may have failed to address some of these issues in a critical analysis.
Side chaining (finding co-citations)
Finally, one of the most powerful but underappreciated methods of discovering literature is looking at co-citations. These are references or citations that have been referenced or cited together by other publications and may indicate that the publication has similar results, similar support, or similar criticism can be applied. Scite citation statements are the most powerful way to discover and evaluate co-citations because they let you see the co-citations in context. In example 4, we see that while the cited work in question ( Reference #33) has been supported, so has Reference #32 and Reference #68. We can see in another citation statement in example 4 even more co-citations which could help us supplement our literature review. All of the references in Scite are clickable and will take you to the supported literature in question.
Critically Engaging with a publication and its results
One of the most important aspects of literature reviews and critical analyses is that you understand how to critically engage with a publication. What this means is that not only should we understand the arguments being put forward by the paper, but also understand the context in which it's written and how it's been received by other papers. In this section we will look at how to use citation statements to help you understand a publication and its place in the literature.
Using results, arguments, methodologies presented citation statements to perform a critical analysis
In some ways citation statements are summaries of a cited paper. By reading citation statements, we are not only understanding differences or similarity of results and arguments, we are also looking at what those results and arguments are. In example 5, we see the following citation statement "Moreover, in accord with our current results, Phan et al [2006] showed that the magnitude of right amygdala activation to harsh faces in the SAD group was significantly positively correlated with severity of SAD as measured by the LSAS'' this citation statement tells us that the cited publication (Phan et al. 2006) is looking at measuring amygdala activation to harsh faces in the context of Social Anxiety Disorder (SAD) as measured by the Liebowitz Social Anxiety Scale (LSAS). That information tells us one of the primary aspects of the study (amygdala activation to harsh faces), one of it's methods (LSAS evaluation), and it's primary variables under investigation (Social Anxiety Disorder). What is interesting about scite, is that we can read publication's results, arguments, and methodologies in conjunction with other experts and thereby form a much more sophisticated understanding of a publication if we were just reading it out of context.
By reading citation statements with a publication, we are better able to critically engage with how other experts have understood the publication. This is even more clear in the context of contrasting citations where a critical difference in results, disagreement, or criticism might be surfaced that you may not have thought of or not have performed the experiments to discover.
Citation statement search
Since a publication may be cited thousands of times, it might be unrealistic to read all of the citation statements. In fact, there may even be more supporting and contrasting citations than you have time to investigate. Not only is it potentially unrealistic, but also some of the issues may be irrelevant. An example might be a publication which is important as a foundation for a methodology so most papers cite it as a symbol to point to where to read more about the method being used.
In order to help you find more relevant citation statements, scite presents the ability to search citation statements for keywords that might help you narrow down or find specific citation statements that are relevant to your literature review or the position you are taking in a critical analysis. Example 6 shows a search for LSAS to illustrate how one might search for citation statements that have some discussion around that chosen method of evaluation.
Adding nuance to your citations
Many people cite works uncritically, the find literature that supports their arguments and cite it without reading the work or how the work has been received. With Scite AI, instead of just finding and citing a paper such as just "Phan et al. (2006), measures amygdala activation to harsh faces in the context of Social Anxiety Disorder (SAD)", you can cite it as experts are citing the paper. For example: "Some evidence suggests amygdala activation in response to harsh faces (Phan et al. 2006). However, Labuschagne et al. (2010) calls this into question by reporting that amygdala hyperactivity was not observed in response to angry faces."
By reading citation statements presented on Scite AI, you can quickly add nuance to the citations you make and ensure that the nuance is supported and agreed upon by other experts.
Identifying gaps in the literature
One of the most challenging aspects of engaging with literature for non-experts and new or new to the field researchers is understanding where the gaps in the literature are. However, it is very important to surface this in a literature review and critical analysis. How can one use Scite AI to surface these literature gaps? In this section we will address three ways you can use Scite AI to surface gaps in the literature. This section assumes you are comfortable with the sections presented above.
Understanding the current state of the art
Earlier we showed how you can use forward chaining to discover the scientific discussion as a whole starting from specific citation statements. We suggested that for a literature review you collect these arguments and use them to frame your writing about a publication. We can also use forward chaining to understand the current state of the art on a topic. What you will do is click (or chain) through citations until you have come upon the most recent citing publications. By looking at these, you are looking at downstream research that has been founded on the work in question you started with. You can now look at how these recent publications have been talked about or just read the publication to understand the current state of the art on a topic. In example 7, we show a set of citations that have lead from Phan et. al. 2006 measuring dimensions of anxiety with respect to amygdala response all the way to the downstream Leppanen et al. 2018 which discusses the intervention of intranasal oxytocin on threat assessment. If you were developing a literature review that included interventions for Social Anxiety Disorder this method of forward chaining to discover the state of a current meta-review on interventions will help you ensure your literature review is addressing current topics of research. By looking at these current studies, you will also easily be able to find gaps in the literature by looking at the conclusions of these papers.
Finding differences in results
Another key area you will need to address in your literature review and critical analysis is if there are any significant debates or disputes and whether or not these have been resolved. By finding a key debate that has been resolved, you will be finding something significant to mention in your literature review or critical analysis. By finding a key debate that has not been resolved you will have found a significant gap that your current research can try to address. In order to do this in scite, you should focus on reading the contrasting citation statements. By reading those you are looking at differences in results that have been found. By forward chaining through contrasting citation statements or looking at recent studies that have both a high support and contrasting citation count you will easily be able to surface on-going debates. Example 8, shows an example looking at a contrast of a paper using corpus linguistics to examine discourses of refugees and asylum seekers in the UK press. The argument is that the contrasting result (whether there is a negative representation of immigrants in the press) on the two studies mentioned may stem from a difference in the text corpora being used, e.g. whether tabloids were included or not. What we can do as researchers is present this difference in a literature to highlight the importance of using similar datasets for language studies and, if relevant to our current research, use it as a "gap" (the fact that a type of corpora was not included) to inform how we might design a new study and see if there really is a difference between Italian and UK representations of immigrants in discourse as is currently suggested.
Finding suggested areas of investigation
Finally, an author might propose or suggest how contrasting results might be addressed or supporting results be even more firmly established. By using scite, you can search citation statements on a publication of interest to see how gaps identified by citing authors might be addressed. Some of these might be unsubstantiated suggestions, implications that an author is suggesting, or proposals of future research programmes. Example 9 presents a study looking at the computational linguistics task of word sense disambiguation, how might a computational model know if the word pen is being used to indicate a writing device or an enclosure for animals. The citation statement indicates that we should investigate models that might be able to be trained on a training set that doesn't explicitly capture word senses and perform well on the word sense disambiguation task as is indicated by the cited paper. We might take this as an indicator that this is an area that we should investigate further with our own research or suggest so in a literature review.