Immerse yourself in the future of research comprehension with Josh Nicholson, the co-founder and CEO of scite.ai. Discover how smart citations, a product of interdisciplinary research, revolutionize our interaction with scientific literature, elevating understanding and collaboration. Delve into his motivation to enrich our knowledge landscape and render research more accessible, addressing issues such as reproducibility in research, and harnessing the capabilities of large language models like ChatGPT. Embark on a journey through the transformative potential of Smart Citations in seamlessly enhancing research discovery, comprehension, and application across a multitude of domains.
Which wall does your research break?
Scientific research informs nearly all aspects of our lives, from topics like the impact of video games on spatial reasoning and the influence of Peppa Pig on children learning English as a second language to the role of SPG20 on cytokinesis in cancer cells. Science touches everything, and the COVID-19 pandemic made us all acutely aware of this simple fact. And yet, the growing rate of new publications, the increasingly interdisciplinary nature of research, issues of reproducibility, access, and jargon all make the process of understanding, trusting, and keeping up with scientific research challenging for both experts and individuals. Our work at scite aims to eliminate some of the inefficiencies created by the antiquated ways we engage with research so that anyone can confidently discover, understand, and evaluate research.
We have spent years building relationships with publishers to create the next generation of citations, which we call Smart Citations. Where traditional citations only show you how many articles are cited in a publication and a list containing their metadata, Smart Citations go beyond by also showing the citation contexts from each cited article, the section where it occurred, and a classification indicating whether it supports, mentionings, or contrasts the cited claims. Although scite originated from our frustration in assessing the reproducibility of studies, our Smart Citations have powered a number of interesting applications that drastically improve how we find ideas worth pursuing, manage the firehose of information, and, more recently, augment and validate answers by ChatGPT and other Large Language Models (LLMs).
What inspired or motivated you to work on your current research or project?
My motivation for working on citations and building the next generation of citations at scite is the same as my motivation for getting into research and doing a PhD in cancer biology, a desire to enhance our understanding of the world to improve people’s lives. Research is amazing, but it is inefficiently used, and that holds us back. If we can make research easier to discover, understand, and work with, we can positively impact all areas of life.
In what ways does society benefit from your research?
Though research is broad in scope and impact, there are many points of friction at each stage of the process, from finding information, understanding it, and actually translating those insights into real, meaningful impacts on the lives of others.
Our database of Smart Citations is able to power a number of features that streamline or eliminate some of these critical touchpoints, which we believe will, in turn, create better outcomes for society at large.
The early version of scite focused on making it easier to screen publications for quality through better quantitative and qualitative overviews. Instead of only seeing that a paper was cited 100 times, we also showed you that we identified maybe 32 supporting, 80 mentioning, and 8 contrasting statements from those 100 papers. By reading those citation statements, it became possible to build a quick intuition for how a paper was received, which is much easier than opening each of those 100 papers and digging and tracking that information manually.
Related to that, the old workflow of using traditional citation counts to screen publications had the unintended consequence of losing out on interesting claims from papers that were new or not published in prominent journals or by famous authors. Scite makes it easier for articles from less-known authors to have more of a voice in the scholarly conversation.
Beyond the individual paper or researcher, we’ve found interesting applications of our data when used by funding agencies and venture capital firms that want to understand their impact or more effectively allocate their capital.
Now with ChatGPT and other language models, everyone is simultaneously excited but bewildered by its potential. Scite is in a unique position to augment the information these models use through Smart Citations while adding guard rails to validate their claims.
We didn’t anticipate building some of these features when we launched scite five years ago, and we’re not sure what exact features we’ll have five years from now. But we’re confident that our Smart Citations will power them.