Understanding factors that affect the direction and pace of scientific progress has been central to the advancement of science and technology. Fostering innovations through public and private investments requires analysing enormous amounts of data and experimental evidence to visibly explain why certain scientific topics receive more attention resulting in increased innovation and follow-on impact. To that end of understanding the dynamics of scientific progress, there are key questions that have been receiving widespread attention in the area of science of science and innovations and still demand more research.
This themed article collection calls for studies on understanding, measuring and accelerating the scientific process. The topics of interest are related (but not limited) to data mining, network science, information retrieval, machine learning and natural language processing (NLP) with patent and scholarly data, citation data, collaboration network data, knowledge graphs, time series data among others to further the research on the processes of metascience and innovation, in addition to predicting as well as recommending emerging areas of innovation rather than simply measuring the innovation pace. Such techniques will have an enormous impact on understanding the determinants of long-term innovations and why certain fields sustain longer than others.
This Research Topic solicits perspectives and survey pieces, blue-sky proposals, empirical analyses, and methodological contributions that cover insights into a range of topics in no particular order:
● AI for understanding and advancing the scientific process - Data mining and machine learning tools can harness user data, user/researcher connections, and scientific databases to uncover patterns behind scientific success. Questions remain on the predictability patterns of success associated with the content of the scientific articles, patent data and metadata. Can scaling laws of article citations over time aggregated from these databases be used to study the hidden dynamics of scientific progress? Can these explain the factors behind the emergence of innovative ideas? What kind of topics stand the test of time, what causes them to be impactful and can machine learning help in the visualisation and interpretability of these topics? Can advances in recommendation systems be used to predict emerging areas of science and innovations personalised to research groups?
● Scientific NLP: Models of language and images have made striking progress in generating new content and additionally, combining existing knowledge to ground new concepts. Studies in this domain include developing new forms of representation learning techniques and information retrieval systems that can query complex texts and summarise them by automatically approximating a researcher’s knowledge, objectives, needs and interests, based on data. Recently, Large Language Models (LLMs) have shown promise with generative decoding. Questions remain on whether these tools can help in the discovery of new knowledge given that much of the information stored in these models as weights results in memorization and hallucination of existing information on the web.
● Science of collaborations and teams - Research collaborations, often large, have become the norm for scientific progress. Recent studies have shown that “large teams develop and small teams disrupt science and technology”. Understanding the social processes of science through computational approaches can enable faster scientific discovery and better productivity of researchers. It remains to be seen what works in these scenarios - what is the optimal size of teams for objective scientific progress, and how does the diversity of thoughts and backgrounds in teams enable better innovations thereby uncovering new avenues of science and innovation directions? Are there ways to utilise user/author network information and structure to select and replace when necessary, team members of specific expertise to achieve a desired goal?
● Mitigating bias in processes of metascience: The presence of homophily in ethnicity, gender and affiliation has been established to be a hindrance to scientific progress. Studying the effect of diversity on scientific impact and developing techniques to mitigate such inherent bias is an important step toward fostering better innovation mechanisms. Similarly, citation bias in peer and author networks as well as bias in peer reviews significantly impact the quality of scientific articles. How can we test for such bias and can machine learning and data mining help in debiasing the citation statistics?
-----------------------------------------------------------------------------------------------------------------
Keywords:
machine learning, network science, science of science, natural language processing, deep learning, citation networks, data science
Important Note:
All contributions to this Research Topic must be within the scope of the section and journal to which they are submitted, as defined in their mission statements. Frontiers reserves the right to guide an out-of-scope manuscript to a more suitable section or journal at any stage of peer review.
Understanding factors that affect the direction and pace of scientific progress has been central to the advancement of science and technology. Fostering innovations through public and private investments requires analysing enormous amounts of data and experimental evidence to visibly explain why certain scientific topics receive more attention resulting in increased innovation and follow-on impact. To that end of understanding the dynamics of scientific progress, there are key questions that have been receiving widespread attention in the area of science of science and innovations and still demand more research.
This themed article collection calls for studies on understanding, measuring and accelerating the scientific process. The topics of interest are related (but not limited) to data mining, network science, information retrieval, machine learning and natural language processing (NLP) with patent and scholarly data, citation data, collaboration network data, knowledge graphs, time series data among others to further the research on the processes of metascience and innovation, in addition to predicting as well as recommending emerging areas of innovation rather than simply measuring the innovation pace. Such techniques will have an enormous impact on understanding the determinants of long-term innovations and why certain fields sustain longer than others.
This Research Topic solicits perspectives and survey pieces, blue-sky proposals, empirical analyses, and methodological contributions that cover insights into a range of topics in no particular order:
● AI for understanding and advancing the scientific process - Data mining and machine learning tools can harness user data, user/researcher connections, and scientific databases to uncover patterns behind scientific success. Questions remain on the predictability patterns of success associated with the content of the scientific articles, patent data and metadata. Can scaling laws of article citations over time aggregated from these databases be used to study the hidden dynamics of scientific progress? Can these explain the factors behind the emergence of innovative ideas? What kind of topics stand the test of time, what causes them to be impactful and can machine learning help in the visualisation and interpretability of these topics? Can advances in recommendation systems be used to predict emerging areas of science and innovations personalised to research groups?
● Scientific NLP: Models of language and images have made striking progress in generating new content and additionally, combining existing knowledge to ground new concepts. Studies in this domain include developing new forms of representation learning techniques and information retrieval systems that can query complex texts and summarise them by automatically approximating a researcher’s knowledge, objectives, needs and interests, based on data. Recently, Large Language Models (LLMs) have shown promise with generative decoding. Questions remain on whether these tools can help in the discovery of new knowledge given that much of the information stored in these models as weights results in memorization and hallucination of existing information on the web.
● Science of collaborations and teams - Research collaborations, often large, have become the norm for scientific progress. Recent studies have shown that “large teams develop and small teams disrupt science and technology”. Understanding the social processes of science through computational approaches can enable faster scientific discovery and better productivity of researchers. It remains to be seen what works in these scenarios - what is the optimal size of teams for objective scientific progress, and how does the diversity of thoughts and backgrounds in teams enable better innovations thereby uncovering new avenues of science and innovation directions? Are there ways to utilise user/author network information and structure to select and replace when necessary, team members of specific expertise to achieve a desired goal?
● Mitigating bias in processes of metascience: The presence of homophily in ethnicity, gender and affiliation has been established to be a hindrance to scientific progress. Studying the effect of diversity on scientific impact and developing techniques to mitigate such inherent bias is an important step toward fostering better innovation mechanisms. Similarly, citation bias in peer and author networks as well as bias in peer reviews significantly impact the quality of scientific articles. How can we test for such bias and can machine learning and data mining help in debiasing the citation statistics?
-----------------------------------------------------------------------------------------------------------------
Keywords:
machine learning, network science, science of science, natural language processing, deep learning, citation networks, data science
Important Note:
All contributions to this Research Topic must be within the scope of the section and journal to which they are submitted, as defined in their mission statements. Frontiers reserves the right to guide an out-of-scope manuscript to a more suitable section or journal at any stage of peer review.