Mike Thelwall

Mike Thelwall, Kayvan Kousha, Mahshid Abdoli et al.

National research evaluation initiatives and incentive schemes have previously chosen between simplistic quantitative indicators and time-consuming peer review, sometimes supported by bibliometrics. Here we assess whether artificial intelligence (AI) could provide a third alternative, estimating article quality using more multiple bibliometric and metadata inputs. We investigated this using provisional three-level REF2021 peer review scores for 84,966 articles submitted to the UK Research Excellence Framework 2021, matching a Scopus record 2014-18 and with a substantial abstract. We found that accuracy is highest in the medical and physical sciences Units of Assessment (UoAs) and economics, reaching 42% above the baseline (72% overall) in the best case. This is based on 1000 bibliometric inputs and half of the articles used for training in each UoA. Prediction accuracies above the baseline for the social science, mathematics, engineering, arts, and humanities UoAs were much lower or close to zero. The Random Forest Classifier (standard or ordinal) and Extreme Gradient Boosting Classifier algorithms performed best from the 32 tested. Accuracy was lower if UoAs were merged or replaced by Scopus broad categories. We increased accuracy with an active learning strategy and by selecting articles with higher prediction probabilities, as estimated by the algorithms, but this substantially reduced the number of scores predicted.

Mike Thelwall, Kayvan Kousha, Mahshid Abdoli et al.

This document describes strategies for using Artificial Intelligence (AI) to predict some journal article scores in future research assessment exercises. Five strategies have been assessed.

Kayvan Kousha, Mike Thelwall

This literature review identifies indicators that associate with higher impact or higher quality research from article text (e.g., titles, abstracts, lengths, cited references and readability) or metadata (e.g., the number of authors, international or domestic collaborations, journal impact factors and authors' h-index). This includes studies that used machine learning techniques to predict citation counts or quality scores for journal articles or conference papers. The literature review also includes evidence about the strength of association between bibliometric indicators and quality score rankings from previous UK Research Assessment Exercises (RAEs) and REFs in different subjects and years and similar evidence from other countries (e.g., Australia and Italy). In support of this, the document also surveys studies that used public datasets of citations, social media indictors or open review texts (e.g., Dimensions, OpenCitations, Altmetric.com and Publons) to help predict the scholarly impact of articles. The results of this part of the literature review were used to inform the experiments using machine learning to predict REF journal article quality scores, as reported in the AI experiments report for this project. The literature review also covers technology to automate editorial processes, to provide quality control for papers and reviewers' suggestions, to match reviewers with articles, and to automatically categorise journal articles into fields. Bias and transparency in technology assisted assessment are also discussed.

Mike Thelwall

Evaluating the quality of academic journal articles is a time consuming but critical task for national research evaluation exercises, appointments and promotion. It is therefore important to investigate whether Large Language Models (LLMs) can play a role in this process. This article assesses which ChatGPT inputs (full text without tables, figures and references; title and abstract; title only) produce better quality score estimates, and the extent to which scores are affected by ChatGPT models and system prompts. The results show that the optimal input is the article title and abstract, with average ChatGPT scores based on these (30 iterations on a dataset of 51 papers) correlating at 0.67 with human scores, the highest ever reported. ChatGPT 4o is slightly better than 3.5-turbo (0.66), and 4o-mini (0.66). The results suggest that article full texts might confuse LLM research quality evaluations, even though complex system instructions for the task are more effective than simple ones. Thus, whilst abstracts contain insufficient information for a thorough assessment of rigour, they may contain strong pointers about originality and significance. Finally, linear regression can be used to convert the model scores into the human scale scores, which is 31% more accurate than guessing.

Mike Thelwall

Purpose: Assess whether ChatGPT 4.0 is accurate enough to perform research evaluations on journal articles to automate this time-consuming task. Design/methodology/approach: Test the extent to which ChatGPT-4 can assess the quality of journal articles using a case study of the published scoring guidelines of the UK Research Excellence Framework (REF) 2021 to create a research evaluation ChatGPT. This was applied to 51 of my own articles and compared against my own quality judgements. Findings: ChatGPT-4 can produce plausible document summaries and quality evaluation rationales that match the REF criteria. Its overall scores have weak correlations with my self-evaluation scores of the same documents (averaging r=0.281 over 15 iterations, with 8 being statistically significantly different from 0). In contrast, the average scores from the 15 iterations produced a statistically significant positive correlation of 0.509. Thus, averaging scores from multiple ChatGPT-4 rounds seems more effective than individual scores. The positive correlation may be due to ChatGPT being able to extract the author's significance, rigour, and originality claims from inside each paper. If my weakest articles are removed, then the correlation with average scores (r=0.200) falls below statistical significance, suggesting that ChatGPT struggles to make fine-grained evaluations. Research limitations: The data is self-evaluations of a convenience sample of articles from one academic in one field. Practical implications: Overall, ChatGPT does not yet seem to be accurate enough to be trusted for any formal or informal research quality evaluation tasks. Research evaluators, including journal editors, should therefore take steps to control its use. Originality/value: This is the first published attempt at post-publication expert review accuracy testing for ChatGPT.

Mike Thelwall, Xiaorui Jiang, Peter A. Bath

Estimating the quality of published research is important for evaluations of departments, researchers, and job candidates. Citation-based indicators sometimes support these tasks, but do not work for new articles and have low or moderate accuracy. Previous research has shown that ChatGPT can estimate the quality of research articles, with its scores correlating positively with an expert scores proxy in all fields, and often more strongly than citation-based indicators, except for clinical medicine. ChatGPT scores may therefore replace citation-based indicators for some applications. This article investigates the clinical medicine anomaly with the largest dataset yet and a more detailed analysis. The results showed that ChatGPT 4o-mini scores for articles submitted to the UK's Research Excellence Framework (REF) 2021 Unit of Assessment (UoA) 1 Clinical Medicine correlated positively (r=0.134, n=9872) with departmental mean REF scores, against a theoretical maximum correlation of r=0.226. ChatGPT 4o and 3.5 turbo also gave positive correlations. At the departmental level, mean ChatGPT scores correlated more strongly with departmental mean REF scores (r=0.395, n=31). For the 100 journals with the most articles in UoA 1, their mean ChatGPT score correlated strongly with their REF score (r=0.495) but negatively with their citation rate (r=-0.148). Journal and departmental anomalies in these results point to ChatGPT being ineffective at assessing the quality of research in prestigious medical journals or research directly affecting human health, or both. Nevertheless, the results give evidence of ChatGPT's ability to assess research quality overall for Clinical Medicine, where it might replace citation-based indicators for new research.

Kayvan Kousha, Mike Thelwall

Academics and departments are sometimes judged by how their research has benefitted society. For example, the UK Research Excellence Framework (REF) assesses Impact Case Studies (ICS), which are five-page evidence-based claims of societal impacts. This study investigates whether ChatGPT can evaluate societal impact claims and therefore potentially support expert human assessors. For this, various parts of 6,220 public ICS from REF2021 were fed to ChatGPT 4o-mini along with the REF2021 evaluation guidelines, comparing the results with published departmental average ICS scores. The results suggest that the optimal strategy for high correlations with expert scores is to input the title and summary of an ICS but not the remaining text, and to modify the original REF guidelines to encourage a stricter evaluation. The scores generated by this approach correlated positively with departmental average scores in all 34 Units of Assessment (UoAs), with values between 0.18 (Economics and Econometrics) and 0.56 (Psychology, Psychiatry and Neuroscience). At the departmental level, the corresponding correlations were higher, reaching 0.71 for Sport and Exercise Sciences, Leisure and Tourism. Thus, ChatGPT-based ICS evaluations are simple and viable to support or cross-check expert judgments, although their value varies substantially between fields.

Mike Thelwall, Abdullah Yaghi

While previous studies have demonstrated that Large Language Models (LLMs) can predict peer review outcomes to some extent, this paper builds on that by introducing two new contexts and employing a more robust method - averaging multiple ChatGPT scores. The findings that averaging 30 ChatGPT predictions, based on reviewer guidelines and using only the submitted titles and abstracts, failed to predict peer review outcomes for F1000Research (Spearman's rho=0.00). However, it produced mostly weak positive correlations with the quality dimensions of SciPost Physics (rho=0.25 for validity, rho=0.25 for originality, rho=0.20 for significance, and rho = 0.08 for clarity) and a moderate positive correlation for papers from the International Conference on Learning Representations (ICLR) (rho=0.38). Including the full text of articles significantly increased the correlation for ICLR (rho=0.46) and slightly improved it for F1000Research (rho=0.09), while it had variable effects on the four quality dimension correlations for SciPost LaTeX files. The use of chain-of-thought system prompts slightly increased the correlation for F1000Research (rho=0.10), marginally reduced it for ICLR (rho=0.37), and further decreased it for SciPost Physics (rho=0.16 for validity, rho=0.18 for originality, rho=0.18 for significance, and rho=0.05 for clarity). Overall, the results suggest that in some contexts, ChatGPT can produce weak pre-publication quality assessments. However, the effectiveness of these assessments and the optimal strategies for employing them vary considerably across different platforms, journals, and conferences. Additionally, the most suitable inputs for ChatGPT appear to differ depending on the platform.

Mike Thelwall

Although both Google Gemini (1.5 Flash) and ChatGPT (4o and 4o-mini) give research quality evaluation scores that correlate positively with expert scores in nearly all fields, and more strongly that citations in most, it is not known whether this is true for smaller Large Language Models (LLMs). In response, this article assesses Google's Gemma-3-27b-it, a downloadable LLM (60Gb). The results for 104,187 articles show that Gemma-3-27b-it scores correlate positively with an expert research quality score proxy for all 34 Units of Assessment (broad fields) from the UK Research Excellence Framework 2021. The Gemma-3-27b-it correlations have 83.8% of the strength of ChatGPT 4o and 94.7% of the strength of ChatGPT 4o-mini correlations. Differently from the two larger LLMs, the Gemma-3-27b-it correlations do not increase substantially when the scores are averaged across five repetitions, its scores tend to be lower, and its reports are relatively uniform in style. Overall, the results show that research quality score estimation can be conducted by offline LLMs, so this capability is not an emergent property of the largest LLMs. Moreover, score improvement through repetition is not a universal feature of LLMs. In conclusion, although the largest LLMs still have the highest research evaluation score estimation capability, smaller ones can also be used for this task, and this can be helpful for cost saving or when secure offline processing is needed.

Mike Thelwall

Researchers may be tempted to attract attention through poetic titles for their publications, but would this be mistaken in some fields? Whilst poetic titles are known to be common in medicine, it is not clear whether the practice is widespread elsewhere. This article investigates the prevalence of poetic expressions in journal article titles 1996-2019 in 3.3 million articles from all 27 Scopus broad fields. Expressions were identified by manually checking all phrases with at least 5 words that occurred at least 25 times, finding 149 stock phrases, idioms, sayings, literary allusions, film names and song titles or lyrics. The expressions found are most common in the social sciences and the humanities. They are also relatively common in medicine, but almost absent from engineering and the natural and formal sciences. The differences may reflect the less hierarchical and more varied nature of the social sciences and humanities, where interesting titles may attract an audience. In engineering, natural science and formal science fields, authors should take extra care with poetic expressions, in case their choice is judged inappropriate. This includes interdisciplinary research overlapping these areas. Conversely, reviewers of interdisciplinary research involving the social sciences should be more tolerant of poetic license.

Ehsan Mohammadi, Mike Thelwall

Reading academic publications is a key scholarly activity. Scholars accessing and recording academic publications online are producing new types of readership data. These include publisher, repository, and academic social network download statistics as well as online reference manager records. This chapter discusses the use of download and reference manager data for research evaluation and library collection development. The focus is on the validity and application of readership data as an impact indicator for academic publications across different disciplines. Mendeley is particularly promising in this regard, although all data sources are not subjected to rigorous quality control and can be manipulated.

Mike Thelwall

Computer systems need to be able to react to stress in order to perform optimally on some tasks. This article describes TensiStrength, a system to detect the strength of stress and relaxation expressed in social media text messages. TensiStrength uses a lexical approach and a set of rules to detect direct and indirect expressions of stress or relaxation, particularly in the context of transportation. It is slightly more effective than a comparable sentiment analysis program, although their similar performances occur despite differences on almost half of the tweets gathered. The effectiveness of TensiStrength depends on the nature of the tweets classified, with tweets that are rich in stress-related terms being particularly problematic. Although generic machine learning methods can give better performance than TensiStrength overall, they exploit topic-related terms in a way that may be undesirable in practical applications and that may not work as well in more focused contexts. In conclusion, TensiStrength and generic machine learning approaches work well enough to be practical choices for intelligent applications that need to take advantage of stress information, and the decision about which to use depends on the nature of the texts analysed and the purpose of the task.

Patrick Kenekayoro, Kevan Buckley, Mike Thelwall

Using raw hyperlink counts for webometrics research has been shown to be unreliable and researchers have looked for alternatives. One alternative is classifying hyperlinks in a website based on the motivation behind the hyperlink creation. The method used for this type of classification involves manually visiting a webpage and then classifying individual links on the webpage. This is time consuming, making it infeasible for large scale studies. This paper speeds up the classification of hyperlinks in UK academic websites by using a machine learning technique, decision tree induction, to group web pages found in UK academic websites into one of eight categories and then infer the motivation for the creation of a hyperlink in a webpage based on the linking pattern of the category the webpage belongs to.