Predicting the results of evaluation procedures of academics

Francesco Poggi; Paolo Ciancarini; Aldo Gangemi; Andrea Giovanni Nuzzolese; Silvio Peroni; Valentina Presutti

doi:10.7717/peerj-cs.199

Predicting the results of evaluation procedures of academics

PeerJ Comput Sci. 2019 Jun 21:5:e199. doi: 10.7717/peerj-cs.199. eCollection 2019.

Authors

Francesco Poggi¹, Paolo Ciancarini^{1

2}, Aldo Gangemi³, Andrea Giovanni Nuzzolese⁴, Silvio Peroni³, Valentina Presutti⁴

Affiliations

¹ Department of Computer Science and Engineering (DISI), University of Bologna, Bologna, Italy.
² Institute of Data Science and Artificial Intelligence, Innopolis University, Innopolis, Russia.
³ Department of Classical Philology and Italian Studies, University of Bologna, Bologna, Italy.
⁴ STLab, Institute of Cognitive Science and Technologies, National Research Council, Roma, Italy.

Abstract

Background: The 2010 reform of the Italian university system introduced the National Scientific Habilitation (ASN) as a requirement for applying to permanent professor positions. Since the CVs of the 59,149 candidates and the results of their assessments have been made publicly available, the ASN constitutes an opportunity to perform analyses about a nation-wide evaluation process.

Objective: The main goals of this paper are: (i) predicting the ASN results using the information contained in the candidates' CVs; (ii) identifying a small set of quantitative indicators that can be used to perform accurate predictions.

Approach: Semantic technologies are used to extract, systematize and enrich the information contained in the applicants' CVs, and machine learning methods are used to predict the ASN results and to identify a subset of relevant predictors.

Results: For predicting the success in the role of associate professor, our best models using all and the top 15 predictors make accurate predictions (F-measure values higher than 0.6) in 88% and 88.6% of the cases, respectively. Similar results have been achieved for the role of full professor.

Evaluation: The proposed approach outperforms the other models developed to predict the results of researchers' evaluation procedures.

Conclusions: Such results allow the development of an automated system for supporting both candidates and committees in the future ASN sessions and other scholars' evaluation procedures.

Keywords: ASN; Academic assessment; Data Processing; Informetrics; Machine Learning; National Scientific Habilitation; Predictive Models; Research Evaluation; Science of Science; Scientometrics.

Grants and funding

This research has been supported by the Italian National Agency for the Assessment of Universities and Research (ANVUR) within the Uniform Representation of Curricular Attributes (URCA) project (see articolo 4 of the ‘Concorso Pubblico di Idee di Ricerca’ - bando ANVUR, 12 February 2015). Paolo Ciancarini was also supported by CINI (ENAV project) and by CNR-ISTC. There was no additional external funding received for this study. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.