Introduction
Technology-focused management and utilization of information about the teacher-student learning experiences can enhance the level of impact of educational-process initiatives on students’ satisfaction. Studies have shown that, frequently, higher education institutions (HEI) rely heavily on the outcomes of the Students Evaluation of Teaching (SET), not only for improving the different learning activities and decision-making strategies about the organizations but also for determining academic performance and assessment of teachers (Badri, Abdulla, Kamali, & Dodeen,
2006; Bianchini, Lissoni, & Pezzoni,
2013; Boring,
2017). Indeed, using information from the learning activities to support decision-making (i.e., learning analytics) (Ferguson,
2012; Gedrimiene, Silvola, Pursiainen, Rusanen, & Muukkonen,
2019; Papamitsiou & Economides,
2014), is useful for the dissemination of student-generated data to create enhanced support for the learners (Slade & Prinsloo,
2013). Also, it is essential to capitalize on the performance indicators when addressing the different challenges and developments associated with educational technologies (Perrotta & Williamson,
2018). For example, Badri et al. (
2006) notes that SET has become a factor in promotion, long-term contracts, merits, and award-related decisions, and contract renewals for teachers in most institutions. Perrotta and Williamson (
2018) note that how the different educational choices and activities are collectively managed mainly depends on the data collection procedures, the methods applied for the sample analyses, and the interpretation and communication of the results or findings from the data.
This study assumes that there is a need for emerging innovative methods to accurately extract contextually-based educational information from the extensive collection of data recorded in SET and to transform it into actionable insights for decision-making. Interestingly, Lau, Lee, and Ho (
2005) note that quite often, it is challenging to combine the qualitative text data with quantitative methods for analysis. For instance, existing research focuses on information extracted from questionnaires or surveys completed by students. These are designed by the researchers to obtain student opinions; they pay little attention to their actual viewpoints or emotions, as written in the SET comments. In this paper, text mining (e.g., sentiment analysis) is considered as a method that supports educational process innovation and information management. Technically, text mining is capable of analyzing large volumes of data that are recorded in various scholastic databases (Lau et al.,
2005) to derive new and relevant information for driving business operations forward. Examples include ensuring the right decision-making strategies and monitoring potential deviations or bottlenecks. We note that by using the information correctly, educational organizations can define procedures and policies aimed to maintain a strong relationship with the stakeholders (Payne,
2006; Piedade & Santos,
2010). The models and methodologies discussed thus far have shown to be essential to promote the personalization of users’ experiences and outcomes, e.g., learning processes, teachers’ performance, and competencies (Pedró, Subosa, Rivas, & Valverde,
2019; Sánchez, Domínguez, Blanco, & Jaime,
2019; Yadav & Berges,
2019).
The rationale of the study
This study proposes the Educational Process and Data Mining (EPDM) model to demonstrate the capabilities of text mining and its related technologies within the educational domain. The main research questions are as follows:
-
How can we analyze the textual data in SET to derive useful information from the opinion of the students to support the different educational processes and decision-making strategies? and
-
How can we utilize the derived data to understand how the students evaluate their teachers and their qualities, considering the gender of the teachers?
Accordingly, the study develops a set of constructs and uses it to perform the investigations and analyses as follows:
-
We identify the most frequent terms used by the student to describe the teachers and the number of occurrences within the comments that reference the teachers’ gender and differences.
-
We also perform sentiment analysis to determine the intensity of the students’ comments towards the teachers by gender.
-
We determine the qualities that the undergraduates consider essential in the evaluation of the teaching-learning process and the performance of the teachers.
-
Finally, we establish the implications of the statistical significance and differences by the gender of the teachers and provide an empirical discussion of the results.
Based on the stated research questions and effort to provide answers to the identified objectives, this study makes the following contributions to knowledge:
1.
It defines an educational process and data mining model (EPDM) that leverages the perspectives or opinions of the undergraduate students; stated in the Students’ Evaluation of Teaching (SET). In turn, the method provides useful information to enhance the end-to-end processes within the educational domain.
2.
It demonstrates the capabilities of text mining technologies and their application within the educational domain.
3.
It describes a method utilized to define how students rate their teachers’ qualities and performance based on gender differences.
4.
It illustrates how data about students’ evaluations of teaching can be analyzed to provide solutions to curricular challenges in a competitive and rapidly changing educational environment. This is paramount during a time when SET data are stored at an unprecedented rate within the different educational information systems or databases. Moreover, there is an ever-increasing need to improve and support the teacher-student learning processes.
The remainder of this paper is structured as follows: The Background Information section discusses the relevant state-of-the-art in this topic area, especially as it relates to analyzing SET outcomes by gender of the teachers. The Methodology section introduces the proposed method, the EDPM model and its main components, the participants’ information, instruments, data analyses, and case study implementation. In the Discussion section, this work covers the outcomes and the impact of the EDPM method, and then concludes and points out the direction for future studies in the Conclusion section.
Discussion
This paper introduces the Educational Process and Data Mining model (EPDM) (Fig.
1) as a method towards discovering and understanding the different sets of qualities that are considered necessary for teaching-learning processes. The current research has shown that students perceive the qualities of the teachers to a significant level. Their comments include feedback on the teaching advice and interactions, intellectual challenge, guidance and supervision, and overall performance, for example. Moreover, this study also indicated that the perception of teachers’ qualities could vary by the gender of the teachers. Interestingly, early studies has shown that the post-course or semester evaluations in SET, to some extent, are significantly biased against female teachers. Also, many of the educational institutions rely exclusively on SET for decision-making purposes regarding the teachers (Boring, Ottoboni, & Stark,
2016; Whitney, Hayter, & Marshall,
2019).
The findings on how the students evaluate their teachers’ performance through the ECOA SET (see: the section on Data sampling and participants) revealed more comments for male teachers (52218) than for females (41072). Perhaps, this could be due to the higher number of male teachers across the institution. However, this work only used the first 41,072 comments for male teachers in its analysis for equality of feedback between the two genders and for conformability and validation of the results. We noted that 19 of the 20 words with the highest frequency were the same for the teachers, regardless of gender (see: Table 4 in
appendix and Figs.
2 and
3).
Also noteworthy is the fact that the two most mentioned words were “knowledge” and “explain,” the first one being on top for males while the second one led for females. Although Table 4 in
appendix does not show whether the comments were positive or negative, this reflects gender bias. The students valued female teachers on their teaching methodology or the way they were able to explain clearly the concepts/materials that they teach; whereas, they valued male teachers for the knowledge they could demonstrate.
Although the frequency of the words appearing in Table 4 in
appendix varies by gender, we note that the only difference between male and female when considering the top twenty words were “problem” and “patience.” While the term “problem” is used to refer to males, “patience” is used to refer to females, respectively. Consequently, in Table 4 in
appendix, we link the two words (problem and patience) with the correlated words to deduce how the terms relate to the different genders independently. Thus, while words such as problems (problemas), resolve (resolver), and imagination (imaginación) were correlated for the male teachers, the term “patience” did not correlate with any other in the data. Nonetheless, these observations also aligned with or supported the fact that the students viewed the male teachers as more knowledgeable and the female teachers as having a better ability to explain concepts and topics very well. Perhaps, this was also the reason why the students used the attribute “patience” to describe the female teachers.
Some examples of specific comments provided by the students related to male teachers were “Smart teacher; however, there is little interaction in classes,” (El Maestro es inteligente, sin embargo, las clases poco interactivas) and “He has general knowledge of all topics.” (Tiene conocimiento general sobre todos los temas). Comments related to female teachers included “You understand the topics very well “ (Entiendes muy bien los temas) and “She always explains differently to make it possible for all students to understand the course content. She is always patient and kind.” (Siempre explica de una forma diferente para que sea posible que todos los alumnos entendamos el contenido del curso. Siempre es paciente y amable). We confirm that the noted affirmations and statements aligned with the results of this study and settle that the outcome and results of the EPDM method described in this paper are strongly evident.
In this study, it is noteworthy that nine of the questions in the ECOA SET survey were mandatory and multiple-choice, while the comments from which the data was extracted were optional. We believe there is a need for educators to emphasize and explore more the impact of the students’ opinions on the teacher-student dynamics in learning and on the educational processes/management at large. Moreover, Slade and Prinsloo (
2013) note that for the stakeholders to achieve an effective datafied-education ethically, the students should partake in defining the context, conditions, and purposes for which the data collected from them are used.
The sentiment chart (Fig.
9) demonstrates that the majority of the available comments show adverse emotions to some extent. It could be that the students who write negative comments do so because they have bad evaluations from their teachers. The figure (Fig.
9) also shows that the assessment of the teachers correlates strongly with feelings beyond satisfaction or dissatisfaction. Perhaps, the two affirmations mentioned previously could suggest that SET is not an objective evaluation for teachers’ performance. Therefore, a more profound and cross-related analysis with descriptive statistics should be done, such as the text mining method and analysis described in this paper.
We found that even though the students tended to confide in the male teachers, there significantly existed also some elements of anxiousness, uncertainty, resentfulness, and dissatisfaction. For the female professors, the overall sentiment ratings showed a mixture of both positive and negative feelings, which perhaps confirms the results of the analysis of emotional valence, as represented in Fig.
5, where the emotional slope moves up and down intermittently. In any case, the students appear to be motivated, contented, and amazed by the teaching qualities of the professors or tutors. Moreover, it is essential to mention that most of the comments provided by the students are considered neutral with little or no sentiment (emotions) attached (see Figs.
4,
5, and
7).
In the wider spectrum of scientific research, several works have studied the impact of the state-of-the-art methodologies on the teacher-student learning process (Engen,
2019; Gallego-Arrufat, Torres-Hernández, & Pessoa,
2019; Gordillo, López-Pernas, & Barra,
2019; Silva, Usart, & Lázaro-Cantabrana,
2019). Particularly, Silva et al. (
2019) note that a larger scale of the early indicators and success factors have been as a result of using the information or digital technologies to improve teaching and professional development (Kori et al.,
2018). Whereas, Engen (
2019) argues that there is also a need for a greater understanding of new technologies and their effective use by stakeholders such as teachers and students. They (Engen,
2019) argue that the different cultural and social aspects of the users of digital technologies for transforming modern educational institutions and models should be considered. Interestingly, the work of (Silva et al.,
2019) comments that there has been much progress demonstrated in organizational management, teaching spaces, planning, and technology to support the student learning process. However, evidence from the existing works has shown that male students achieved a higher level of competencies compared with female students (Silva et al.,
2019). Perhaps, future research questions might address whether such differences in skills also exist for male and female teachers (Sánchez et al.,
2019; Exter et al.,
2019; Yadav & Berges,
2019; Crues et al.,
2018).
Nonetheless, the results of this paper show that by analyzing the real comments (opinions) of the students, we were able not only to provide a better way to understand their feelings and what they expect from the teachers but also what would help inform the decision-making processes and strategies of higher education institutions. Not only does the EPDM method show a high level of aptitude to offer solutions for the different educational frameworks and practices, but also the model can provide new and better ways to monitor and improve the educational processes.
There are two main drivers for the method introduced in this research. First, data about student evaluations of teaching are captured and stored at an unprecedented rate within the educational information systems/databases. Second, there is now more than ever an increasing need to improve and support the teacher-student learning processes because of the competitive and rapidly changing educational environment and curricula. Thus, we developed the EPDM model as a theoretical bridge for the aforementioned gaps or challenges. Moreover, the EPDM design framework can easily be applied by the owners of educational processes, innovators, and educational advisory boards to understand the organizational structures and policies they need for informed teaching-process decisions and the provision of valuable support to the stakeholders. This also includes improvement and monitoring of the various activities that underly the educational processes and learning in general.
The concept of
datafication used and explained in this study cannot be fully emphasized or practically applied without acknowledging the ethical considerations that surround the resultant methods. Therefore, we turned our attention to some of the ethical implications, especially the sociotechnical perspective on data usage in education. Whereas Slade and Prinsloo (
2013) highlight the need to harvest data under provisions that ensure trust among the different institutions and students, we took steps to ensure that the contexts in which the readily available datasets were analysed and the suggested predictions were made are within the social construction and moral objectives of technical expertise (Perrotta & Williamson,
2018; Prinsloo,
2017; Slade & Prinsloo,
2013). We bore in mind the need for suitable methods of data-driven segmentation and diversification (Perrotta & Williamson,
2018) and the ethical framework and practices that require the HEIs to offer context-appropriate solutions or strategies that increase the quality and effectiveness of teaching and learning processes (Slade & Prinsloo,
2013). Whereas Prinsloo (
2017) notes that the socio-technical imaginary of HEIs and algorithmic decision-making methods (such as the EPDM model developed in this paper) offer huge potential; we must also acknowledge the risks and ethical concerns that are attached to such data-driven methods. Besides, it does no harm to map out the opportunities and threats in applying new algorithmic or technology-focused decision-making practices within the higher educational settings, especially when adequate procedures such as sensing, processing, acting, and learning are put in place (Prinsloo,
2017).
Indeed, this study has identified and discussed the different problems with current tools and methods that support the SET evaluations, especially as it concerns the genders of the teachers. This work considers how to resolve the identified challenges, especially regarding new technologies and data-driven approaches such as text mining (natural language processing) in the education domain. To this effect, this paper proposes the EPDM model, which shows to be effective in the analysis of the textual data collected in SET for improved educational process management and technology-based decision-making. However, while we believe that the method is suitable for a contextual-based analysis of the different opinions and perspectives of the students in understanding the teacher-student processes, there might also exist a number of limitations or threats to its validity. For instance, although this paper has introduced the framework for the sets of descriptive and quantifiable text analyses that provide ample understanding of the student comments found in the datasets, there could be potentially many ways to approach this, and there may be bigger areas of considerations not yet addressed. This is due to the fact that the text analysis methods (e.g. sentiment analysis) represent a new area within the educational domain, and there are not many tools or methods that support this approach currently in the literature. Therefore, the study assumes that this work is an incentive and methodological road map for more robust and intensive research to come, particularly within the broad and overlapping field of educational-process mining and innovation.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.