On the Perceptions of Online Learning Due to COVID-19 Pandemic. Case Study: University of Foggia, Italy

Abstract

Due to COVID-19, higher education institutions all over the world transitioned to online learning. The sudden and forced transition to this new learning methodology pushed the Universities to rapidly adequate to the needs, upgrading their digital platforms to comply with the new requirements. In the same way, teachers had to adapt their teaching to fit the new medium’s potentials and limitations. The final receivers of this striking change, the students, had to adequate to the novelty approach, though this process has not been painless. Several difficulties, challenges and opportunities arose in this transition process for students, and the full digital class delivery also stressed them emotively. This study explores University of Foggia students’ perceptions of the emergency online learning. The factors analyzed involved their perception about the University implementation of the online class delivery, their consideration about the future of online learning and their emotional impact in attending courses with this methodology. Quantitative and qualitative data were collected from 3,140 participants. The findings present how students have globally appreciated online learning, but they mostly prefer a blended learning approach. Furthermore, by using segmentation variables, differences emerged among the participants’ groups, indicating online learning can have great potentials, but more complex and integrated approaches are needed to fulfill the different learning needs.

1 Introduction

The world is facing a health crisis due to the COVID-19 pandemic. As a result of COVID-19, higher education all over the world has moved to deliver courses online during Spring 2020 [2, 7, 12, 21]. In particular, in Italy, online learning has been expanded to Fall 2020 and also Spring 2021 [3, 11]. However, students have expressed stress related to online learning and difficulties compared to the traditional learning in physical classrooms [10]. By understanding students’ challenges and preferences, universities can develop strategies to assist students in case there are further weaves of COVID-19 or any other disaster requiring an emergency transition to remote learning. The shift towards online education during the pandemic of COVID-19 has led many studies to focus on perceived learning outcomes and student satisfaction in this new learning environment [1, 4]. This paper explores Italian University students’ perceptions about online learning after COVID-19 government measures (stay-at-home and/or physical distance), and in particular students from University of Foggia.

To reduce transmission of the COVID-19, several countries established measures on infection prevention and control by limiting contact between people¹. Governments suggested or ordered physical distancing and movement restrictions².

Universities aimed to slow down the spread of COVID-19 by protecting all the individuals involved in the education compartments – students, staff, and faculty members – and to help ensure a safe and healthy learning environment [5, 13]. Many universities transitioned to remote learning where classes were held online [2, 13]. Some universities were offering asynchronous classes where instructors prepare assignments or record lectures and students can complete them at their own pace [12]. Some institutions used synchronous learning that occurs at a specific time via a specific medium. In this paper, we focus on the case of the University of Foggia (Southern-Italy), a young University (about 20 years since its foundation) with about 11 thousand students in six departments (Table 1). The University of Foggia, even before the COVID-19 pandemic, was engaged in online teaching. Specifically, 7 courses were in blended-form and the University of Foggia was leading the Eduopen project (http://www.eduopen.org/), a platform providing more than 300 massive open online courses (MOOCs), involving 27 higher education institutions and 262 tutors and teachers. When the COVID-19 pandemic forced the University of Foggia (together with all the Italian education system) to switch to a complete online teaching, its previous experience helped this transition to be not so critical. The University of Foggia was already equipped with a Moodle platform, a Virtual Classroom Tool together with an e-learning center (founded in 2015) with specialized administrative staff. As a result, the University of Foggia response to the COVID-19 outbreak was not so dramatic. After the Italian President of the Council of Ministers’ decree³, enacted on 4 March 2020, all the activities were switched to a virtual form; lessons, exams, seminars, meetings were held online, and all planned academic activities have been confirmed, including graduation sessions, orientation events and research activities. In order to understand the effect of the COVID-19 outbreak on the students’ community, the Italian National Agency for the Evaluation of Universities and Research Institutes (ANVUR – www.anvur.it) and the European Association for Quality Assurance in Higher Education (ENQA – www.enqa.eu) have promoted and established a working group to develop a survey to analyze the didactic experiences made in Italian universities during the COVID-19 health emergency, also in order to offer valuable elements in view of the strategies that the universities will have to adopt once the current pandemic phase will be over. In this paper, we analyze the questionnaire provided by the University of Foggia according to the indications received by ANVUR and ENQA.

Table 1. Students at University of Foggia in academic year 2020–2021. Unique = Unique Cycle (Master degree), also known as Long First Degree Courses at Master level.

This paper is organized as follows: Sect. 2 provides detail on the methodology adopted to survey students at University of Foggia. Section 3 presents the obtained results and discuss about them. Finally, Sect. 4 concludes with final remarks and recommendations.

2 Method

To study the perceptions of online learning during COVID-19 pandemic at University of Foggia, we built a questionnaire that circulated among students by emails, messages, and word of mouth. Also University professors during lessons asked students to fill the questionnaire sharing the link. Such questionnaire includes questions to get preliminary information of participants (e.g., University degree course they are enrolled in), questions about the perceptions of online learning due to COVID-19, and about preferences of learning modalities for future. More details are reported in Sect. 2.1. Next, we analyzed the proposed questionnare. In this work, we select the Pairwise Markov random field method (PMRF), which relies also on graphs. Therefore, in this section, we briefly describe graphs in general (Sect. 2.2), and then how they are applied within the PMRF (Sect. 2.3).

2.1 The Questionnaire

In order to study the perceptions of online learning during COVID-19 pandemic at University of Foggia, we built the questionnaire reported in Table 2. The questionnaire is composed of 10 questions about perceptions related to online learning during COVID-19 (questions with Q identifier). To this we added further four questions asking participants general information (GQ identifier), their preferences on learning modalities for the future, and suggestions (LQ identifier). The answers to questions Qs are set on a 4-point likert scale (“Absolutely No”, “More No than Yes”, “More Yes than No”, “Absolutely Yes”). LQ is a nominal question and LQ2 is an open-ended question. Some of the questions (Q1, Q3, Q4, Q5, Q6, Q7, Q9, Q10) are written with a positive formulation: it means that a high agreement received as answer implicates a positive result. We refer to these questions as positive questions. Other questions (Q2, Q8) are written with a negative formulation: it means that a high agreement received as answer implicates a negative result. We refer to this latter group of questions as negative questions.

We shared the questionnaire by emails, messages, and word of mouth starting from 1st April 2021. The last student participant submitted on 19 April 2021.

Table 2. List of the questions (Q) about online/remote learning posed to students at University of Foggia.

2.2 Briefs on Graphs

Network theory is the study of graphs, representations of relations occurring between discrete objects. Deeply rooted in graph theory – a branch of mathematics whose origins can be traced back to 1735, when the Swiss mathematician Leonhard Euler solved the Königsberg bridge problem – network theory has been applied in disciplines spanning from statistical physics to computer science, from electrical engineering to biology or climatology [6, 14, 16,17,18,19,20]. In particular, as we will see in the following section, graphs can be applied to study questionnaires.

Graphs are very abstract conceptual structures that can be used to model relations and processes taking place in extremely different systems. A network is a graph with N nodes (or vertices) and L links (or edges) that can be weighted or unweighted, directed or not. An unweighted network is completely represented by its $N \times N$ adjacency matrix A such that $A_{ij} = 1$ if node i points to node j, $A_{ij} = 0$ otherwise. Let $G = (V, E)$ be a graph, where V is the set of its vertices such that $|V| = N$ and E is the set of its edges such that $|E| = L$. Edges may denote just the connection among two nodes or being labeled with a number indicating weights assigned to them. In the latter case, the graph is called weighted.

2.3 The Pairwise Markov Random Field Method

Pairwise Markov random field (PMRF) is a well-known network model, used for estimating psychological networks. The PMRF are networks where the nodes represent variables connected by undirected networks, i.e., edges where the relation between nodes can be traversed by both ways. The edges between the nodes (variables) indicate that there is conditional dependence between the two variable, while low weight or absent edges indicate that the two variable are independent after conditioning on other variables [9]. When data are multivariate normal, such a conditional independence would correspond to a partial correlation being equal to zero [9]. According to [8] this method can be applied also on ordinal data, which are available in the dataset we gathered for this research. We can imagine a network model $G=(Y,E)$ with three nodes $y_{1}, y_{2}, y_{3}$ and undirected edges $(y_{1}, y_{2})$ and $(y_{2}, y_{3})$. We can consider $y_{1}$ conditionally independent from $y_{3}$, given $y_{2}$ that may be a mediator or a moderator in a network psychometric. But if we remove the $y_{2}$ from the network, the two variable may be correlated. The two variable $y_{1}$ and $y_{3}$, may be the answers to the questions “Are you happy to go to work?” and “Do you consider your work environment comfortable?”, the two question have a high chance to be correlated, but their correlation could be explained by the variable $y_{2}$ “Is it your dream job?”.

When the data follow a multivariate normal density, the appropriate PRMF model is called the Gaussian graphical model (GGM; [15]), in which edges can directly be interpreted as partial correlation coefficients [9].

Such network is encoded in a symmetrical and real valued $p\times p$ weight matrix $\mathbf {\Omega }$, which elements $\omega _{jk}$ represents the edge between the node j and k given by $Cor(y_j, y_k|\textbf{y}^{-(j,k)}) = \omega _{jk} = \omega _{kj}$ where $\textbf{y}^{-(j,k)}$ is the vector of variables excluding the variable $y_j$ and $y_k$ [9]. The partial correlations coefficient can be obtained from the inverse of variace-covariance matrix $\mathbf {\Sigma }$, also named the precision matrix $\textbf{K}$ [15]. In case of ordinal data are used polychoric correlations [8].

The GGM model used is defined by:

$$\begin{aligned} \mathbf {\Sigma } = \mathbf {\Delta }(I - \mathbf {\Omega })^{-1}\mathbf {\Delta } \end{aligned}$$

(1)

where $\mathbf {\Delta }$ is a diagonal matrix with $\delta _{jj}=k_{jj}^{-\frac{1}{2}}$ and $\mathbf {\Omega }$ has zeros on the diagonal.

The edge missing between nodes means the two variable are conditionally independent [9].

In this work, we used GGM as a way to visualize the correlation between different segments of the sample we used. In Fig. 7 the tree different models show the variable interaction, and are helpful to explain what factors are more connected by the typology of students.

The GGM model have been generated using the R package “psychonetrics”⁴.

Figure 10 shows the most used keywords in the conclusive open question “Further observations”, and is built with the R package “tm”⁵.

3 Results

In this section, we report the results obtained from the analysis of questionnaire. We gathered a total of 3,140 participants, i.e., 26.5% of total students ad University of Foggia. The analysis is split as follows: Sect. 3.1 presents the result obtained analyzing GQs; Sect. 3.2 shows the results obtained studying Qs through both descriptive and PMRF analysis; lastly, Sect. 3.3 illustrates the future preferences of students with respect to learning modalities and the analysis of their suggestions.

3.1 Analysis of GQs

In this section, we analyze the questions GQs (Table 2). Information about GQ2, GQ4 are reported in Table 3. To ease the reading, results of GQ1, GQ3 are reported in Fig. 3 and Fig. 4, respectively.

The distribution of participants with respect to total students per degree course is shown in Fig. 1a. We observe that, except for Humanities and Law (20% or lower), all degree courses provided a similar percentage of participants (more than 25% and less than 35%). In Fig. 1b, we show the distribution of participants with respect to total students at University of Foggia based on the year of enrollment. We notice that the majority of students of the last years were involved as participants in the questionnaire, with respect to what happens for the first three years (less than 30% of students participate to the questionnaire). Figure 2 also show the distribution of student’s enrollment year by course type, indicating a great part of respondents are enrolled at first year of a bachelor course.

Fig. 1.

Distribution of participants to questionnaire with respect to students at University of Foggia.

Fig. 2.

Participants to questionnaire split based on the type of degree course and the year of enrollment to that specific course.

The distribution of participants by degree course is depicted in Fig. 3; the participants were equally distributed across the different courses at University of Foggia (from 14% of Clinical and Experimental Medicine, to 22% of Humanities).

Fig. 3.

Distribution of participants per Department at University of Foggia.

The participants were mainly enrolled in Bachelor degree courses (66.3%); the rest are evenly split between Master degree courses and Unique Cycle degree courses, 17.8% and 15.7% respectively (Table 3). The vast majority of them are enrolled in the first and second year of University (63% of total participants). We gathered answers from a small percentage of participants from four, fifth, sixth and successive years (Fig. 4). Furthermore, the vast majority of them do not live in the city of Foggia: in particular, 37.8% of them live within 50 km from University of Foggia, while 33% live more than 50 km far from the University (Table 3).

Table 3. Distribution of students by course degree and distance from Foggia

Fig. 4.

Distribution of participants per year of enrollment at University of Foggia.

3.2 Analysis of Qs

The analysis of Qs (Table 2) is split into descriptive and PMRF analysis. In the following, we first dwell on the descriptive study and then on the one conducted with PMRF method.

Descriptive Analysis. For this analysis we have chosen boxplots. Boxplots are an effective way to display results, comparing the distributions of samples. The boxplot uses as main information for the structure of the graphic, the median and quartiles. The median is represented in the graphics of this work with the red line. The upper and lower limits of the white box represents respectively, the thid quartile and fist quartile. Thus, in the white box falls the 50% of the distribution. The whiskers encompass the values falling in a range of 1.5 of the interquartile range (third quartile minus first quartile) out of the upper and lower margin of box. Values outside this range are represented a outlier points.

Fig. 5.

Participants’ answers by their home distance from University of Foggia. Foggia = Students that live in the University’s town, <50 km = students living within a radius of 50 km from University of Foggia, >50 km means students living more tha 50 km far from University of Foggia.

In Fig. 6, we display the boxplots answers for questions Qs based on participants home distance from University of Foggia. In particular, we split between those living in Foggia city, those living within a radius of 50 km from Foggia, and the ones living more than 50 km far from Foggia (including foreign students).

Overall, the figure shows how the majority of questions have received positive answers towards remote and blended learning (according to the positive and negative formulation of the question) by all three groups of participants. The most extreme results come from the question Q4, about the utility of recorded lessons, and secondly from question Q1, about the easiness of e-learning platform. Questions about the increased class attendance (Q3), Instructor proficiency for remote teaching (Q5), and possession of adequate devices and connection (Q6) show a wide positive accordance among groups. Few students had difficulties in finding didactic material (Q2) and all the three groups showed a high overall satisfaction for remote teaching (Q7). While all three groups experienced a improvement in time management (Q8), as we could expect the answers are more positive for students living outside Foggia city. The reasonable explanation is that these students have spend more time traveling to Foggia with traditional learning. This gap between students living in Foggia city and other cities is substantially suppressed by remote learning. The question Q9, about better conversation with teacher, shows a higher variance among groups. In particular, students living in Foggia city perceived a worsening of interaction “quality” with teachers during COVID-19 pandemic, while students outside experienced an improvement. Speculation about this difference may settle in the facts students from Foggia city always had better opportunity to attend class dialog with teachers directly, while students from outside experienced more limitation due to public transportation time constrictions and generally more difficulties in attending classes. The question Q10, about the negative impact on the emotional state, also highlight that students living in Foggia city were more negatively impacted from remote learning. The reasons may lie in the advantages and easiness of reaching the University facilities and higher opportunities in establishing relationship with other students.

Fig. 6.

Participants’ answers by typology of degree course. Unique = Unique Cycle Master course.

In Fig. 6, we display the answers boxplots for questions Qs based on which typology of degree course the participants are enrolled in. In particular, we split between Bachelor, Master, and Unique Cycle (Master degree). In red, we depict the median. Overall, the figure shows how the majority of questions have received positive answers towards remote and blended learning (according to the positive and negative formulation of the question) by all three groups of participants.

The most extreme results come from the question Q4, about the utility of recorded lessons, and secondly from question Q1, about the easiness of e-learning platform. Questions about the increased class attendance (Q3), Instructor proficiency for remote teaching (Q5), and possession of adequate devices and connection (Q6) show a wide positive accordance among groups. Few students had difficulties in finding didactic material (Q2) and all the three groups showed a high overall satisfaction for remote teaching (Q7). While all three groups experienced an improvement in time management (Q8), Master’s students showed a more significant improvement. Possible explanation settles in the long experience of such students in managing studying activities and because they often have a job or internship. Question Q9, about better conversation with teacher, shows a high variance among groups. In particular, Bachelor students perceived a worse kind of interaction with teachers during the COVID-19 pandemic. The main reasons for this difference maybe they are less experienced and new in the University environment, and the bachelors’ courses are usually more crowded, affecting more the possibility of a fruitful interaction on an e-learning platform⁶. Lastly, Q10 indicates that Bachelor students had the most negative impact on the emotional state. This is not unexpected, considering that most of the participants in our questionnaire were enrolled on the first year (Fig. 4). These students experienced the shifting from High schools to University during the COVID-19 emergency and subsequent restrictions. Therefore, they had few opportunities to settle in the environment and establish a relationship with other students.

PMRF Analysis. Figure 7 and Fig. 8 shows the graphs created according to the methodology described in Sect. 2. The edges in the graphs indicate the presence of a connection between answers of the questions, over the threshold of 0.15, empirically chosen as equilibrium point between informativity and legibility. The thicker the edges, the stronger is the connection between the variables. The blue edge color indicates a direct connection between the questions, i.e., a higher(lower) score of question A matches a higher(lower) score of question B. Instead, the red edge color indicates an inverse connection between the nodes, i.e., a higher(lower) score of question A matches a lower(higher) score of question B. A particularly prominent example of inverse connection in our work is the relation between node Q8 and Q10. Q8 asks the student to provide a higher score when remote learning improved his/her time management – a positive feature for remote learning. On the other hand, Q10 asks to provide a higher score if the remote learning negatively affected her/his emotional state – a negative feature for remote learning. Although the two questions ask about two different features of remote learning, some sort of connection between them is reasonable, as the emotional state can impact time management and vice versa. The way the question has been asked imply that if students are in a negative(positive) attitude toward remote learning they (should) give a lower(higher) score to Q8 and higher(lower) score to Q10. In doing this, participants create an inverse connection reflected trough the red edges in the graph. One or more latent factors likely impacted the student’s answers to the questionnaire; however, the questionnaire has not been structured to be managed with proper methodologies relative to the detection of the latent factors.

In Fig. 7 the three graphs have been generated filtering questionnaires based on where the participants declared to live. International students have been merged with the group of students living more than 50 km far from the University of Foggia. This segmentation is because students from Foggia have reduced problems in getting to the University facilities. Instead, students living in a town within a radius of 50 km are likely to be commuters and have more difficulties reaching the university facilities, as they need to spend more time traveling. The last group are the students coming from town distant more than 50 km. These are likely to be off-site students, with typical challenges of young students living in other cities far from relatives and hometown-friends.

The node numbers in the graphs match the questions of Table 2 and the features reported in Table 4. In all three graphs of Fig. 7, a strong connection between some features is evident. For example, the edge between Q5 and Q7 shows a connection between the score of Instructor proficiency for remote teaching and the overall satisfaction for remote learning. The edge between Q3 and Q8 indicates a strong connection between the increased class attendance and the time management improvement as expected. Nodes Q8 and Q9 indicate a connection between the improvement in time management and the conversation with the teacher. The edge between Q8 and Q10 indicates an inverse connection between time management improvement and the negative impact on the emotional state. Likewise, the edge between Q9 and Q10 indicates an inverse connection between improvements in conversion with the teacher and the negative impact on the emotional state. Figure 7b and 7c shows commuters and offsite students have a higher inverse connection between Q7 and Q10, i.e. the overall satisfaction for remote and blended learning and the negative impact on the emotional state, and also a negative correlation between Q2 and Q5, i.e., difficulty of finding didactic material and Instructor proficiency for remote teaching. For students living more than 50 km far from University of Foggia (Fig. 7c) there is also a stronger connection between Q8 – time management improvement – and Q3 – increased class attendance. Moreover, Q3 is also inversely connected with Q10 – the negative impact on the emotional state. These connections may be because many students very distant from university do commuters or often do not attend class. Students living within a radius of 50 km from the University of Foggia in Fig. 7b, shows a higher connection between Q1 – easiness of e-learning platform – and Q5 – Instructors proficiency for remote learning –, and also a connection between Q3 – increased class attendance – and Q7 – the overall satisfaction for remote learning. Students living in Foggia in Fig. 7a, showed a higher connection between Q4 and Q7, utility of recorded lessons and the overall satisfaction for remote learning, and between Q7 and Q9 – better conversation with the teacher. Both students living in Foggia and a town within a radius of 50 km from Foggia showed an inverse connection between the easiness of the e-learning platform and the difficulty of finding didactic material.

Table 4. List of feature measured with the questions (Q) about online/remote learning posed to students at University of Foggia. The exact corresponding questions are indicated in Table 2

Fig. 7.

Relation between the questions based on participants home distance from University of Foggia. Nodes represent questions (Q) – see Tables 2 and 4. Blue edges indicate a direct relation between the questions, red edges indicate an inverse relation between the questions. The thicker the edges the stronger relation between the questions. (Color figure online)

Figure 8 shows the three graphs generated filtering questionnaires based on the type of course attended by students. We can expect the students attending a bachelor are less experienced than Master degree students. They are younger, and usually, they just come directly from high schools. On the other hand, master degree students are more experienced in the university environment and probably more used to the University of Foggia itself. Finally, there is a more complicated category, i.e., Unique cycle students, that encompasses new students and older ones. However, these students are mainly enrolled in the program of Law and Medical and Surgical Sciences.

The consideration above pushed to explorer if these differences could affect the connection of the question. Many of the general considerations about the connection between questions made above for Fig. 7 still apply in this case. In particular, it mainly holds true for the nodes Q5, Q7, Q8, Q9 and Q10. However, it is relevant to shed light on some relations for Master students in Fig. 8b. There is a stronger connection between Q6 and Q7, i.e., possession of adequate devices and connection and the overall satisfaction for remote learning, deserving deeper future investigation, and inverse connection between Q2 and Q5, i.e., difficulty of finding didactic material and Instructor proficiency for remote teaching. Particularly interesting is the network of students enrolled in a Unique cycle Master program depicted in Fig. 8c, which shows more edges than the others. For example, the edge between Q1 and Q5, highlight a connection between the easiness of the e-learning platform and the Instructor proficiency for remote teaching. The edges between Q3 and Q10 indicate an inverse connection between increased class attendance and negative impact on the emotional state, while the edge between Q5 and Q8 also shows an inverse connection between Instructor proficiency for remote teaching and time management improvements. Lastly, there is a connection between Q4 – utility of recorded lessons – and Q7 – overall satisfaction for remote blended learning.

It is worth highlighting that the edges shown in this section’s graphs have to be interpreted only in terms of connection and not as indicators of a more positive (or negative) factor for one students’ group with respect to the others.

Fig. 8.

Relation between the questions based on participants’ type of course. Nodes represent questions (Q) – see Tables 2 and 4. Blue edges indicate a direct relation between the questions, red edges indicate an inverse relation between the questions. The thicker the edges the stronger relation between the questions. (Color figure online)

3.3 Analysis of LQs

In this section, we analyze the two LQs (Table 2).

The LQ1 question asked students what type of teaching methodology would they choose in future.

Fig. 9.

Participants’ preferences for the future learning approach. * More than $\frac{1}{10}$ but less than $\frac{2}{3}$ of lessons and other teaching activities are held in traditional learning modality; the rest are held online. ** More than $\frac{2}{3}$ of lessons and other teaching activities are held online.

Figure 9a shows the distribution of participants answers by where they live. For students living in Foggia city, the most relevant answer are Traditional learning only and Bledended learning, with more than $\frac{1}{10}$ but less than $\frac{2}{3}$ of lessons. In contrast, students living within a radius of 50 km from Foggia prefer Bledended learning, with more than $\frac{1}{10}$ but less than $\frac{2}{3}$ of lessons, and students living more than 50 km far from Foggia prefer Online learning only. The different propensity of the three groups can be easily justified by the difficulties these students face in reaching the University facilities: clearly, this impacts more students living outside Foggia.

Figure 9b shows the distribution of participants answers by the typology of degree course they are enrolled in. The preferred answer for all the three groups is Bledended learning, with more than $\frac{1}{10}$ but less than $\frac{2}{3}$ of lessons. For Master students, there is also a strong preference for Online learning only, while Unique cycle Master students highly prefer Traditional learning only. Overall, by analysing the results obtained, blended learning approaches are the more prominent among all groups. The predilection of master students for online learning may be justified by the need to conjugate study with a job.

Fig. 10.

Wordcloud of the top keywords (uni-grams) used in answers to the open-ended questions concerning “Further observation” in the questionnaire.

LQ2 is an open-ended question, inquiring students to provide further observations. Five hundred students answered this question with comments whose length varied from just one word to long comments. Figure 10 shows the word cloud of the top-100 words used by students in their comments on this question. The comments cover different areas and topics. Many of them contain an appreciation for remote learning (in particular for recorded lessons). Such comments come mainly from commuters and students with a job. They stated they benefited from remote learning and wish to continue to benefit in future.

Other comments focus on appreciation for the e-learning platform and how the University of Foggia managed it. On the other hand, others complain about issues experienced with the platform. Moreover, some comments provided suggestions to improve the management and experience of remote learning. Lastly, several students manifested discomfort for remote learning approach, complaining about the lack of interaction and relationship with students, apathy and alienation of attending class from a device, anxiety and discomfort in taking exams online, and unsuitability for matters requiring practical assistance experience.

In summary, the comments suggest that remote learning may be a viable and appreciated way to deliver classes but does not fit all the students’ needs.

4 Conclusion

In this work, we performed an exploratory analysis of the impact of remote and blended learning methodologies due to COVID-19 on the students’ community at the University of Foggia. For this reason, we surveyed students perceptions with a questionnaire whose items aimed to provide to the University a feedback about the quality of the implementation of the e-learning platform and teaching, how these modalities have been accepted by students and the students’ propensity for the next future to the different learning approaches.

The analysis showed how the overall opinion of students about the experience offered is positive. However, despite the global feedback, the analyses showed how different students’ segments had slightly or substantially different experiences. In particular, students living outside Foggia city have benefited a lot from remote learning, while students living in Foggia experienced more difficulties and reduced human contact with the online learning approach. These aspects have been also confirmed by the different preferences these segments expressed about the future learning modality at University.

Moreover, the students enrolled in Bachelor programs had more difficulties integrating into the university environment due to their direct transition from high school to University with e-learning and COVID-19 restrictions.

The Pairwise Markov random field analyses provided insights on some peculiar interactions between the measured factors specific for some segments of students. However, this analysis has been limited by the shortcomings of the questionnaire adopted, with too limited values of Likert items, and lacking some crucial questions to better segment the students and more fine-grained questions targeting students’ psychological experience. On the other hand, the insights gained through this survey and the speculation inherent some exciting results provide the foundation to elaborate more sophisticated research, aiming to answer some of possible emerged research questions.

In summary, the data lead us to consider that the implementation of remote learning at the University of Foggia was successful, despite some difficulties. However, the appreciation and the emotional experience diverges across different segments of students. This means that remote learning cannot be the ultimate solution to fit the complexity of learning needs, but more tailored programs and learning strategies have to be implemented to cope with the needs of different students at the University of Foggia and in Italy in general.