Critical factors affecting student satisfaction in a distance learning environment

Student satisfaction refers to the students’ perceptions about their learning experiences and the perceived value of their distance education course (Kuo et al., 2013). It is a very important factor in the learning process. Student satisfaction is one of the four factors to be considered when designing a learning environment (Liaw & Huang, 2007).

Sun et al. (2008) showed that learners’ computer stress, their attitude towards e-learning, the flexibility of e-learning courses, the quality of e-learning courses, the perceived usefulness, the perceived ease of use and the diversity in the evaluation are critical factors affecting the perceived satisfaction of learners. Gao et al. (2020) found that the emotional engagement and the perceived playfulness of the platforms used in a blended learning environment affect satisfaction. Moreover, with emotional engagement acting as a mediator, perceived usefulness, ease-of-use and interaction also affect satisfaction.

Lim (2001) found that computer self-efficacy was a significant predictor of learners’ satisfaction in online courses as well as of their intention to attend future webbased courses. Lin et al. (2008) found that self-efficacy, social ability and task value significantly affect satisfaction from online learning courses. Student satisfaction along with self-efficacy can act as a substantial mechanism that stimulates students’ academic performance and the overall satisfaction with their studies (Kostagiolas et al., 2019).

According to Joo et al. (2013), a very important predictor of learners’ satisfaction, learners’ achievements and their persistence is the perceived value of students’ work. Therefore, the design of courses and the guidance of tutors should be directed to make learning experiences more useful for students. Task value, satisfaction and achievement were also found to have significant implications in the persistence of students. This means that students with high achievement levels tend to continue their studies, and therefore reduce the drop-out rate.

Liaw (2008) found that the perceived self-efficacy, multimedia teaching and the quality of the e-learning system are predictors of perceived satisfaction of learners. A study in the Hellenic Open University (HOU) has shown that there is a statistically significant relationship between student satisfaction and selfesteem (Vakoufari et al., 2014). In particular, it was found that as the levels of self-esteem increase, the greater the satisfaction becomes. Furthermore, significant correlations have been found between student satisfaction and tutors’ performance, interaction between students and the evaluation of the course (Anagnostopoulou et al., 2015).

Li (2019) found that the massive open online courses (MOOCs) learners’ satisfaction can be predicted by the academic level, by the number of online courses previously taken, which is an indicator of previous experience in distance learning, by the usage of selfregulated strategies and perceived learning.

Another trend in research regarding student satisfaction is the effort to link learning design metrics with student satisfaction; it was found that learner satisfaction was significantly affected be learning design (Rienties & Toetenel, 2016) and that assimilative learning design activities positively correlated with learner satisfaction. However, activities related to communication and finding information were negatively correlated with satisfaction. Learning design-related factors, such as delivery of teaching materials and learning activities, had a significant impact on learners’ overall satisfaction irrespective of student- or module-related characteristics (Rienties et al., 2015).

Computer self-efficacy refers to a person’s perceived ability to use a computer (Compeau & Higgins, 1995) and to carry out some activities related to computers. It is based on the concept of Bandura’s self-efficacy (1997), which is the conviction of an individual for his/her abilities. The measurement of computer selfefficacy indicates the confidence that one has in the use of computers and in the accomplishment of specific activities. According to the theory of self-efficacy, individuals undertake activities they believe they can accomplish, and they avoid situations where they believe they will fail. However, individuals with a strong sense of self-efficacy think that they can complete even difficult tasks. They treat them as challenges to be mastered, rather than as threats to be avoided (Bandura, 1994). People’s beliefs about their selfefficacy can be developed through the influence of four factors: mastery experiences, vicarious experiences provided by social models, social persuasion and physical and emotional state. The beliefs about self-efficacy determine how people feel, how they think, how they motivate themselves and how they behave. These beliefs produce diverse effects through four major processes: cognitive, motivational, affective and selection (Bandura, 1994).

Shen et al. (2013) found that female students are likely to have higher self-efficacy in e-learning than male students, suggesting that female students can be more active, seek more help or work better than male students. Students who participated in more online courses were more likely to have higher self-efficacy in e-learning and to successfully complete an online course. Moreover, they were more likely to communicate and collaborate with other students in academic work. According to Joo et al. (2013), students with higher perceived levels of internal control, self-efficacy and task value were more satisfied with online courses. These three predictors exercised an indirect significant impact on persistence, since satisfaction exerts mediating effects between predictors and persistence. Chien (2012) found that high computer self-efficacy leads to higher effectiveness of training. Self-efficacy was also found positively correlated with online learning satisfaction (Al-Nasa’h et al., 2021; Carranza Esteban et al., 2022), while it reduced general anxiety and fear (Al-Nasa’h et al., 2021). Furthermore, students exhibiting high self-efficacy showed lower levels of depression and emotional exhaustion (Carranza Esteban et al., 2022).

Knowles (1975), Moore (1980) and Brocket and Hiemstra (1991) defined self-regulated learning as a learning process where a student determines and has the ultimate responsibility for the purpose, content, method and evaluation of this process. Moore (1980) believed that when learners deal with difficult situations, they can seek information to acquire skills, set goals and determine the criteria for a successful effort in order to solve their problems.

Self-regulated learning includes metacognitive processes, procedures and behavioural motivation processes (Zimmerman, 1989). In terms of metacognition, self-regulated students plan, set goals, organise, self-monitor and self-evaluate during the procedure of knowledge acquisition. In terms of motivation, students exhibit high self-efficacy, self-performance and intrinsic interest. From a behavioural point of view, self-regulated students choose, plan and create environments that optimise learning. In the present study, we focus on metacognitive self-regulated learning, which is measured as the students’ perceived ability to regulate their studies.

According to Pintrich (2000), self-regulated learning is defined as a dynamic and creative process in which students set learning goals and try to guide, regulate and control their reactions, which are determined or limited by their targets and the environmental context.

Hase and Kenyon (2000, 2007) introduced the concept of ‘heutagogy’, in relation to self-determined learning, as a holistic approach to developing learner capabilities, with learning as an active and proactive process and learners serving as ‘the major agent in their own learning, which occurs as a result of personal experiences’ (Hase & Kenyon, 2007). The tutor facilitates the learning process by providing guidance and resources, while fully relinquishing ownership of the learning process to the learner, who negotiates learning and determines what and how will be learned (Hase & Kenyon, 2000).

According to Corno (2001), the characteristics of students who regulate their learning are the following:

They use cognitive strategies, such as repetition, processing and organisation, to represent, organise, edit and retrieve information.

Self-regulated learning includes the efforts made by the students to systematically manage learning processes to achieve their goals (Zimmerman & Schunk, 2011). Often, multiple structures explain the selfregulated learning of students (Artino, 2009; Azevedo, 2005; Cho & Jonassen, 2009; Zimmerman & Schunk, 2011). These structures include target setting, academic self-efficacy and adjustment to the learning context (Pintrich, 2004). The competent self-regulated students were reported to have higher intrinsic orientation of goals and higher academic self-efficacy than the less-skilled on self-regulation students. Furthermore, the skilled self-regulated students regulate and adapt better to the learning process in different learning contexts (Pintrich, 1999, 2004; Zimmerman & Schunk, 2011).

Studies have reported that students’ self-regulated learning is important to determine successful learning experiences, such as satisfaction and achievement in technological learning environments (Artino, 2008; Greene & Azevedo, 2009; Li, 2019). Moreover, it was shown that students’ self-regulated learning has a parallel mediating effect in the relationship between interaction and satisfaction (Wu et al., 2023). However, self-regulated learning is difficult for many students, especially in an online learning environment that may lack direct support; as a consequence, students may feel lost and socially isolated (Cho et al., 2010; Sun & Rueda, 2012). Ali and Leeds (2009) found significantly higher drop-out rates in online courses than in face-to-face classes. Lee and Choi (2011) found that the lack of capacity for self-regulation of online learning students is an important reason for the high number of dropouts.

According to Moore (1989), there are three types of interaction – and therefore three types of transactional distance – namely learner–tutor interaction, learner–learner interaction and learner–content interaction. The present study examines two of the three types of interaction that Moore suggested: learner–learner and learner–tutor interactions. Interaction is the good and effective communication between the teacher and the student, as well as between the students themselves; it contributes to the gradual self-confidence of the student and consequently in the achievement of his/ her goals. The tutor through this interaction covers the needs of the students both in the learning level and in the sentimental one (Holmberg, 1995). Interaction also encompasses constructive discussion, which is achieved through technological means, including synchronous communication. The lack of interaction leads to the increase of the transactional distance, that is, of the psychological and communicative space between a teacher and a student, or otherwise the distance between the teacher’s input and the student’s actual perception in an educational programme (Moore, 1989).

Learner–learner interaction is the interaction between two or more learners, alone or in group settings, with or without the presence of the tutor (Moore, 1989), by which they exchange information, thoughts or ideas about course content (Moore & Kearsley, 1996). The interaction between students is measured as the level of interaction that students perceive that they experience with their peers. It refers to the exchange of ideas, cooperation in carrying out assignments, as well as to the provision of comments and feedback.

Learner–learner interaction includes a mutual bidirectional communication between two or more students, which is extremely useful and essential for the learning process (Moore, 1989). While interacting with other students, the students can exchange ideas and get feedback from each other. An ideal distancelearning programme gives learners the opportunity to communicate with each other synchronously via teleconferencing, just as they would in traditional classroom situations, and asynchronously via message boards and email lists. Some programmes include distance and/or face-to-face meetings to provide group interaction (Moore & Kearsley, 1996). Due to the use of high-tech devices for interaction in distance education, Hillman et al. (1994) suggested a fourth type of interaction, that is, between learner and interface.

Fotiadou et al. (2017) found a statistically significant positive correlation between autonomy and both learner–learner interaction and teacher–learner interaction. Kassandrinou et al. (2014) found that most participants in their survey, who were postgraduate students at the HOU, were reluctant to communicate with their fellow students or did not consider it necessary to take the initiative to communicate and interact with other students, especially during group counselling meetings. Similar results were obtained in other studies on communication and interaction between students within the HOU courses (Angelaki & Mavroidis, 2013; Loizidou-Chatzitheodoulou et al., 2001; Tzoutza, 2010). Learner–learner interaction helps the exchange of ideas and information and enhances experiential learning. The lack of interaction between learners is a big problem in distance education. The absence of activities favouring interactions leads to the isolation of the learners (Belanger & Jordan, 2000). However, in modern e-learning environments, the synchronous or asynchronous interaction allows students not only to share information but also to determine how to retrieve useful information (Liaw, 2008). In this context, instructors are required to promote learning tasks based on collaboration and discussion to enhance learner–learner interaction (Kara et al., 2021).

Regarding learner–tutor interaction, the tutors seek to stimulate or maintain the students’ interest and to motivate the students to learn, by considering self-direction and self-motivation (Moore, 1989). They provide advice, support and encouragement to each student, depending on the educational level, the personality and the philosophy of each student and other such factors (Moore, 1989). Learners are not perceived as passive recipients but as autonomous and responsible individuals, who actively participate in the learning process (Lionarakis, 2009). Many studies, such as those of Tait (2003) and of Thatch and Murphy (1995), have shown that interaction with the tutor affects the effectiveness of distance learning, as the skills regarding immediacy and individual feedback contribute to satisfaction and learning. Anderson (2003), Battalio (2007) and Jung et al. (2002) observed that an increase in the teacher–learner interaction leads to increased satisfaction. Moreover, Turhangil Erenler (2020) examined all types of interaction within course delivery and found that they had significant positive effects on student satisfaction.

On the other hand, there are occasions where the interaction between learners and participation in group activities leads to reduced satisfaction (Berge, 1999; Northrup et al., 2002). There are also cases where the relationship between the interaction of learners and satisfaction was moderate (Baturay, 2011) or they are sometimes positively correlated. Overall, it is considered that the interaction between learners is a poor predictor of satisfaction, while learner–teacher interaction (and learner–content interaction) is a good predictor (Kuo et al., 2013). Gray and DiLoreto (2016) also found that learner–learner interaction does not have a significant impact on student satisfaction, while the instructor’s presence significantly affected student satisfaction, a relationship which was partially mediated by students’ engagement.

The main purpose of this study is to examine the relationship between students’ internet self-efficacy, student–student and student–tutor interaction, selfregulated learning skills of students and the satisfaction that students receive from participating in distancelearning courses. Furthermore, the effect of gender, age and other demographic characteristics on satisfaction, self-regulated learning, internet self-efficacy and learner–learner and learner–tutor interaction is estimated.

The research questions of this study are:

RQ1: What is the extent of (a) satisfaction of students from their learning experience, (b) self-regulating learning, (c) internet self-efficacy, (d) interaction with their tutor and (e) interaction between students?

RQ2: Are there any statistically significant differences in the examined parameters in relation to the different categories of demographic factors (i.e., gender, age, employment status, academic level, number of modules, previous experience in distance education, graduate curriculum, school)?

RQ3: What is the correlation between student satisfaction from their learning experience and the other factors under consideration?

RQ4: Can a regression model adequately estimate the probability that distance learning participants feel satisfied, when the values of the independent variables included in the model are known?

The HOU consists of four schools, namely applied arts, humanities, science and technology, and social sciences. It offers both undergraduate and postgraduate courses through open and distance education, using various methods for distance learning. The participants of this study were 122 postgraduate students, who attended three different distance-learning postgraduate programmes at the HOU. In particular, 66 (54.1%) students attended the programme ‘Master in Education’ of the School of Humanities, 34 (22.9%) were enrolled in the programme ‘Waste Management’ and 22 students (18.0%) in the programme ‘Environmental Design of Cities and Buildings’ of the School of Science and Technology. Students attending these postgraduate programmes have to successfully complete four course modules and to submit a postgraduate dissertation to obtain their postgraduate degree (more recently, this practice has changed and students can choose between attending a fifth course module and handing in a dissertation). For each course module they enrol in, students must hand in five written assignments, throughout the 10-month academic year, and take up exams at the end of it. Furthermore, each course module includes five face-to-face counselling group sessions (CGS). Participation in a CGS is not compulsory. Students should plan their own study during each course module while they are continuously supported by their tutor. The educational material is developed based on distance learning principles and is available to students in hardcopy before the beginning of the course. The course content, additional learning resources, the academic schedule, the submission of written assignments and the feedback from tutors, as well as communication with tutors and co-learners (through course and individual group fora) are accessed through an online learning management system.

An overview of how each of the examined parameters is concretely implemented in HOU is provided below, especially since student satisfaction emanating from each of these parameters is implicitly linked to the way it is implemented in a specific learning environment:

Self-regulated learning, through: written assignments, self-assessment exercises, studying in the period between CGSs.

The data for this study were collected in 2015, that is, in the pre-COVID-19 period. The results of this study would be of further interest for educational research and practice in the post-COVID-19 era, due to the increased interest in distance education and for improving not only academic results but also student satisfaction under these conditions.

The socio-demographic characteristics of the students participating in the survey are presented in Table 1.

The tool that was used to collect the data was a questionnaire consisting of six subscales (see Appendix). The first part deals with questions on the demographic characteristics. These are gender, age, employment status, academic level, previous experience in using computers, number of modules, previous experience in distance learning and the graduate study programme that the students attended. In order to measure ‘learner–learner interaction’ and ‘learner–tutor interaction’, two subscales derived from the distance education learning environments survey (DELES scale) by Walker and Fraser (2005) were used. ‘Self-regulated’ learning was measured by the ‘Motivated Strategies for Learning Questionnaire’ MSLQ by Pintrich et al. (1993), as adapted by Kuo (2010). ‘Satisfaction’ of students participating in distance education was measured using a questionnaire based on the scale of Arbaugh (2000), adapted to the HOU by Vakoufari et al. (2014). Internet self-efficacy was measured using a scale developed by Eastin and LaRose (2000) to measure the belief of people about the use of technologies based on the internet (web-based technology), as adapted by Kuo (2010).

The calculation of Cronbach’s alpha reliability coefficient for all subscales of the questionnaire used in this survey indicated the existence of internal coherence: for learner–learner interaction, it was equal to 0.827; for learner–tutor interaction, it was equal to 0.919, for self-regulated learning, it was equal to 0.701, for satisfaction, it was 0.909 and, finally, for internet self-efficacy, it was 0.908.

The data were processed and analysed using the IBM SPSS 21 software package. The absolute and relative frequencies and descriptive statistics were calculated. The Kolmogorov–Smirnov normality test was used to determine whether data analysis will be conducted with parametric or nonparametric tests. T-test for independent samples was used to check the differences between the mean values of satisfaction, learner–learner interaction, learner–tutor interaction, self-regulated learning and internet self-efficacy between the categories of the following variables: sex, faculty/school, and previous experience in distance learning programmes. Oneway Analysis of Variance ANOVA was used to test the differences between the mean values of satisfaction, learner–learner interaction, learner–tutor interaction, self-regulated learning and internet self-efficacy, between the categories of the following variables: age, employment status, graduate curriculum, number of modules and academic level. The Pearson correlation coefficient was used for calculating the correlation between satisfaction, learner–learner interaction, learner–tutor interaction, self-regulated learning and internet self-efficacy. Finally, logistic regression was used to build a model that predicts the probability of distance learning participants feeling satisfied.

In respect to the first research question, the mean (M) and standard deviation (SD) of the variables were calculated (Table 2). Learner–learner interaction levels were moderate (M = 2.53, SD = 0.86), contrary to the values of learner–tutor interaction (M = 3.95, SD = 0.77), self-regulated learning (M = 3.55, SD = 0.51), student satisfaction (M = 3.67, SD = 0.91) and internet selfefficacy (M = 3.77, SD = 0.79), which were higher.

As the five variables follow a normal distribution, independent samples t-test were used to identify any differences in mean values between males and females. In all five variables, the p-value is greater than the a = 0.05 significance level, which does not allow us to reject the null hypotheses, according to which there are no differences in mean values between men and women. To test the differences in mean values between the groups of students with or without previous experience in distance learning, an independent samples t-test was also used. The average learner–learner interaction, learner–tutor interaction, self-regulated learning, student satisfaction and internet self-efficacy among individuals, who have previous experience in distance learning and those who do not, did not differ statistically significantly. In all five variables, the p-value is greater than the significance level a = 0.05, which does not allow us to reject the null hypotheses, according to which there are no differences in mean values between individuals who have or do not have previous experience in distance learning. Similarly, an independent samples t-test was used for controlling the differences between the respondents participating in courses of the School of Humanities and the School of Sciences and Technology. All the p-values are greater than the significance level a = 0.05, which does not allow us to reject the null hypothesis, according to which there are no differences in mean values among students from the two schools. Furthermore, the differences in the averages of the five variables by age group (≥30 years, 31–40 years, 41–50 years and >50 years) were studied with one-way ANOVA. The results showed that there were no statistically significant differences between the age groups for all five variables, as the p-value in all cases is greater than the a = 0.05 significance level.

Regarding the effect of the employment status on the variables: learner–learner interaction, learner–tutor interaction, self-regulated learning and internet self-efficacy, the p-values are greater than the significance level a = 0.05 (Table 3). For the variable student satisfaction, ANOVA showed a statistically significant difference between the categories of employment status (F = 3.366, p = 0.038 <0.05). The averages of satisfaction in each employment status category were compared using the Tukey and Scheffe post-hoc criteria. This analysis indicated a statistically significant difference between the mean value of satisfaction among the categories of unemployed and full-time employed persons. More specifically, the Tukey and Scheffe test gave p = 0.032 <0.05 (p = 0.042 <0.05), with a mean difference of the satisfaction of the unemployed minus the satisfaction of full-time employed equal to -0.6005. Given that the average satisfaction of the unemployed is less than the average satisfaction of full-time employed, one-sided control with null hypothesis, ‘H0: The mean satisfaction of the unemployed is not different from the average satisfaction of full-time employed’, and alternative hypothesis, ‘H1: The mean satisfaction of unemployed is less than the average satisfaction of fulltime employed’, resulted in p = 0.016 <0.025 (p = 0.021 <0.025). Therefore, the null hypothesis is rejected, which means that the satisfaction of the unemployed is less than the satisfaction of the fully employed.

One-way ANOVA was also used to find out if there are statistically significant differences in the averages of the five variables in relation to (a) the level of academic studies (tertiary technical [ATEI]), tertiary university [AEI], Master’s and PhD), as well as (b) the three postgraduate programmes (Studies in Education, Waste Management, Environmental Urban Planning and Building). The results showed that there were no statistically significant differences, as the p-value in all cases is greater than the significance level a = 0.05.

Moreover, differences in the averages of the five variables in relation to the number of course modules (n = 1, 2, 3, 4) that the students have already followed were investigated. Because the category ‘1 module’ consisted only of one case, this category was rejected by the ANOVA analysis because the post-hoc criteria could not be calculated. It was found that the averages of all the variables apart from the learner–tutor interaction do not differ significantly (Table 4); for learner–tutor interaction, the ANOVA test showed that there is statistically significant difference regarding the number of modules (F = 5,993, p = 0.003 <0.05). According to the post-hoc criteria of Tukey and Scheffe, there is not only a statistically significant difference in mean learner–tutor interaction between the category of ‘two modules’ and the category of ‘four modules’, but also between the category of ‘three modules’ and the category of ‘four modules’. More specifically, the Tukey and Scheffe test gave a value of p = 0.014 <0.05 (p = 0.019 <0.05) with a mean difference ‘two modules’ – ‘four modules’ equal to -0.55 and p = 0.025 <0.05 (p = 0.034 <0.05) and mean difference ‘three modules’ – ‘four modules’ equal to -0.4144. As the average value of the learner–tutor interaction in the case of ‘two modules’ is lower than in the case of ‘four modules’, the one-sided control gave p = 0.007 <0.025 (p = 0.0095 <0.025), and therefore the null hypothesis is rejected, and the mean value of the learner–tutor interaction in the case of ‘two modules’ is smaller than the average value of the interaction in the case of ‘four modules’. Similarly, because the average value of learner–tutor interaction in the case of ‘three modules’ is lower than in the case of ‘four modules’, the one-sided control gave p = 0.0125 <0.025 (p = 0.017 <0.025), and therefore the alternative hypothesis “The mean value of the learner–tutor interaction in the case of “three modules” is smaller than the average value in the case of “four modules”’ is accepted.

Regarding the third research question, correlations between the variables (a) student satisfaction received from distance learning, (b) self-regulated learning, (c) internet self-efficacy, (d) learner–tutor interaction and (e) learner–learner interaction were assessed by calculating the Pearson correlation coefficient, after the Kolmogorov–Smirnov normality test reported that the variables follow the normal distribution. Table 5 shows that there are mainly small but statistically significant correlations between variables. More specifically, small or moderate statistically significant positive correlations at a significance level a = 0.01 appear between the variables (a) learner–tutor interaction and learner–learner interaction, (b) learner–tutor interaction and selfregulated learning, (c) learner–learner interaction and self-regulated learning, (d) learner–learner interaction and satisfaction, (e) internet self-efficacy and interaction with tutor, (f) internet self-efficacy and self-regulated learning and (f) self-efficacy and satisfaction, as the p-values are less than the significance level a = 0.01. Furthermore, the correlation between the variables learner–learner interaction and internet self-efficacy is statistically significant, with a positive small correlation at the 5% significance level. Finally, moderate, positive correlations that are statistically significant at a confidence level of 1% were observed between the variables satisfaction and learner–tutor interaction, as well as between the variables satisfaction and selfregulated learning.

In order to proceed with the development of a logistic regression model, a new dichotomous variable was defined. The cases with average satisfaction ≥3 were defined to reflect a value of 1 in the new satisfaction variable. That means that those students feel satisfaction. On the other hand, the cases with average satisfaction <3 were defined to correspond to a value of 0 for the new variable of satisfaction, and thus we accept that those students do not feel satisfaction from their studies. The aim of the development of the logistic regression model is to estimate the probability that distance learning participants feel satisfied, when the values of the independent variables included in the model are known.

The method that was chosen to introduce the independent variables in the model is the Forward Likelihood Ratio (LR) and the variables were included into the model according to their usefulness. In the first model (Block 0), only the intercept is included. Thus, if it is assumed that someone is satisfied with distance education, without having any other information, then our assumption is correct by 77% (Table 6).

Wald test equals 31,643 beyond p < 0,01, which means that the model is statistically significant (Table 7). The conclusion is that this model is better than random guessing about student satisfaction when any knowledge of the independent variables is absent.

As different variables (self-regulating learning, internet self-efficacy, learner–tutor interaction and learner–learner interaction) were tested to be added to the model, learner–tutor interaction was embedded in model 1, while in model 2, self-regulated learning was also incorporated. The -2log likelihood, which is a statistical measure of deviation between the observed and predicted values of the dependent variables, has fallen from 114.931 in the first model to 109.374 in the second model (Table 8). This indicates that adding the additional independent variable (self-regulated learning) in model 2 leads to an improvement.

The Hosmer–Lemeshow test (Table 9) indicates that the -2 log likelihood is statistically significant since χ² is statistically insignificant for both models (χ² = 11,798, p = 0.16 >0.05) and (χ² = 4,165, p = 0.251 >0.05), respectively, and thus they are effective.

In Table 10, the observed and predicted values of the dependent variable for both models are presented. Model 1 classifies incorrectly 21 out of 28 students who felt satisfaction, predicting incorrectly that they felt satisfaction. Similarly, 4 out of 94 students who felt satisfaction were incorrectly classified as having felt no satisfaction. So, model 1 classified correctly 79.5% of the data. In model 2, however, the incorrect rates fell slightly: 19 out of 28 students who felt no satisfaction were incorrectly classified as satisfied from their participation in distance education. Thus, compared to model 1, the percentage of correct classification of unsatisfied students increased from 25% to 32.1%. However, the percentage of correct classification of satisfied students remained 95.7%. Overall, model 2 had an increased classification rate of 81.1%.

The values of coefficients B are statistically significant (Table 11). In model 1, learner–tutor interaction’s coefficient equals 1.175 (Wald = 13.792, p = 0.000 <0.05) and the constant equals -3.244 (Wald = 7.468, p = 0.006 <0.05), whereas in model 2, learner–tutor interaction’s coefficient is 1.005 (Wald = 10.087, p = 0.001 <0.05), self-regulated learning’s coefficient is 1.101 (Wald = 5.166, p = 0.023 <0.05) and the constant equals -6.408 (Wald = 11.379, p = 0.001 <0.05).

Thus, model 2, which has a greater classification percentage is given by Eq. (1): (1)logp^1−p^=−6.408+1.005×(Learner- tutor interaction)++1.101×(Self − regulated learning )$$\matrix{ {\log \left( {{{\mathop {\rm{p}}\limits^\^ } \over {1 - \mathop {\rm{p}}\limits^\^ }}} \right) = - 6.408 + 1.005 \times ({\rm{Learner - tutor interaction}}) + } \cr { + 1.101 \times ({\rm{Self\;}} - {\rm{\;regulated learning\;}})} \cr } $$ where p^$${\mathop {\rm{p}}\limits^\^ }$$ is the probability of feeling satisfied by distance learning. Therefore, the probability of feeling satisfied is given by Eq. (2): (2)P(X)=11+e−(−6.408+1.005×(LTI)+1,101×(SRL))$${\rm{P}}({\rm{X}}) = {1 \over {1 + {{\rm{e}}^{ - ( - 6.408 + 1.005 \times ({\rm{LTI}}) + 1,101 \times ({\rm{SRL}}))}}}}$$

where LTI is learner–tutor Interaction and SRL is self-regulated Learning.

Practically, when both LTI and SRL equal 3, the probability of feeling satisfied by distance learning is approximately 0.477 or 47.7%. If the value of LTI remains constant at 3 while SRL increases by one unit, that is, 4, then the likelihood of distance learning satisfaction will be approximately 0.733 or 73.3%. Correspondingly, if LTI and SRL equal 5, then the likelihood of feeling satisfied is approximately 0.984 or 98.4%. On the other hand, if LTI and SRL both equal 1, that is, if the student does not show any self-regulated learning abilities and does not interact with the tutor at all, then the probability of distance learning satisfaction is 0.013 or 1.3%.

In Table 11, the Exp (B) column is related to the odds ratio for all predictor variables obtained in the model. In model 2, the odds ratio of self-regulated learning is greater than that of the interaction with tutors. This means that self-regulated learning is a better predictor of satisfaction than learner–tutor interaction. The Exp (B) value for self-regulated learning is 3.008. This indicates that if self-regulated learning increases by one unit and interaction with teachers remains constant, then the relative probability of feeling satisfied will increase by 200.8% (3,008–1,000 = 2,008), where relative probability is the ratio of probability of satisfaction from distance learning to the probability of not being satisfied. Correspondingly, the exp (B) value for LTI is 2,732, indicating that if LTI increases by one unit and self-regulated learning remains constant, then the relative probability of feeling satisfied will increase by 173.2% (2,732–1,000).

The results of the present study have indicated that the levels of learner–tutor interaction, satisfaction, self-regulated learning and internet self-efficacy were high between the participants. It is therefore indicated that students participating in this study are provided with advice, support, encouragement and motivation. Moreover, learners show characteristics that are crucial for regulating the way they learn and carry out their studies, as well as the skills needed for the use of internet. Learner–learner interaction was at lower levels, which is partly explained by the individual writing assignments that give no incentives to the learners to collaborate and interact. Learner–tutor interaction is influenced by the number of modules that the students have attended; interaction is reinforced as learners gain experience in distance education. Furthermore, the employment status affected the student satisfaction, which was attributed to the flexibility offered by distance programmes to the fully employed as well as by the overall lower satisfaction levels related to unemployment.

The logistic regression model that was built to predict the probability of learners being satisfied by distance education consists of learner–tutor interaction and self-regulated learning as independent parameters. Self-regulated learning, in fact, contributes more to the prediction of satisfaction than learner–tutor interaction.

To this end, it is recommended that educational practitioners take into account the findings of studies like the present one, and use techniques and activities that enhance the key parameters of distance learning and therefore student satisfaction. The inclusion of learner–tutor interaction in the satisfaction model underlines the importance of introducing relevant practices. Therefore, it is proposed to HOU to maintain and enhance the effectiveness of group counselling meetings and introduce other methods that enhance the interaction of learners with tutors, such as the use of further online meetings. Since the second variable that contributes to the prediction of satisfaction is the self-regulated learning, the need to design courses that promote selfregulated learning and also the need for enhancement of self-regulation skills by the educational system are highlighted. In addition, the establishment and use of online collaborative groups and the establishment of group assignments are important, as there is a general perception that collaborative learning supports self-regulation, since classmates model and discuss their own learning strategies and motivations, which are then distributed throughout the group. Thus, individuals can use and modify them to meet their own needs (Boekaerts & Corno, 2005).

Learner–learner interaction could be improved by using to a greater extent group-learning techniques in group meetings and beyond. This is now allowed also for online meetings (such as CGS through teleconference) through online platform features such as ‘break-out groups’. Students can therefore work together in small groups, which enables achieving higher student satisfaction than in larger groups (Chou & Chang, 2018), while allowing more direct and meaningful interaction, and they can more effectively assist each other in the learning process. Tutors can also encourage students to participate in online learning and discussion groups (either during a CGS or between CGS), for example, by asking them to prepare a joint assignment or even by asking to evaluate some simple activities/small tasks by colearners (peer evaluation) and by discussing these evaluations in small groups. There are several ways to set up such activities; the tutor needs to be an initiator and a moderator whenever needed, but should also enable the students to take initiatives, to effectively interact with each other, and to move outside their comfort zone of conventional instruction.

More research could be conducted to provide further insight, taking into account also the limitations of the present study, such as the relatively small sample size. A study with a larger sample that also covers undergraduate programmes of HOU, or even programmes from other distance learning environments, would allow further generalisation. Moreover, the findings of a qualitative research would provide further insight into the relations between the variables under consideration and would allow the validation of the findings through triangulation. Another limitation of our study is related to the fact that there are many factors/dimensions that have an influence of student satisfaction, and we only examined a limited number. Future research could examine a larger number of parameters or completely different parameters that can influence student satisfaction. For example, another variable that would be interesting to investigate is learner–educational material interaction and how this affects student satisfaction. Furthermore, the development of a predictive model that includes a greater variety of variables, which have emerged from the literature as important for student satisfaction, but also in general for open and distance education, would be of interest. Such variables could include learner–educational material interaction, learner–interface interaction, social presence, internal control and perceptual self-esteem. Finally, the design and implementation of a similar research in the post-COVID-19 era will be interesting to allow comparison and to identify any possible differences or new features.