Chapter 4: Findings of the study
explained by the fact that the survey was mostly held among students that studied programs which are female dominant. The percent of those who are enrolled in the abovementioned programs, namely Program 2, Program 3 and Program 4, is equal to 58.3% of the total which is 91 students.
In addition to gender and program, participants were asked to indicate their form of study, the language of instruction, enrollment year and GPA. So, most of the students, namely 147, pay for their education, another 7 students have state grants and 1 student writes that he or she has a rector grant. More than 50% of students which are 86 students study their subjects in the Russian language, while the other 68 study in the Kazakh. Regarding their year of study, 51.3% are first-year students, 31.4% - fourth-year students and 17.3% of them are in their second year. Most of the students noted that their GPA varies from 4.0 to 3.67, at 73 students, and from 3.3 to 2.67, at 75 students, which is A and B grades. Only 8 students noted that their GPA equaled 2.33-1.67.
Table 2
Demographics of the Sample
Variable Category N %
Gender Female 114 73,1
Male 42 26,9
Program Program 1 65 41,7
Program 2 62 40
Program 3 21 13.4
Program 4 8 5,1
Form of study Tuition fee 148 94,9
State grant 8 5,1
Language of Instruction
Kazakh
69 44,2
Russian 87 55,8
Enrollment Year 2018 27 17,3
2017 80 51,3
2016 49 31,4
GPA 4.0-3.67 73 46,8
3.3-2.67 75 48,1
2.33-1.67 8 5,1
Students’ Perception of the Academic Integrity Policy
All participants were asked to evaluate their experience of the academic integrity policy (hereinafter referred to as the AIP) existing at the university (Table 3). Descriptive analysis showed that about 47% of students happened to have a medium level of awareness of the existing policy, 38% of them noted that their awareness level was ‘high’ and ‘very high’, while only 15% said it was low.
From the point of view of the respondents, students at the university frequently and always follow the policy, 30.8% and 35.9% respectively. Another popular response was
‘sometimes’ which was 25%. 34.6% and 22.4% of the participants indicated that university teachers follow the policy ‘often’ and ‘always’, 25% said that it happened ‘sometimes’, whereas for 17% it was ‘rarely’ or ‘never’.
The participants were asked to assess the effectiveness of the policy. So, 35.3% reported that they found the policy ‘effective’, while approximately the same proportion of students, at 31.4%, thought that it was ‘moderately effective’. 10.9% of students hesitated and noted that the policy was ‘somewhat effective’. Almost the same number of students found the policy very
‘effective’ and ‘ineffective’, 17 and 15 respectively.
Table 3
Students’ experience of Academic Integrity Policy
Variable Responses N %
What is your
awareness of the AIP
?
very low 9 5,8
low 14 9,0
medium 73 46,8
high 50 32,1
very high 10 6,4
never 3 1,9
rarely 13 8,3
To what extent do students follow the university AIP?
sometimes 33 21,2
often 48 30,8
always 56 35,9
other 3 1,9
To what extent do teachers follow the university AIP?
never 7 4,5
rarely 19 12,2
sometimes 39 25,0
often 54 34,6
always 35 22,4
other 2 1,3
To what extent is the university AIP effective?
ineffective 15 9,6
somewhat effective 17 10,9
moderately effective 49 31,4
effective 55 35,3
very effective 17 10,9
other 3 1,9
Further, ordinal logistic regression was conducted to reveal how nominal independent variables or factors predict students’ experience of the AIP expressed by ordinal variables, in other words whether such factors as gender, GPA, a form of study, academic program, the language of instruction and enrollment year affect the dependent ordinal variable. First of all, PLUM was carried out to see if the abovementioned predictors influence students’ awareness of the AIP. The calculation showed that significance level for all 7 factors was higher (p-value >
.05) than the cut-off point which means that it was statistically insignificant and, consequently, the changes in either gender, GPA, a form of study, program, the language of instruction and enrollment year did not cause the changes in students’ awareness of the AIP at the university. No predictions were also found to the extent students followed the AIP. The significance level was also higher than the cut-off point, i.e. p-value > .05. This implied that the gender of the
respondents did not explain the responses on how frequently students follow the AIP. The same influence neither had their program, GPA, form of study, year of study and the language of instruction.
When it came to analyzing the data for ‘To what extent do teachers follow the academic integrity policy’, regression revealed that the only factor that affected the students’ perception of the extent teachers follow the AIP was academic program. In this case, the analysis showed that the final model with two predictors fitted better than the baseline model without predictors, p- value = .002, but the level of fit was poor (Cox and Snell R2= .079, Nigelkerke R2= .083). The given predictor had two categories: ‘Program 1’ for category 1 and ‘Other’ for category 10, which is presented with Program 2, Program 3 and Program 4. ‘Other’ had zero coefficient in
‘Estimate’ column because it was used as a reference category. It implies that category 1 was significantly related to the responses on dependable variables ‘To what extent do teachers follow the academic integrity policy’ (p-value=. 001) with ‘estimate’ equal to 1,232 (see Table 4). This finding means that students of Program 1 were likely to believe that their teachers follow the AIP or the guidelines prescribed in it more than students of other programs. Moreover, the more students of Program 1 believe that teachers follow the AIP, the more students of other programs do so.
Table 4
Factors predicting students’ perception of the effectiveness of the AIP
Independent variable Dependent variable Estimates Sig.
Program To what extent do teachers follow the AIP
1,232 .001
Students’ attitude towards academic dishonesty acts
During the survey, respondents were asked to evaluate the acceptability of various academic dishonesty acts which were the examples of cheating, plagiarism, fabrication and facilitation. To analyze the data, descriptive statistics and ordinal regression were performed.
Further, in the fourth section correlation analysis was held to define the association with the likelihood of students’ involvement in academic dishonesty.
The survey had seven practices that respondents were asked to assess. These practices include the cases when students cheat using others’ work, allow others to cheat from their work and contract-cheating. The descriptive statistics showed that the majority of the participants did not approve of the use of cribs or messengers during the test. The number of those who ‘strongly disagree’ or ‘disagree’ was equal to 93 respondents on the former and 122 on the latter.
However, 20.5% of respondents hesitated and could not express their agreement or disagreement if it was acceptable to use cribs. It should also be noted that 31 respondents chose ‘strongly agree’ or ‘somewhat agree’, 9 and 22 respondents respectively. There were less respondents, namely 15, who ‘neither agree nor disagree’ on the acceptability of using messengers like what’s up, SMS and other ones during the test or exam. 19 students approved it by choosing ‘somewhat agree’ or ‘strongly agree’.
In terms of copying other students’ work during the exam or test or just homework, most of the respondents did not support these kinds of cheating. 86 respondents ‘strongly disagreed’ and 43 ‘somewhat disagreed’ that it was acceptable to copy other students’ answers during tests or exams. Moreover, 116 respondents did not favour even copying homework.
However, 23 participants did not find anything wrong in cheating on homework. As for the responses to ‘allowing to copy your work during a test or exam is acceptable’, 120 respondents did not support it opposite to 16 students who agreed with the statement. Students’ attitudes to contract-cheating, namely ordering papers online or from third parties, showed a similar pattern: the number of those who disagreed exceeded those who thought it was acceptable, 118 and 15 respectively. Regarding the practice where students act like suppliers or do the
homework or write the essays for money or other benefits, the data demonstrated no changes:
110 for ‘strongly disagree’ and ‘somewhat disagree’ and 20 for ‘somewhat agree’ and ‘strongly agree’.
When it came to plagiarism, students had to assess five cases. The case with the most frequents response ‘strongly disagree’ was ‘copy-pasting another students’ work and submitting it as own work is acceptable’ with 107 students. Another 33 participants indicated ‘somewhat disagree’. Moreover, this case acquired the fewest responses ‘neither agree nor disagree’ with 8 students and ‘somewhat agree’ and ‘strongly agree’ with 8 students which probably means that they have a clear understanding that such acts of academic dishonesty is wrong. With reference to other cases, most of the students did not approve of self-plagiarism and plagiarizing without references and in-text citations. The proportion of these responses constituted from 58% to 75%, unlike those who supported these acts – from 9% to 15%. What was interesting in these responses is that 43 participants, or 27.6%, neither agreed nor disagreed on ‘Using your own work more than one time without citing or referencing is acceptable’. Indeed, this case seems to be confusing for students as it is likely that they believe that there is nothing wrong with that.
There were only two cases offered to assess students’ attitudes towards facilitation and one case for fabrication. The frequency distribution shows that most respondents chose that it was normal to help other students to complete their assignments, at 28.8%. But, nearly the same percentage of students responded ‘strongly disagree’, at 26.9%, and ‘somewhat disagree’, at 23.7%. In terms of the situation when students seek help themselves, most students find it unacceptable for them, 34.0% for ‘strongly disagree’. The percentage for those who opted for
‘somewhat disagree’ is about the same as ‘somewhat agree’, 22.4% and 21.2% respectively. As
for fabrication or falsification, just under half of the total number of students find it wrong, 46.2% - ‘strongly disagree’ and only 6% agree that it is normal.
Overall, frequency distribution analysis demonstrated that students’ attitudes towards various academic dishonesty acts is mostly negative. But, there were no zero responses to such categories as ‘somewhat agree’ and ‘strongly agree’ which say that there is a share of students who do not find it unacceptable.
To reveal the factors that affect students’ attitudes towards academic dishonesty, ordinal regression, which results are shown in Table 5, was performed. Firstly, cases describing
cheating were analyzed. The PLUM showed that students’ attitudes to six of seven dishonest acts are likely to be caused by one or two factors. The only variable that did not experience any influence from independent variables was ‘allowing to copy your work during a test or exam is acceptable’. The relationship was not statistically significant because the p-value was higher than 0.05 (p>0.05), which means that the attitude to this case probably does not depend on students’ programs, gender, the language of instruction, GPA etc. As for the other cases, they can be divided into those which are predicted by enrolment year, the language of instruction and both enrollment year and gender. Students’ attitudes towards the dependent variable ‘Using cribs to answer test or exam questions is acceptable’ is influenced by enrollment year. The calculation showed that the p-value for category 1 (2018) is equal to .012 which shows the statistically significant relation to the year of study (p-value<.05). However, the significance level in ‘model fitting information’ showed that the analyzed model does not fit better than the baseline model (p-value=. 116) with no predictors and the improvement of this model over the baseline model is poor (Cox and Snell R2= .046, Nigelkerke R2= .049). Moreover, the
goodness-of-fit demonstrated that the given model predicts the outcomes different from actual
outcomes and this model does not fit the data (Person=.025, Deviance = .013). Regarding the variable ‘Using What’s up, SMS and other text messengers to get answers to test or exam is acceptable’, the calculation shows that it also had a statistically significant relation to
enrollment year. P-value for Category 1 (2018) is equal to .022 and for Category 2 (2017) it is .012. The coefficient for Category 3 (2016) is 0 since its significance is calculated and used as a reference category. It can be concluded that attitude to cheating by means of messengers may change upon the year of study which is likely to be caused by the length of study. It can be seen that the older the student, the more significant the relation. The significance level in ‘model fitting information’ was just above the cut-off point (p-value=.080), but still demonstrated that the model does not fit better than the baseline model with poor improvement fit (Cox and Snell R2= .052, Nigelkerke R2= .056). As for the goodness-of-fit, the difference between model outcomes and actual outcomes is not significant and could fit the data (Person=.328, Deviance
= .256). The data are rather contradictory as it implies that the model fits the data but does not fit better the baseline model.
There were two factors that are likely to predict students’ attitude to copying off other students during the exam or test: gender and enrollment year. According to the calculations, the relationship between gender and attitude to cheating from other students is statistically
significant. The significance level showed p-value=.034 for females and used males as a reference category. The information on whether the model fits better than the baseline model is not in favor of the model with p-value=.095 but with modest or moderate improvement over the baseline model (Cox and Snell R2= .030, Nigelkerke R2= .033). Goodness-of-fit demonstrates that the difference between model outcomes and actual outcomes is not significant and could fit the data (Person=.755, Deviance = .671) which again means that even if the model is worse than
baseline model it still fits the data. The enrollment year shows contradictory results which say that the relationship is statistically significant for Category 1 (2018 enrolment year) with p- value=.017 but insignificant for Category 2 (2017 enrolment year) with p-value = .400. The coefficient for the 2016 enrollment year is 0 as it is used as a reference category. Correlation analysis showed dubious data because the Pearson correlation demonstrated significant positive association between enrollment year and students’ attitude to cheating from other students, (r (156) = .184, p = .022), while the Spearman correlation defined insignificant positive
association, (rs(156) = .156, p = .051). The analysis of the model fit shows that despite being worse than the baseline model (p-value=.068) with poor improvement (Cox and Snell R2= .030, Nigelkerke R2= .033), it fits the data (Person=.537, Deviance = .442).
Further ordinal regression indicated that such factors as the language of instruction may predict the following dependable variables ‘Copying homework from another student is
acceptable’, ‘Ordering a paper online or from peer and submitting it as own is acceptable’ and
‘Doing the homework or writing an essay for your peers for money, for other benefits or for free is acceptable’. The information shows that there is a statistically significant difference in the attitude to cheating on the homework (p-value=.000) and contract-cheating (p-value=.000) between the students who study in the groups with different languages. The coefficient for Category 2 (Russian language) is equal to 0 as it is used as a reference category. The information about the model fit reports that it does not fit better than the baseline model (p-value= .089) but with strong improvement (Cox and Snell R2= .050, Nigelkerke R2= .054). Pearson and Deviance coefficients mark the model as fitting the data (Person=.979, Deviance = .971). The same pattern can be applied for the model analysis of variables describing contract-cheating cases: ‘Ordering a paper online or from peer and submitting it as own is acceptable’ and ‘Doing the homework or
writing an essay for your peers for money, for other benefits or for free is acceptable’. It can be interpreted with the fact that students whose language of instruction is Kazakh experience some difficulties related to the access to Kazakh online and published sources. Therefore, probably the contract-cheating rate among students with Russian-language of instruction is higher. According to cross-tabulation, such likelihood is supported for cheating homework, where the number of Russian-speaking students (n=16) who approve of cheating is twice higher than Kazakh ones (n=
7). However, there is no significant difference in the number of responses for the other two practices
Ordinal regression did not show much evidence that independent variables predict the outcomes for cases of plagiarism, fabrication and facilitation. For most of them, the significance level was statistically insignificant (p-value>.05). An exception were the cases of copy-paste plagiarism and helping other students. The PLUM defined that gender has an influence on students’ attitudes to ‘copy-pasting another student’s work and submitting it as own work’ (p- value=.026). Though, the model analysis demonstrated that this model did not fit better than the baseline model (p-value=.077) with moderate improvement (Cox and Snell R2= .032, Nigelkerke R2= .038). Cross-tabulation revealed that there are more females (93%) for whom copy-paste is unacceptable than males (81%), but still the proportion for both is too high to predict actual data for population.
The last but not least, ordinal logistic regressions identified the statistically significant relation of helping other students with their assignment to enrollment year (p-value=.034), namely only Category 2. On the other hand, the coefficient for Category 1 was insignificant.
Such discrepancy can be explained with the fact that the given model with independent variables does not fit the data better than the baseline model (p-value=.285).
Table 5
Factors predicting students’ attitude towards academic dishonesty Factors/independent
variable
Dependent variable Estimates Sig.
Enrollment Year Using cribs to answer test or exam questions is acceptable
-1.131 .012 Enrollment Year Using What’s up, SMS and other text
messengers to get answers to test or exam is acceptable
-1.051 -.860
1= .022 2= . 012 Enrollment Year Copying from another student during a test
or exam is acceptable
-1,249 .017 Gender Copying from another student during a test
or exam is acceptable
-.881 .034
Gender Copy-pasting another students’ work and submitting it as own work is acceptable
-1.020 .026 Language of
Instruction
Copying homework from another student is acceptable
15.671 .000 Language of
Instruction
Ordering a paper online or from peer and submitting it as own is acceptable
15.740 .000
Language of Instruction
Doing the homework or writing essay for your peers for money, for other benefits or for free is acceptable
16.123 .000
Language of Instruction
Seeking help of other students on your projects, exams or other individual works is acceptable
17.026 .000 Enrollment Year Helping someone to complete projects,
exams or other individual works is acceptable
-.702
1= .381 2= . 034 3= . Likelihood of students’ involvement in academic dishonesty acts
Having defined the factors that affect students’ attitudes towards academic dishonesty, the research continues with an attempt to discover if these factors influence the likelihood of students’ dishonest behavior. Descriptive statistics were used to reveal the frequency
distribution of students’ responses to statements about the involvement in various dishonest acts.
As descriptive statistics shows, most of the students do not consider violating academic integrity rules and, therefore, choose ‘definitely not’ or ‘probably not’. First of all, regarding
cheating, 87 students would not use cribs or notes to do a test or an exam and 107 students note that they would not use messengers for cheating. 130 said they would not be involved in
copying off other students and for 107 students there is no likelihood to copy the homework. It is interesting that the number of those who hesitated and chose ‘possibly’ comprises a fair share of the total responses. So, 44 students probably would resort to cribs and other notes, 32
students are likely to use the messengers, 34 would possibly copy homework and only 18 probably would copy off other students. As for the contract-cheating, a high number of students noted that they were not likely to buy papers or do the work for other students for money or other benefits, 125 students and 119 students respectively. The number of those who hesitated accounted for 23-24 for both variables.
When it came to plagiarism, 136 students pointed out that they would not use copy- paste from students’ papers in their work and 115 said that they would not copy word for word from various sources. Finally, 100 students wrote that they would not paraphrase without proper citation. It should be noted that 34 and 30 students expressed their hesitation in
answering the questions about paraphrase and copy-paste from sources. It could be connected with the fact that some students are confused with citing the paraphrase because they think that if they paraphrase with their own words then there is no authorship. As for the copy-paste from sources, this type of plagiarism is widespread since students come from schools where it is not punished and when they enter the university environment they might be confused. Opposite to copy-paste from sources, copy-pasting other students’ tasks are punished at schools which may explain that only 18 students would do that at university. Approximately the same figures were for self-plagiarism, 105 students chose ‘definitely not’ or ‘probably not’ and 35 students chose
‘possibly’.