Determinants of moral judgments about academic dis-honesty: an experimental approach

Ðàçìåùåíî íà http://www.allbest.ru/

NATIONAL RESEARCH UNIVERSITY HIGHER SCHOOL OF ECONOMICS

DEPARTMENT OF SOCIOLOGY CHAIR OF ANALYSIS OF SOCIAL INSTITUTIONS

Final qualifying work

Determinants of moral judgments about academic dishonesty: an experimental approach

Bykov A.V.

Moscow 2020

Abstract

The current investigation is an attempt to define the determinants of moral judgments about academic dishonesty. Factorial experiment with the use of hypothetical vignettes was conducted to examine the effects of description of presented situation (subjective vs. intersubjective vs. objective) in Study 2, protagonist's gender (male vs. female), language of instruction (native vs. non-native), social origins (high vs. low) and geographical origins (urban vs. rural) in Study 1 on first-year university students' judgment about acceptability and dishonesty of cheating. As hypothesized, decreasing the level of subjectivity in description increases students' assessment of acceptability and honesty of transgressive behavior in Study 2. We did not accept most of the alternative hypothesis in Study 1, however, the main hypothesis that the assessment of acceptability significantly differs from the assessment of honesty was confirmed in both Studies. We explain such results through the theory of moral judgments. The “first-order” judgments are more principle and rely on internal feelings of rightness and wrongness, while “second-order” judgments are personal obligations to act accordingly with a “first-order” judgments. Participants find cheating dishonest but acceptable under some circumstances. Current investigation illustrates that deontic understanding of academic dishonesty is rather problematic since students perceive cheating and plagiarism as quite honest and acceptable behavior.

Key words: Academic Dishonesty; Acceptability of Cheating; Moral Judgments; Framing Effect; Factorial Experiment.

Introduction

Background. The phenomenon of academic dishonesty has been well studied by scholars for generations. The primer goal of such studies is to define factors that increase or decrease academic dishonesty among students, mainly internal (Anderman & Danner, 2008; Anderman & Koenka, 2017; Koul, 2012) and external (McCabe, Trevino, Butterfield, 2001a; Murdock et al., 2008). We can find first attempts to cover this problem at the beginning of the 20th century (James, 1933) that demonstrate an unpleasant result for the academy and recent investigations show no difference (McCabe, Trevino, Butterfield, 2001b). The core assumption, that scholars of academic dishonesty use, is an assumption that students see cheating, fabrication, facilitating in cheating and plagiarism as dishonest behavior by itself. It is perceived as students are well aware that cheating is a dishonest behavior because they do everything to make sure their teachers do not see them as cheaters (Albas & Albas, 1993), so the own concept of dishonesty is not questioned and seen as a dichotomous deontic variable: either you obey the rules being honest or you break them being dishonest. However, we see this view on dishonesty rather problematic due to the fact that the definition of academic dishonesty is made by teaching staff and researchers (e. g., Cizek, 1999; Pavela, 1997) that contradicts the idea of honesty as a moral concept (Burton, 1963) and can be easily misinterpreted as academic ignorance or even if students know that they act “against the rules” it does not mean that they perceive such behavior as dishonest, because under some conditions it would be “fair” to cheat. Thus, the investigation on student's lay theories and lay judgments about transgressive behavior within the academy should be investigated to begin the discussion which scholars of academic dishonesty rely on.

Problem statement. There are some attempts that use the acceptability of dishonest behavior as a dependent variable in relation to different factors (Jensen et al., 2002; Murdock & Stephens, 2007) though the assessment of dishonesty itself has never been tested. Moreover, previous studies have mostly used cross-sectional correlation data to test hypothesis on factors of academic dishonesty except for rare quasi-experiments (Rettinger & Kramer, 2009; Rettinger, Jordan, Peschiera, 2004; Jensen et al., 2002; Murdock & Stephens, 2007; Wryobeck & Whitley, 1999) what does not allow building causal relationships and define the determinants of transgressive behavior. Hence, we suggest conducting two studies with the use of vignettes to be able to directly manipulate independent variables and answer the main research questions: which factors determine the lay judgments of students on academic dishonesty, and would different factors cause different assessments of honesty and acceptability, and would some factors mediate that difference? We will test our hypotheses on the dependent variables “assessment of dishonesty” and “assessment of acceptability” with relation to the set of independent socio-demographic (gender, language of instruction, social origins and geographical origins) variables in Study 1, and with relation to the level of subjectivity in the description of a situation in Study 2.

Assumptions. In the current research, we investigate how socio-demographic factors of the protagonist determine student's lay judgments about academic dishonesty: gender, language of instruction, social origins and geographical origins. In Study 2 we test only one independent variable “level of subjectivity”. By conducting this study, it can be shown that dishonesty and acceptability are not an autonomous concept from any social factors but perceived in relation to them.

1. Literature Review

1.1 Dependent variables

There are two related types of moral judgments: "first-order" and "second-order". A “first-order” judgments are judgments about the rightness or wrongness of an action; it is derived from the moral principles of the individual. A "second-order" judgments are an individual's judgments of a personal obligation to act in accordance with a "first-order" judgment (Kohlberg & Candee, 1984). An individual can also externalize responsibility for their own behavior. It should be mentioned here that the use of the vignette experiment in this study is also justified by the fact that even when making judgments about another person, the individual still puts himself in the place of this other.

Dishonesty

In the context of academic dishonesty, “first-order” judgments can be attributed to evaluating dishonesty of “against the rules” behavior. And though the concepts of honesty and dishonesty are embedded in moral reasoning (Ayal & Gino, 2011; Friedman, Blau, Eshet-Alkalay, 2016; Murdock & Jason, 2007), in the current investigation we understand morality as a general term that is present in lay consciousness. Moreover, some studies show that people perceive honesty as morality itself (Xu & Ma, 2015).

It is remarkable how both of these concepts (honesty and dishonesty) are represented in the academic community. If we try to find any article about honesty as a concept, we will easily find some, but if we do the same with dishonesty, we will get the result of “academic dishonesty”. Thus, honesty is a question of morality, but dishonesty is a statement of morality. This understanding of morality from above requires reconsideration.

So students may not rate “against the rules" behavior as dishonest, and here we want to answer the first main research questions: which factors determine the lay judgments of students on academic dishonesty, and would different factors cause different assessments of dishonesty, and would some factors mediate that difference.

Acceptability

«Second-order» judgments can be attributed to the concept of acceptability of cheating. In some studies, scholars of academic dishonesty investigate the perception of acceptability of the transgressive behavior (Jensen et al., 2002; Murdock & Stephens, 2007). There are many attempts that use the acceptability of dishonest behavior as a dependent variable in relation to different factors (Jensen et al., 2002; Murdock & Stephens, 2007).

Research shows that students who think cheating is morally wrong nonetheless find ways to neutralize or rationalize cheating (Murdock, Beauchamp, Hinton, 2008), to externalize responsibility for one's own behavior (Kohlberg & Candee, 1984). For example, students justify their transgressive behavior by finding the course material is too complex, that it is useless, that they risk losing their scholarship, that they do not have enough time to combine work with study, that the teacher is incompetent, or that the teacher is too demanding. Or, for example, the fact that cheating does not harm anyone, that everyone around is also cheating, people nearby do not try to hide their answers, friends asked for help with cheating or the fact that the teacher does not follow the students on the exam (Haines, Diekhoff, LaBeff, & Clark, 1986).

So, the second main research question is which factors determine the lay judgments of students on academic dishonesty, and would different factors cause different assessments of acceptability, and would some factors mediate that difference.

Thus, the concepts of academic honesty/dishonesty and the acceptability/unacceptability of cheating differ, and an individual may consider an action dishonest (a “first-order” judgment), but acceptable (a “second-order” judgment). An additional question in this study is to determine the difference between dishonesty and acceptability assessment.

1.2 Independent variables

Gender

Perception of gender roles is a part of everyday thinking (Butler, 2011), so we assume that this notion could be found not only on the scientific explanatory level of transgressive behavior but also in moral judgments among adolescents. So, in the present study we expect to find a significant difference between the perception of male and female degree of dishonesty.

In the question of moral reasoning and ethical decision-making the role of socialization is crucial. In the study of academic dishonesty researches have shown a relationship between self-esteem and likelihood of cheating depending on the student's gender: high-self-esteem in the case of female students reduced the likelihood to cheat, while the relationship between high-self-esteem in the case of male students and the likelihood of cheating was insignificant (Eisen, 1972; Ward, 1986). Scholars explain these findings through the difference in gender socialization: men's high-self-esteem leads to the independence of thought and action, while women's high-self-esteem leads to submission to the rules of the game. There are also some evidence to this theory based on the neutralization attitude measurement which shows that female students who cheated were more likely to neutralize their behavior, while the correlation between cheating behavior and tendencies to neutralize was not founded among male students (Ward & Beck, 1990). In other words, women are more likely to invoke the mechanism of excuse-making to justify transgressive behavior than men do.

The meta-analyses conducted by Whitley, Nelson and Jones, also supported the differential socialization theory of moral reasoning and showed that male students self-report about cheating behavior more than female students do and have a more positive attitude towards cheating in general (1999). Vignette experiments also show that perception of cheating behavior differs depending on the gender of participants (Rettinger, Jordan, Peschiera, 2004), though whether or not the gender of the protagonist itself affects judgments towards academic dishonesty has not been tested.

In our vignettes we use the factor of protagonist's gender as a between-subjects variable (male and female). It is done to exclude the possibility of attentional adjustment of judgments. Gender in the vignette scenarios is changed by using common male or female names.

According to socialization and gender role theories (Tibbetts & Herz, 1996; Sinn, 1997), we assume that participants regardless of gender will find female students' actions as less acceptable and more dishonest than in situations with male protagonists.

Language of instruction

One of the least studied factors of academic dishonesty is language of instruction. There are some investigations that show a positive correlation between non-native language of instruction and the likelihood of cheating or plagiarism (Eccles et al., 2006; Park, 2003; Hosney, Fatima, 2014). Language proficiency is related to student's concerns about their academic abilities and increases the possibility of transgressive behavior by students who are non-native speakers. Also, language proficiency is strongly related to necessary studying skills for the learning processes, such as note-taking, essay writing, questioning and academic referencing (Dunkel, Mishra, Berliner, 1989; Burnett, 2002). Non-native speakers found themselves in a more difficult position than their native-speaking peers, which could also explain a higher probability of turning to plagiarism or cheating. For them it is simply harder to learn, especially having performance orientation.

Moreover, language can also indicate the difference between cultural perception towards academic dishonesty. It is assumed that students from non-English speaking countries are more lenient towards academic dishonesty: in doing so and in perception of it (Eccles et al., 2006). One of the studies conducted on Chinese students by G. Deckert (1993) shows that understanding of the concept of plagiarism itself differs depending on the culture. While American students point out mostly external aspects of plagiarism that include concern for the author, peers, ownership rights, and responsibility for doing independent work, Chinese students recognize only internal aspects -responsibility for doing independent work - and perceive it egocentrically trough concern for hampering their own learning. Deckert even suggests, that instructors in Western universities “should not be perplexed when ESL students, who represent different ideas and educational experiences and who lack confidence in using English, violate Western standards of scholarship” (Deckert, 1993, p. 132).

In the global process of digitalization of education which also includes the unification of language (Crystal, 2012; Nunan, 2003) the question of academic dishonesty is significant because further changing of the curriculum into English-taught subject without taking into account students perception towards such changes could increase fraud and impair the desire to master a profession. In the vignette scenarios we include “language of instruction” as between-subjects variable (non-native and native). It is operationalized through language of instruction of the course in which the protagonist decided to cheat: English / Russian - foreign language and mother-tong respectively.

We hypothesize that all difficulties that relate to the learning process will decrease the assessment of dishonesty and increase the assessment of acceptability in a situation of transgressive behavior. So, the assessment of dishonesty will be lower if the course instructed in a foreign language and cheating will be more acceptable.

Social origins

Parental social status has a major impact on children's socialization and their life course. Many studies that were conducted show that good social origins of a child: high parental education, high parental social position, high parental income have a positive impact on children's mobility, educational attainment, future income, and vice versa (Byrne, Chandola, Shlomo, 2018; Bukodi & Goldthorpe, 2013; Erola, Jalonen, Lehti, 2016; Gugushvili, Bukodi, Goldthorpe, 2017). It is also suggested that the parental social class has a greater impact than, for example, parental education or housing ownership status on children's life course (Bukodi & Goldthorpe, 2009; Byrne, Chandola, Shlomo, 2018). In this sense it can be suggested that students with better social origins would less likely to cheat on an exam and see cheating behavior as dishonest than students in a worse situation as findings show in the study of likelihood of criminal behavior (Akee et al., 2010). However, it is unclear whether or not the perception of such inequality of “glass floor” and “glass ceiling” in relation to dishonesty presents in everyday consciousness.

That is why in the present study we decided to directly manipulate social origins of the protagonist in addition to the standard questions of participant's socioeconomic status. In vignettes we use the factor of protagonist's social origin as within-subject variable (high and low). In the participant's mind difference in parental social class interpreted through working position could also lead to future inequality of their children, due to the possibility of parents to help their offsprings with acquiring a better job and sharing social networks (Macmillan, Tyler, Vignoles, 2015). Socioeconomic status in the vignette scenarios is operationalized through parent's working position: engaged in manual labor at a textile factory / hold prestigious positions in the banking sector both manual and both non-manual respectively which is a standard dichotomized measurement for parental social class. And though there are some limitations related to more precise measurement of mother's contribution on children through education because unlike man's education woman's education poorly converted into employment and they are less likely to work full time (Korupp, Ganzedoom, Van Der Lippe, 2002), we expect that it is easier for participants to understand the socioeconomic status of protagonist more accurate if both parents would work under the same conditions.

Our compensatory hypothesis for social origins is that participants will consider a cheating behavior of student with low social origins as less dishonest and more acceptable and vice versa. Geographical origins

The latter factor is specific for modern Russia: Russian people in the economic and socio-cultural context divide the country into two conditional worlds: “Moscow” and “Non-Moscow”. The distance between these two worlds is perceived negatively by both residents of non-Metropolitan settlements and Muscovites themselves, sympathizing with the province and at the same time feeling irritated by the influx of migrants (Sedova, 2019). Another major center is Saint Petersburg, but since the current study is focused on Moscow university we will consider Moscow situation.

Migration in Russia has a pronounced centripetal character, when the population - mainly young people - tends to migrate to Moscow, St. Petersburg and the agglomerations of the same name. At the same time, young people migrate in the absolute majority (78% of them) “for education percusses”. This migration is often one-sided, because while studying, a person adapts to an independent life, finds a job and does not return to their hometown (Mkrtchan, 2017). Moreover, migration is often planned, and households in small cities invest as much as possible in the student's relocation and living in a large city, considering this as a strategic investment (Mkrtchan, 2017).

The provincial youth themselves associate moving to a large city primarily with getting a job (this is important for 95% of Russians when choosing an educational institution), high salaries and opportunities for professional self-realization that are not available in the province, and not with studying at a specific University or obtaining a specific specialty (WCIOM, 2020). Also in large cities leisure and other infrastructure is more developed: this includes the improvement of the city, shopping and sports centers, cafes (Mkrtchan, 2017).

4 out of 5 Russians believe that an applicant should choose a University in the regional capital or Moscow to "gain a foothold” in a major city. 43% believe that higher education is of higher quality in major cities. Also, one of the arguments supported by 37% of Russians who approve of educational migration is the expectation of getting a job with a high salary after graduation in a large city (LEVADA, 2018). Thus, we can see a high level of external educational motivation, when getting an education serves to achieve other goals, and this is realized by the students themselves (Bugrimenko, 2006). Such students are more likely to commit academic fraud than intrinsically motivated students who are focused on gaining knowledge and skills (Anderman, Koenka, 2017). Using Becker's theory of crime (Becker, 1968) we can hypothesize that transgressive behavior of students from rural areas in universities in large cities can appear to participants as rational acts to maximize utility. Making the decision to cheat depending on the possible costs and benefits, where the benefit is a successful educational migration to a large city (without loss of household investment) and the realization of opportunities in it, because the province is perceived by all layers of Russian society as a "hopeless place" (LEVADA, 2018).

At the same time, Muscovites understand that it is difficult to gain a foothold in Moscow only by getting an education in the capital, because there is higher competition in the labor market (LEVADA, 2018). The majority of opponents of educational migration to large cities (37%) believe that quality education can be obtained in small cities. And some of them, in support of their position, on the contrary, called the poor quality of education in rural areas, which is why graduates will not cope with studying in a large city (LEVADA, 2018).

The main role in the attractiveness of an educational institution for an applicant is played by the level of economic development of the region in which it is located (Prakhov, Bocharova, 2016). In general, the flow of students coincides with the general migration trends in the country, is associated with the heterogeneity of the development and distribution of universities in the country and is directed to major Metropolitan areas - Moscow and St. Petersburg (Mkrtchan, 2003; Kozlov, Platonova, Leshukov, 2017). As of 2015, the largest number of higher education institutions was concentrated there: 183 in Moscow and 77 in Saint Petersburg. At the same time, as of 2020, the top positions of the national rating of Russian universities, according to Russians, are mainly occupied by Moscow universities - MSU, MGIMO, MSTU of Bauman, Higher School of Economics (WCIOM, 2020).

In this regard, we decided to add the "geographical origins" factor to determine whether the origin of the protagonist affects the assessment of acceptability and dishonesty in a situation of transgressive behavior. Will the participants justify the actions of a visiting student in the capital due to the greater need to gain a foothold in the city at all costs? This factor contains two levels (urban and rural) in the vignette scenarios and is operationalized through the line that tells that protagonist was born and finish school in Moscow (urban) or in Ryazan region (rural).

We assume that the assessments of acceptability and honesty will be more favorable to non-Moscow protagonists due to their greater need to “get a better life”.

Subjectivity

In Study 2 the same dependent variables were measured on the same sample. However, instead of testing another hypothesis of which protagonist characteristic determines lay judgments towards academic dishonesty, we decided to investigate how the presentation of the information about transgressive behavior in vignette's scenario itself influences assessments of acceptability and dishonesty.

Our core assumption is rooted into framing effect. In their research, Tversky and Kahneman showed that the presentation of information significantly affects decision-making. For example, if a participant is given a contract describing the possible gains from risk, he or she is more likely to refuse to sign it than in a situation where the description focuses on the losses he or she will incur if a participant do not take the risk (Tversky & Kahneman, 1981). Other studies also confirm that different descriptions of the same situation lead to different judgments. For example, Levin and co-authors vary the amount of information and its presentation - negative and positive - and found out that it affects participants' consumer judgments (Levin et al., 1985). In our study, we would like to test whether the perspective of the written vignette determines the perception of honesty and the acceptability of transgressive behavior. We will consider three possible perspectives of situations described in more detail by Traugott (1984): subjective, intersubjective, and objective. The "subjective" perspective should be understood as a description that depends on the “experiential control" of the protagonist, his evaluative perspective. The "intersubjective" perspective can be described as shared perspective, it is not necessary true but detached from protagonist's perspective. The "objective" or external description does not involve any intervention on the part of the protagonist, it is completely focused on the given object.

The matter of subjectivity is a concern of “cognitive grammar”. A framework that considers the use of subjective statements as a key to coherent storytelling. Subjectivity is seen as a situation of maximum asymmetry between the observer and what is observed. However, this is also an interesting point, because when reading the description of the situation, we find the narrator (protagonist), on the one hand, responsible for observing the scene, and as an actor of the scene, on the other hand, having thus dual role as the observer and as the object of observation (Langacker, 1985). As the narrator becomes involved in the situation, his responsibility for what is happening increases - it is he who focuses on these specific details, he who considers them important, and this is only his personal opinion about what is happening. The very first-person description of reality is not actual knowledge, instead it is «feeling of knowing” or “subjective knowledge” (Raju, Lonial, Glynn-Mangold, 1995). As the perspective shifts from external to internal, which expresses the observer's internal experience, or to intersubjective, a whole range of linguistic effects can occur, such as purely immediate implied meanings or diachronic change in meaning (Krawczak, 2016). Since meanings themselves are not directly determined by objective reality - they change depending on how the situation is constructed and presented (Langacker, 1985), we expect participants to pay attention to the degree of reliability of the story - the opinion of one interested person, the entire class, or objectively presented information.

In the study of Academic Dishonesty the results of testing this hypothesis could be significant, especially for future investigations using vignette experiments since experiments are sensitive to changes and additions of any factors. A significant difference in the perspective of the story will help to better formulate scenarios and neutralize the possible effects of empathy, through detached storytelling.

To be able to manipulate these three levels of one factor within vignettes we use one scenario which awakes neutralizing attitude (Murdock, Beauchamp, Hinton, 2008) in participants' minds. First, a tolerant environment towards cheating: a high percentage of cheating students leads to the normalization of cheating and, as a result, to an increase in the share of cheating students (McCabe, Trevino, 1997; Pulvers, Diekhoff, 1999). This was described through the line that tells that "Artem's (the protagonist's) peers often copying homework from each other". Second, performance goal orientation (Anderman, Griesinger, Westerfield, 1998), when the learning process is organized in such a way that its main goal is to demonstrate abilities and knowledge, rather than learning the material and really developing skills. One of the signs of performance goal orientation can be a focus on achieving high scores. So, this was reflected in the text of the vignette through the line "For him the main goal in education is getting good grades". Third, bad teaching competence (Murdock, Miller, Kohlhardt, 2004) - “the teacher does not prepare well for classes and generally poorly understand his subject”. All this was done to reduce the likelihood of getting an "zero" rating, meaning "absolutely unacceptable" and "absolutely dishonest" in terms of transgressive behavior. By filling the scenario with contextual meaning that causes the participant to have a neutralizing attitude, we hoped to get a difference in responses due to the description of the situation itself instead of meaningful changes within the situation.

The participant received one randomly selected vignette, and then answered two questions about the acceptability and dishonesty of the protagonist's actions. In the vignettes, only a few words were changed, reflecting the unique perception of the protagonist, the shared perception of the protagonist's peers, and an impersonal description. In the case of a “subjective” level, the situation was accompanied by personalized words such as “Artem seems...”, “Artem thinks...” and “Artem believes...”. The “intersubjective” level assumed being in a shared reality condition where “Artem's peers seem...”, “Artem's peers think...” and “Artem's peers believe...”. On the last - “objective” - level all words related to opinion were excluded and the situation is presented in the form of a fact description.

We assume, that more detached description will affect more favourable assessments, since reading scenario would appear as factual knowledge instead of “feeling of knowledge” and therefore participant would consider a scenario as objectively bad and unfair for protagonist.

2. Methodology

2.1 Justification

Scientific justification: lack of experimental designs in the study of academic dishonesty. At the same time, research claims the presence of determinants based on correlation analysis, which does not allow to determine causal relationships. Existing research either identifies the characteristics of a typical cheater or experiments illuminate actual cheating, but miss the reasons behind these actions. academic dishonesty transgressive behavior

Social justification: misinterpretation of the determinants of academic dishonesty and attitudes towards it does not help, but rather hinders the conduct of a competent academic policy against plagiarism and cheating

2.2 Method

For the current investigation we decided to conduct a factorial experiment using vignettes for 3 major reasons. First of all, the topic of transgressive behavior in academia is very sensitive, especially if a student is asked by a professor from the same university. Self-reporting questions can mislead the researcher with interpretation: instead of measuring dishonest behavior on the exams, scholars measure how honest a student, answering this question. Vignette's scenarios about the protagonist neutralize this problem by transferring the judgments to the other. Second, only the experiment allows to build causal relationships and define not only correlation but causation between variables. Also, we can define a major problem of measurement in the previous studies relying on the use of the questionnaires which are usually based on self-reporting questions (e.g., Dans, 1996; Hilbert, 1985; McCabe & Trevino, 1996). Again, it is crucial not to analyze reporting of dishonest behavior as dishonest behavior itself for the simple reason that not every cheater would confess its own fraud, especially to a faculty member, moreover, not every cheater thinks about himself as a cheater, which also actualizes the usage of vignettes. In the current investigation the empirical objects are judgments about academic dishonesty. Subjects of research are determinants of moral judgments about academic dishonesty. Here we would like to highlight that our empirical objects are judgments and not a certain social group, however, judgments of students are more valuable in the sense that they are themselves active participants of academic reality and academic policy towards academic dishonesty concerns them the most.

2.3 Measurements

In our vignettes the dependent variables were measured by:

“Assessment of dishonesty” which is measured by the question about dishonesty of protagonist in the case of transgressive behavior (cheating during the exam): “do you think, that (name of protagonist) acted honestly or dishonestly in this situation, and if so, to what extent? (11 points scale from 0 to 10, where 0 is completely dishonestly and 10 is completely honestly)”.

“Assessment of acceptability” which is measured by the question about the acceptability of actions of protagonist in the case of transgressive behavior (cheating during the exam): “do you think, that actions of (name of protagonist) acceptable in this situation, and if so, to what extent? (11 points scale from 0 to 10, where 0 is completely unacceptable and 10 is completely acceptable)”.

2.4 Pilot Study

The pilot study was divided into two parts: cognitive interviews and an online pilot questionnaire.

The tasks of the first stage were:

1. To determine what meanings the subjects put in the proposed vignettes and how understand key concepts of "honesty” and" acceptability” and evaluate the semantic adequacy of responses;

2. To identify the process of forming judgments by subjects, including explanation of the reasons for the response, confidence in giving the answer, the basis for making judgments;

3. To identify the main difficulties encountered by the subjects when reading situations and answering subsequent questions, including the completeness of the proposed information;

4. To determine how believable the proposed situations seem to the subjects;

5. Identify the weak points of the welcome page and the block with socio-demographic questions.

The tasks of the second stage:

1. To check the reliability of measurements and correct display of questions and answers;

2. To evaluate the average time spent on completing the questionnaire: check for the absence of monotony of responses, and evaluate the response rate.

When compiling the questions (see the Guide in appendices) we were primarily guided by the method of cognitive interview, the method of verbal Protocol, and the method of expert assessments, described in detail by G. Willis (1999). In total, we interviewed 7 subjects, all of them second-year students of different programs at the HSE. The average time for conducting cognitive interviews with each subject is 30 minutes.

As a result of the interviews, it was revealed that the subjects perceive the basic concepts of honesty and acceptability in accordance with the given meanings. Several adjustments were made to the wording and order of the proposals. The main changes were with the change of the experimental design from mixed factorial design to between-subjects design since a large number of vignettes increased the cognitive load of subjects and they had a desire to prematurely drop the study. The online source that we used to conduct the research worked without distortion and was user-friendly. The average passage time of the already corrected survey did not exceed three minutes.

2.5 Participants

Participants were 2204 (29,2% of the population) students of the first-year cycle from one University - Higher School of Economics (HSE) Moscow campus. Participation was voluntary for students if they fit the criteria of being first-year student of the first cycle (bachelor or specialist) studying in Moscow campus. The population (all first-year students from HSE Moscow campus) equals 7540. The sample included 37,8% of male students and 62,2% of female. 51,5% of them were from Moscow (finished school in Moscow) and 48,5% were from other places (finished school not in Moscow). The sample was not representative of the population since participation was voluntary and we had high non-response for the invitation to participate in research. Nonetheless, with such number of participants the statistical power is high enough to decrease the probability of making a type II error and to define even small effects as significant.

2.6 Procedure

Initially, the entire population of first-year students at the HSE Moscow campus was gathered by comparing academic ratings for the first and second modules of the academic year. After that, the students were mixed up and a letter with an invitation to participate in the experiment was sent to the corporate addresses of 7540 students, of which 4129 followed the link to the survey, but only 2204 reached the end. The current research was approved by the HSE Ethics Commission: “it is fully appropriate to conduct this survey in the NRU HSE”.

After answering the first question about whether or not they are first-year students of first cycle of HSE Moscow campus, participants were randomly assigned to read one of three possible vignettes from Study 2 and answer two questions: the first one was about acceptability of protagonist's action (cheating) and second was about dishonesty of such actions. So, here we use between-subjects factorial design with one factor (presentation of situation) on three levels: subjective, intersubjective and objective. An example vignette (subjective presentation) translated into English read:

Artem is a first year student at one Moscow University. He believes that everyone around him often cheats on exams and copies each other's homework. Artem himself thinks that the main goal in education is getting good grades. At one of the courses, it seems to him that the teacher does not prepare well for classes and generally poorly understand his subject. At the exam in this discipline, for which Artem does not have time to prepare, he sees the answers from a neighbor on the desk and copies them.

Then participant completes the same questions following another random vignette from the whole vignette population from Study 1 without possibility to go back and change answers following previous vignette. And though vignettes' scenarios were completely unrelated, we wanted to neutralize “order” and “learning” effects by separating Studies and using between-subjects factorial design in both cases. Another reason to use between-subjects factorial design in Study 1 is the number of similar stories (16 in total) that increase cognitive load decreasing concentration needed to focus on differences between vignettes. Moreover, within-subjects factorial design is appropriate only for octagonal factors that is not the case in Study 1. The total population of hypothetical vignettes equals 16 - 2 (gender) X 2 (language of instruction) X 2 (social origins) X (geographical origins). Operationalization of each factor is presented in Table 1. An example vignette (a1b1c1d1) translated into English read:

Andrew was born and graduated from school in Moscow. Andrew's mother is engaged in manual labor in a textile factory, and his father - in a steel factory. Now Andrey is a first-year student at a Moscow University. He doesn't have time to prepare for the exam in one of the subjects taught in Russian. In this exam, Andrew sees answers from a neighbor on the desk and copies them.

Table 1 Experimental factors and its levels for Study 1

After answering the questions about acceptability and dishonesty of protagonist's action, students are asked to complete the last three questions about their sociodemographic characteristics: their gender, where did they finish school and their bachelor or specialist program. The last section was not obligatory and was used if a participant have questions for the researcher.

3. Results

3.1 Hypothesis Testing

To test the hypothesis that presentation of the situation, gender of protagonist, language of instruction, social origins and geographical origins of protagonist affect students' perception of acceptability and dishonesty of transgressive behavior, we performed mixed ANOVA with the usage of factors as character vectors for both Studies. We use mixed ANOVA since the variable type of question is within-subject variable and therefore we cannot use full between-subjects design. All covariates (gender and geographical origins) were not found significant in experiment and therefore gender and geographical origins do not affect moral judgments toward academic dishonesty. In the present study we use III type Sum of Squares that corrects each factor in relation to other elements of the model. By using these main effects of the factors may occur significant rather than effects of interaction if we use II type (Langsrud, 2003) which satisfies our hypothesis. During the analysis and testing of hypotheses that means are the same for assessments of acceptability and dishonesty, confidence intervals were constructed using its correction by the Bonferroni method, which controls the group level of á and protects against errors of the first type.

Current answers are not distributed normally due to Kolmogorov-Smirnov and Shapiro-Wilk tests (see Table 4 and Table 5 in appendices). However, the ANOVA model can be considered stable to an non-normal distribution under certain conditions: first, according to the Central limits theorem, in order for the data to be considered normally distributed, it is necessary to have a sample of at least 30 observations; second, the variation of residuals within groups must be the same for all values of the dependent variable (the Levene test showed that we cannot reject H0, and our residuals are homoscedastic) for both Studies (see Table 2 and Table 3 in appendices); third, the number of the least populated group should not be less than 20% of observations and our data also meets these condition (Lunney, 1970, p. 267). Due to the lack of within-subjects variables containing more than three levels, there is no need to check data for sphericity.

3.2 Study 2

All in all, 844 participants completed “Subjective” vignette, 818 - “Intersubjective” and 815 - “Objective” in Study 2. Box's test of Equality of Covariance was not significant on 95% confidence level, so we cannot reject null hypothesis of equality of covariance across groups (see Table 6 in appendices). The multivariate tests showed a significant main effect of “question type” as within-subject variable F(1, 2474) = 1057, p < .05 with partial ç2 = 0.299. That means that 29,9% of variance in acceptability and honesty estimation is explained by the type of question and not the level of “subjectivity”. The interaction effect between “question type” and “subjectivity” was not significant on 95% confidence level (see Table 7 in appendices). Test of Between-Subjects Effects showed a significant main effect of “Subjectivity” factor that contains three levels and have main effect F(2, 2474) = 21, p < .05 with partial ç2 = 0.017 (see Table 8 in appendices). That means that 1,7% of variance in acceptability and honesty estimation is explained by the level of “subjectivity” and not by type of question.

Pairwise-Comparisons was conducted on 95% confidence level to answer the question, whether or not the “Subjectivity” differs on its levels. For such purposes we compared means estimated depending on both assessments since the effect of interactions between two variables was not found significant (see Table 7 in appendices). As we can see in Table 9, the same line pattern occurs. “Objective” settings are significantly differ from “subjective” and “intersubjective” settings. On average, students evaluate transgressive behavior as more acceptable and more honest in “objective” settings than in “intersubjective” (on 0,644 points less) and even less in “subjective” (on 0,703 points less). Estimated Marginal Means were 5,975 for “objective” settings, 5,331 for “intersubjective” and 5,272 for “subjective” settings (see Table 10 in appendices). Interestingly, the relation between “intersubjective” and “subjective” settings was not found. The results mean that students regardless of the question type would assess cheating more favorably in “objective” settings than in “intersubjective” and even more compared to “subjective” settings (more acceptable and honest).

Table 9 Pairwise Comparisons

Same Pairwise-Comparisons was conducted on 95% confidence level to answer the question, whether or not the assessment depends on “Type of question”. For such purposes we compared means estimated depending on two questions: about acceptability and about honesty since the effect of interactions between two variables was not found significant (see Table 7 in appendices).

Table 11 Pairwise Comparisons

The found difference was significant, which means that students assess two questions differently. Students regardless of experimental manipulation with the level of subjectivity would assess cheating as more acceptable behavior than honest behavior (mean difference equals 1,664) as seen in Table 10. Estimated Marginal Means were 6,358 for acceptability and 4,694 for honesty (see Table 12 in appendices).

3.3 Study 1

For Study 1 we conducted mixed Anova as well to take into account not only four between-subjects variables, but also one within-subjects - type of question. Box's test of Equality of Covariance was not significant on 95% confidence level, so we cannot reject null hypothesis of equality of covariance across groups (see Table 13 in appendices). Conducting Multivariate tests (see Table 14 in appendices) with tests of Within-Subjects Effects (see Table 15 in appendices) and Tests of Within-Subjects Contrasts (see Table 16 in appendices) on 95% confidence level showed the significant main effect of “question type” as within-subject variable F(1, 2357) = 683, p < .05 with partial ç2 = 0.225. That means that 25,5% of variance in acceptability and honesty estimation is explained by the type of the question and not by other factors. The effects of interaction also occur in the case of “type of the question” and “social origins”: F(1, 2357) = 4,362, p < .05 with partial ç2 = 0.002. This effect is rather small it explains only 0,2% of variance in acceptability and honesty estimation by such interaction and not by other factors. Another effect of interaction occurred between three factors: “type of the question”, “gender” and “language of instruction”: F(1, 2357) = 4,025, p < .05 with partial ç2 = 0.002. This effect is rather small it explains only 0,2% of variance in acceptability and honesty estimation by such interaction and not by other factors. The last effect of interaction is between “type of the question”, “gender”, “social origins” and “geographical origins”: F(1, 2357) = 10,782, p < .05 with partial ç2 = 0.005. This effect is a little bigger than effects of interactions before and it explains only 0,5% of variance in acceptability and honesty estimation by such interaction and not by other factors. All of these small effects could be interpreted as a measurement error due to the large effect size that determines even small differences.

However, if we exclude within-subject variable “question type” and conduct Tests of Between-Subjects Effects (see Table 17 in appendices) it will show a different result. On 95% confidence level the main effect of “social origins” is significant: F(1, 2357) = 7,246, p < .05 with partial ç2 = 0.003. That means that still only 0,3% of variance in acceptability and honesty estimation is explained by the social origins of protagonist and not by other factors. The effect of interaction also occurred here between two factors: “gender” and “geographical origins”: F(1, 2357) = 4,688, p < .05 with partial ç2 = 0.002. This effect is rather small it explains only 0,2% of variance in acceptability and honesty estimation by such interaction and not by other factors. The last effect of interaction occurred between two factors: “social origins” and “geographical origins”: F(1, 2357) = 7,605, p < .05 with partial ç2 = 0.003. This effect is rather small it explains only 0,3% of variance in acceptability and honesty estimation by such interaction and not by other factors.

Other main effects of independent variables “gender”, “language of instruction” and “geographical origins” were not found significant. That means that “gender”, “language of instruction” and “geographical origins” of protagonist do not affect students' judgments about acceptability and honesty in situation of transgressive behavior. Even “social origins” which is significant on 95% confidence level is almost indistinguishable between its levels as shown in Table 18. Still, we can say that students regardless of the question type and other factors would assess cheating more favorably if protagonist's social origins are low (estimated marginal mean = 4,8) than high (estimated marginal mean = 4,6) (see Table 19 in appendices).