The ability to collaborate effectively is increasingly crucial for professional success. Teamwork is highly valued in business environments, impacting employee performance, productivity, and creativity. However, not all students appreciate or benefit from group work, potentially due to a lack of experience and skills. Education should foster social skills through collaborative learning, considering student preferences. While student preferences for class organization vary and depend on context, existing educational data lacks adequate methods for evaluating preferences towards study modes and group work, often overlooking unobserved heterogeneity. Previous research using Likert scales has limitations in interpretation and may yield biased results. These scales assess preferences in isolation, failing to capture the complexity of human opinions and the relative importance of different attributes. This study addresses these methodological challenges by employing a Random Utility Model-based Discrete Choice Experiment (DCE) to analyze students' preferences for different study modes and school subjects in the context of achievement, measured using standardized tests linked to the international PISA scale. The study aims to understand students' attitudes toward group work compared to self-study and tutoring, examining how preferences vary across subjects and achievement levels. This approach allows for a precise measurement of preference variation and its explanation using achievement tests and student characteristics. It is, to the authors' knowledge, the first to use a DCE framework to analyze students' preferences for study mode and school subject simultaneously, while controlling for student learning outcomes.

The literature highlights the importance of group work and social-emotional skills for global competence and success in academic and professional settings. While group work offers advantages like knowledge exchange and enhanced critical thinking, it also presents challenges regarding individual responsibility and evaluation. High-achieving students often view group work as collaborative learning, emphasizing equal task distribution and quality, while low-achieving students may expect less effort due to task division. Preferences for learning modes vary across students, depending on performance levels, personality characteristics, and subject matter. Past experiences significantly shape these preferences, with negative experiences potentially biasing future perceptions. Positive student-student interaction is linked to more positive views of group work. The preference for working alone or with others is associated with sociological preferences and personality traits, with introverts tending to prefer individual work. High-achieving students often prefer individual work, possibly due to a desire for control and a focus on individual success, while low-achieving students might find group work beneficial for decision-making, particularly when paired with higher-ability peers. However, the literature presents contradictory findings on this aspect. The preference for working alone or in groups can also change with age and grade, reflecting a growing awareness of performance-based judgment in later educational stages and professional settings. Overall, preferences are multifaceted, influenced by beliefs, habits, social norms, emotions, moral principles, and personality. Satisfying these preferences contributes to well-being and may improve educational outcomes. Current research often oversimplifies learning preferences, lacking context and using potentially biased methodologies. This study aims to overcome these limitations through the use of a DCE approach.

This study uses data from the "Competences 2018" survey conducted in Poland, involving over 34,560 students. The sample, while not random, showed comparable achievement levels to the national average based on PISA scores. The study comprised three parts: preference assessment using DCE, competency assessments in mathematics, reading, and foreign language (linked to the PISA scale), and a general questionnaire. The DCE approach is based on the economic theory of consumers, the characteristics theory of demand, and the random utility model (RUM). Respondents were presented with hypothetical choice sets involving school projects, described by time commitment, working mode (individual, group of five, with parent/tutor), and subject (mathematics, Polish, English, geography). They selected their preferred alternative in six choice tasks. The data comprised 125,652 individual choices from 20,942 students (approximately 50% female). To model preferences, multinomial logistic (MNL) and random parameters logistic (RPL) regression were used. The MNL model assumes homogeneous preferences (fixed coefficients), while the RPL model allows for heterogeneity in preferences by letting coefficients vary randomly across individuals. Both models provided estimates for the marginal rates of substitution (MRS), representing the willingness to pay (WTP) for different attribute levels. The utility function considered studying time, study modes (with group work as reference), and subjects (with Polish as reference). Interactions were introduced into the model to explore heterogeneity. The IRT (Item Response Theory) model was employed to measure student achievement in mathematics, reading comprehension, and foreign language, which were linked to the PISA scale. The IRT-scaled achievement measures were used to assess the relationship between preferences and educational outcomes. Individual-level parameters were calculated using Revelt and Train's (2000) approach. A mixed logit model in WTP space was estimated to address issues with the ratio of two randomly distributed parameters in the random coefficient models. Multilevel analysis was conducted using individual parameter coefficients from the RPL model, incorporating student and school-level random effects.

The MNL and RPL models revealed consistent results. Studying time negatively impacted utility. Group work was generally preferred over other study modes (tutoring and parental help), although preferences varied. Among subjects, English was most preferred, followed by mathematics (counter to common perceptions of mathematics aversion). WTP analysis, calculated as the ratio of attribute coefficients to the time coefficient, provided a common metric for comparing attributes. The RPL model indicated significant heterogeneity in preferences across all attributes. A substantial proportion of students showed positive preferences for study modes other than group work, highlighting preference diversity. Analyzing WTP for high and low-performing students revealed that high-performing students were more willing to sacrifice time to choose group work over other modes, contradicting the notion that higher achievers consistently prefer individual work. Gender differences also emerged, with males showing more aversion to working with adults than females. The inclusion of interaction terms between study mode and subject revealed that preferences for study modes were not consistent across subjects, suggesting that the choice of study mode is context-dependent. Multilevel analysis showed the largest variation in preferences for studying with a tutor, both at the student and school levels, indicating considerable heterogeneity at the student level. The school level explained only a small percentage of the total variance in preferences, suggesting the student-specific aspects play a dominant role.

This study's findings challenge the common perception that Polish students dislike group work. The DCE approach revealed a general preference for group work, yet the specific features of group work and available alternatives greatly influence this preference. Students were willing to spend significantly more time working collaboratively with peers than with adult supervision. The observed heterogeneity of preferences underscores the importance of considering individual characteristics and achievement levels when designing educational interventions. The strong preference for group work among high-achieving students contradicts some prior findings, highlighting the limitations of using simplistic methods like Likert scales. The findings emphasize the importance of providing a range of learning modes to cater to diverse preferences. Contrary to some literature, there was not a strong correlation between gender and preference for group work. The significant interaction effects between study mode and subject suggest that the suitability of different learning modes is highly context-dependent and should be considered in the context of subject matter. The school level variance was relatively small, confirming the dominant role of student-specific preferences, suggesting that individual student preferences are more influential than school-level factors.

This study demonstrates that Polish students generally prefer group work but that this preference is influenced by several factors including the subject, the availability of alternatives and individual characteristics. The DCE methodology is shown to be a powerful tool for uncovering these preferences. Further research should investigate the role of prior experience, expectations, different types of group work, and other contextual factors. In addition, further research should replicate this study in different settings to test the generalizability of the findings.

The study's findings are based on a sample of Polish students, and the generalizability to other contexts may be limited. The survey did not explicitly measure prior experience with group work and the frequency of group work activities, which could influence preferences. Furthermore, the study did not differentiate between types of group work (e.g., collaborative vs. cooperative), nor did it control for group size, which could also influence preferences. Future studies could address these limitations and explore the influence of these factors on student preferences.