Collaborative learning has been extensively documented in the last few decades ^[1][2][1,2] and it is quite common that university teachers assign collaborative tasks in higher education [3]. Collaborative tasks contribute to the development of collaboration and communication skills ^[4][5][4,5]. Moreover, students also acquire domain-specific knowledge ^[3][6][3,6] through several cognitive activities that are taking place during collaboration, such as collecting and sharing information, creating explanations for concepts, negotiating ideas, and building common ground [7].

One specific collaborative task, often implemented at university, is collaboratively writing papers ^[8][9][8,9] Besides the abovementioned general benefits of collaborative learning, collaborative writing aids students in learning to select relevant and reliable information, integrate information from various sources, and to distribute their own ideas to a broader audience ^[9][10][11][9,10,11]. Collaborative writing is nowadays often computer-supported. For example, students use collaborative writing tools such as Google Docs or Etherpad to collaboratively produce texts ^[12][13][14][12,13,14], and can use online meetings to interact and provide feedback while working remotely ^[15][16][15,16].

Notwithstanding the consensus in the literature regarding the many benefits of collaborative learning ^[17][18][19][17,18,19], difficulties and issues are also reported. Common issues include interpersonal conflicts, free-riding ^[1][20][1,20], and the imbalanced skills and knowledge of collaborating students [21]. These issues obviously impact the collaborative process and subsequently the group product. Taking this into account, it can be said that group formation, as a first step in collaborative learning [22], is crucial.

Several approaches can be followed in forming groups. Groups can be assembled by university teachers, deliberately or at random, or by students themselves ^[23][24][23,24]. When not forming groups randomly, university teachers often depart from a cognitive perspective ^{[21][25][26][27]}[21,25,26,27]. However, university students’ perspectives on group formation are less clear.

Collaborative writing is a specific type of collaborative learning. While collaborative learning entails “all instructional arrangements that involve two or more students working together on a shared learning goal” (p. 71) [28], collaborative writing can be defined as “the coauthoring of a single text by two or more writers, where the coauthors are involved in all stages of the composing process and have a shared ownership of the text produced” (p. 1) [29]. Collaborative writing tasks are very specific and complex, requiring extensive knowledge on vocabulary, grammar, content, sourcing, information selection, and writing ^[9][30][9,30]. The contribution of individual group members heavily determines the outcome. Well-composed texts can result from unequal participation [31] as each member brings in other competencies and knowledge [32].

A group consists of at least two students collaborating on the same product. The optimal group size may vary according to the goal and the type of the task ^[33][34][35][33,34,35]. The composition of a group greatly impacts the interactions between the group members and consequently students’ learning or task quality [33]. Group composition has been well-documented during the last few decades ^[26][33][26,33] and refers to the homogeneity or heterogeneity of a group in terms of, for example, gender [36] or competence level [26], or the degree to which group members are familiar with each other ^[37][38][37,38].

The process leading to group composition is called the group formation, which is a critical phase within collaborative learning [22]. When strategically employed, it can improve group performance ^[39][40][39,40].

Two main questions can be put forward with regard to the group formation process: (1) who is forming the groups and (2) what is considered when they are formed. Regarding the first question, a distinction can be made between teacher-assigned and student-selected groups. The term student-selected group formation is used interchangeably with self-selected group formation. Prior research is inconclusive regarding which of these two methods is best. Several studies have shown that students who are allowed to select their own collaboration partner(s) indicate that this method leads to a fairer group composition than being assigned to a group ^[41][42][41,42], as students perceive themselves as being more able than their teachers to select appropriate group members [41]. In particular, when teachers do not know their students very well, teacher-assigned group formation might not be the favorable group formation method, according to Lambić et al. [43].

Regarding the impact of student-selected versus teacher-assigned group formation on the collaborative process, students are more positive when collaborating with self-selected group partners as opposed to teacher-assigned group partners. More specifically, several studies have elicited that students experience more enjoyment, supportive behavior, and at-ease communications ^[23][40][44][23,40,44]. However, in other studies, students in self-selected groups realized that their partner, despite being a good friend, was not a good collaboration partner ^[41][45][41,45]. Meanwhile, whilst some scholars have indicated a greater extent of equal participation and a fairer task division in self-selected groups ^[23][40][23,40], others have established no differences [24], or contradictory findings, i.e., student-selection leading to lower levels of participation and more off-task talk ^[20][24][42][20,24,42].

Concerning group outcomes, some studies have shown higher quality writing tasks in teacher-assigned dyads in comparison to student-selected dyads ^[24][46][24,46]. In contrast, other researchers have concluded that student-selected groups outperform randomly assigned groups ^[47][48][47,48] and teacher-assigned groups that were based on self-reported ability [48]. Tsoi and Aubrey [44] found no general difference in language learning between teacher-assigned or student-selected group formation. However, when studying perceived performance, student-selected groups report higher scores which do not necessarily correspond to their actual grades [23]. Mitchell et al. [40] suggest that the appropriate group formation method depends on the task and that especially for tasks requiring much collaborative effort, self-selection might be eligible.

In summary, there are contradictory findings regarding the impact of student-selected or teacher-assigned group formation on both the collaborative process and outcomes. Some of these differences may be related to the way teachers or students actually form their groups [49], which brings us to the second question concerning group formation: what is considered when groups are formed?

The literature shows that teachers assign groups either at random [24] or purposefully homogeneously and/or heterogeneously based on student characteristics such as gender [36], ability ^[25][26][25,26], personality [50], and prior knowledge ^[6][27][6,27].

Regarding student-selected group formation, the research literature consistently points to group familiarity, and more in particular friendship, as the main determining motive for self-selecting a partner for collaboration ^{[24][40][41][42][44][48]}[24,40,41,42,44,48]. Group familiarity is defined as the extent to which students know each other prior to the collaborative task [38]. The more familiar members are with each other, the quicker they can advance to the core of the collaborative task, as they need less time for regulating their collaborative process ^[23][38][23,38]. Furthermore, group familiarity in general is positively related to teamwork satisfaction ^[23][37][23,37]. In addition, choosing someone familiar decreases uncertainty about the course of the collaborative process, and students tend to prioritize certainty and predictability for academic tasks [49].

Given that students collaborate to perform a particular learning task for which specific skills are needed [49], one could expect that they take their peer’s specific ability into consideration when selecting a partner for collaboration. Several studies indicate the benefits of grouping students based on ability ^[26][43][26,43]. Although ability is one of the main characteristics that university teachers take into account in teacher-assigned group formation, limited research focuses on whether students consider a partner’s ability when forming groups. Ideally, students should consider peers with complementary skills [49]. According to Chen and Gong [48], students heavily rely on friendship for group formation, thereby ignoring the specific abilities of their group members. Fischer et al. [47] analyzed the group formation behavior of 672 higher education students and observed that student-selected groups were significantly more homogeneous than randomly assigned groups in terms of ability, gender, and pro-sociality (i.e., voluntary behavior aimed at benefiting others) [51]. In their study, it is unclear, however, whether students deliberately chose a partner of the same ability.

Students tend to rely on someone’s willingness to contribute to the task. In particular, reputational information on a student’s ability, work ethic, or task approach is considered independent of their actual ability [49]. Objective information is most often unavailable; hence, they rely on either their own experience from collaborating with a particular group member or on information from someone else [49].

In general, as people tend to be attracted to others with similar characteristics, attitudes, or beliefs [52], it can be hypothesized that students, whether consciously or not, self-select peers who share the same characteristics, attitudes, or beliefs for academic collaboration. Student-selected group members quickly recognize similarities in their chosen partners and perceive these as positive. Teacher-assigned group members, on the other hand, rather see differences, regardless of being randomly or purposefully heterogeneously assigned [23]. Collaborating with students similar to oneself can result in easier communication and increases the predictability of others’ behavior and values [49]; hence, improving the predictability of the collaborative process. However, choosing similar students for group formation results in low-diversity-groups [49].

In summary, there is ample research on how university teachers form groups and which student characteristics they take into account. However, while there are studies available on the aspects on which students are focusing in view of selecting group members, detailed information on students’ reasoning to actually select a partner is lacking. This is particularly the case for collaborative writing tasks.