**5. Discussion**

Our paper addressed the interrelatedness of multiparty dynamics at different system levels. We focused on the entwinement between expectations and social interactions and analysed positive and negative interdependence (the interplay between collaboration and conflict). Moreover, we tested bottom-up and top-down influences, that is, how within- stakeholder party dynamics affect those at the system level, and, how system level dynamics affect dynamics at the stakeholder level, in a multiparty context. Building on social interdependence theory (Deutsch 1949; Johnson 2003) we argued that positive interdependence (facilitating collaborativeness) at the system level increase the degrees of freedom for elemental components of the system (allowing stakeholder parties to engage with the task and thus engage in within-group task conflict), while negative interdependence (facilitating conflictuality) decreases these degrees of freedom and stimulate relationship conflict within stakeholder parties.

As top-down influences we have tested the effect of the changes in collaborativeness and conflictuality perceived at the system level on the changes in task and relationship conflict experienced within stakeholder parties. Regarding bottom-up influences, we have tested the extent to which changes in task and relationship conflicts experienced within the stakeholder parties co-evolve with changes in collaborativeness and conflictuality at the system level. To disentangle the top-down and bottom-up influences we have used different referential models for collaboration and conflict. Overall, bottom-up processes received stronger support than top-down processes. Changes in task conflict within stakeholder parties are positively related to changes in perceived collaborativeness at the system level (that is, perceived collaborativeness of all stakeholder parties combined) while changes in within-group relationship conflict are positively associated with system conflictuality. From the two top-down influences investigated, only the effect of change in conflictuality at the system level on the changes in relationship conflict experienced within stakeholder parties was significant. This finding next to the significant bottom-up association between relationship conflict and perceived conflictuality supports the pervasive nature of especially relationship conflict in social systems (Van Bunderen et al. 2017; Pluut and Cur¸seu 2013). Conflict contagion or the transfer of negative cathexis (Johnson and Johnson 2005) seems to be an influential process in multiparty systems. Social interdependence theory (Deutsch 1949; Johnson and Johnson 2005) also claims that positive cathexis could be transferred among the parties engaged in the system. This claim warrants further empirical exploration as positive cathexis is claimed to be beneficial for collaboration (Johnson and Johnson 2005).

The cognitive synergy path (conceptually described in Cur¸seu and Schruijer 2017) and the positive interplay between changes in within-party task conflict and system collaborativeness are only supported for the bottom-up test. Changes in task related debates, in this case engaging in task conflict, within the stakeholder parties, are proportionally related to changes in parties' perceived collaborativeness. In other words, the more stakeholder parties engage internally with the different opinions regarding addressing the task, the more they have the potential to contribute to the

overall collaboration in the system. Using the terms of social interdependence theory, task conflict seems to increase the perceived inducibility (openness to being influenced by others) of the parties. As inducibility increases, parties engage with the multiparty task at hand and promote the achievement of their collective goal (Johnson and Johnson 2005).

We call for more research integrating systems-psychodynamic theorizing (Schruijer 2016; Vansina and Vansina-Cobbaert 2008) and social interdependence theory (Johnson and Johnson 2005) as they both apply to multiparty systems. In social interdependence terms, for multiparty collaboration *substitutability* needs to be low (parties have unique competencies or resources that cannot be substituted by the other parties in the system), while *inducibility* is required (parties need to be open to social influence in the process of developing a collective goal) and *cathexis* is present (the transfer of positive and negative evaluations/emotions within and among parties). A key claim of social interdependence theory is that the structure of the individual goals in a given situation determines how participants interact. The structure of the stakeholder parties' goals in our simulation require collaboration or promotive interactions as parties cannot deal with the complex situation on their own. However, as our results show, parties often engage in conflictual (contrient) interactions and the emotional dynamics override the normative (cognitive and rational) need for collaboration. Exploring the (conscious and unconscious) emotional dynamics or the cathexis using a systems psychodynamic perspective could help us gain a deeper understanding of the dynamics of multiparty systems.

So far, we have offered a system interpretation of our results. In our simulation, different individuals engaged in interactions as representatives of the seven parties. They had to embark on these interactions without the restrictions of particular (individual) role prescriptions. As such, their interpersonal skills, personality and behavioural tendencies may have played an important role on how the interactions between group members unfolded. The aim of the simulation was not to attend to individual behaviours and no personal feedback was given to participants. Moreover, as the participants knew each other (they all took part in either a degree program lasting for 16 months or a professional development program), interpersonal histories may have been carried into the simulation and interpersonal conflicts from outside the simulation or interpersonal attraction for that matter, could have influenced the inter-group dynamics.

Our paper has several limitations. First, for the test of top down influences, data to test the models were collected from the same source for the mediator and dependent variables, therefore these analyses are susceptible to common source bias. However, given the fact that the evaluation of within-party conflict and the conflictuality and collaborativeness had different reference points and different rating intervals, we could argue that the common method bias concerns are attenuated. Second, in our study we only evaluated task and relationship conflict as experienced within the stakeholder parties and future research could also evaluate task and relationship conflict in the whole system and in this way test top-down and bottom-up influences in a different manner. Third, we have used aggregated individual scores to obtain evaluations of system collaborativeness and conflictuality. Future research could use global evaluations made by external observers for these two forms of social interdependence. Fourth, in terms of alternative interpretations, one could argue that the bottom-up influences as operationalized in our study actually reflect self-fulfilling prophecies (Merton 1948). Participants' expectations about positive and negative interdependence may lead them to behave in ways that are consistent with these expectations and as a consequence the way they are perceived by others in interactions is congruen<sup>t</sup> with their own initial expectations. Moreover, the way we have operationalized the top-down influences may conflate false consensus effects (Ross et al. 1977). In other words, participants' experiences of positive or negative interdependence in their own parties may lead them to evaluate other parties as displaying the same forms of interdependence. Although, based on the data collected in this study we cannot refute these two alternative explanations, we argue that the aggregation of individual evaluations may have alleviated the self-fulfilling prophecy and false consensus influences. Finally, given the time-intensive nature of the behavioural simulations, our sample is small and future research could attempt to replicate our findings in larger samples and in different settings.
