**5. Discussion**

This study explores the implications of trust self-enhancement (i.e., a group assigns more trust to oneself as compared to the trust assigned to the other stakeholder groups) for the evolution of perceived centrality (expected before the intergroup interactions and experienced during the intergroup interactions) in the MPSs. More exactly, we analyzed the impact of trust self-enhancement expectations on betweenness and closeness centrality indices as they evolved during the intergroup interactions in an MPS composed of six stakeholders dealing with a complex decision task.

As the results show, trust self-enhancement was a precursor of network centrality in the sense that trust self-enhancement at Time 1 (one's own group is perceived as more trustworthy as compared to the other stakeholders in the MPS) led to increased expectations regarding one's own betweenness and closeness centrality at Time 1 (prior to the interactions with other stakeholders). Then, these expectations sequentially increased the perceptions of experienced betweenness and closeness centrality at Times 2, 3, and 4.

Stakeholders that perceived themselves as being more trustworthy (i.e., more honest, reliable, and less likely to exploit the others), than the other stakeholders in the MPS, were motivated to seek and maintain a central position in the MPS network in order to maintain a high group distinctiveness and ultimately acquire more power. Central stakeholders could make useful contributions to the task at hand by sharing relevant information and by trying to integrate the differences stemming from the other stakeholder groups in the system that might not be otherwise connected through betweenness centrality (ultimately central actors may become more powerful this way). Moreover, expectations regarding one's own betweenness centrality in the collaboration network were further positively associated with perceptions regarding one's betweenness centrality observed in real inter-group interactions throughout the simulation. As previously argued, enacting such a bridging role within the MPS was an indicator of the stakeholder's power and influence within the system. Such a privileged position may have a beneficial impact for the success of the collaboration process, since, in order to achieve a comprehensive decision, the system is required to integrate the needs and concerns of all actors involved in the process (Schruijer 2006; Gray 2007; Curs,eu and Schruijer 2017) and a stakeholder with high betweenness centrality can facilitate this process. However, such a stakeholder may also act as a gatekeeper, blocking the information flow and thus impeding the collaboration process (Burt 1995). In MPSs involved in decision-making tasks, as was the case in our study, such central stakeholders acting as gatekeepers may have prevented consensus to emerge.

Similarly, trust self-enhancement was positively associated with closeness centrality in the collaboration network throughout the simulation. Viewing one's own group as more trustworthy led first to an anticipation of having a more central position in the system, which, in turn, materialized in experienced closeness. The emergence of closeness centrality was relevant for the collaboration process, as the stakeholder groups that relationally are on average closer to the other stakeholders in the system (i.e., they have a high level of closeness centrality) could more easily gather useful information, communicate interests throughout the network, and help with integrating the diverse points of view. On the other hand, high levels of closeness centrality may also impede the potential for fruitful collaboration, as such central stakeholders may push forward their (hidden) agenda, by suppressing diversity and prevent participation, especially from less powerful stakeholders in the system (Curs,eu and Schruijer 2017).

These results extend the expectations state theory (Berger et al. 1974; Berger et al. 1977) by showing that prior to engaging in intergroup interactions, self-enhanced perceptions of trust lead to expectations about a high centrality in the MPS network. These expectations are probably rooted in two mechanisms: self-esteem and self-defense. On the one hand, high self-trust mobilizes the stakeholder to actively engage in intergroup interactions, and on the other hand, rather low trust of others may stimulate the stakeholder to reach out to others in order to establish and maintain control over their actions in an MPS. Seeking and maintaining a high centrality in the MPS network is likely to be a mechanism that allows stakeholders to maintain their high distinctiveness as a group related to trust self-enhancement.

An important claim based on these results is that collaboration starts in the minds of those involved and it is later on shaped by the contextual dynamics and social complexities emerging in MPSs. This adds value to practice as well. The onset of interactions in MPSs seems to be crucial. Prior to any kind of interactions between the stakeholders, consultants and managers can guide the stakeholders to work with themselves and engage in reflective processes about the positive impact of their role in the system and empower them to create expectations about the contribution they can bring to the system.

Our findings also revealed an additional effect regarding the implications of trust self-enhancement for collaborative betweenness network centrality in interactions (not mediated by betweenness centrality expectations evaluated at Time 1). The indirect effect of trust self-enhancement via experienced collaboration betweenness at Time 2, and the sequentially at Time 3 and Time 4, was negative and significant (effect size −0.12, SE = 0.10, 95% CI [ −0.55, −0.01]). Through this mediation path that excluded expectations of betweenness, trust self-enhancement actually decreased betweenness after the onset of intergroup interactions (although this effect was not significant). It was therefore likely that this negative significant indirect effect was explained by the sequential positive association between collaboration betweenness at Times 2, 3, and 4. A potential explanation is that initial trust self-enhancement may lead to overconfidence concerning one's centrality position in the MPS network. When the interactions begin, the stakeholders with high trust self-enhancement may discover that the other stakeholders do not perceive them as being so central in the system. Overconfidence could therefore be an alternative mechanism that explains the workings of trust self-enhancement in MPS.

This emergen<sup>t</sup> result points towards a dual mechanism that connects trust self-enhancement with betweenness centrality. On the one hand, trust self-enhancement influences the expected network position and enhances one's centrality in the collaboration network. On the other hand, trust self-enhancement may generate overconfidence in one's position, that leads to a decrease in collaboration centrality after real between-group interactions start. We could fully explore the first mechanism through the sequential mediation analyses reported in Figures 1 and 2. However, we did not collect data on overconfidence that could elucidate the second mechanism likely to be involved here. Future research could further explore the co-existence of these two mechanisms linking trust self-enhancement expectations to centrality in the collaborative network.

To conclude, the main contribution of this study is the result concerning the sequential development of betweenness and closeness centrality, in line with the view of multiparty systems as dynamic entities (Curs,eu and Schruijer 2017). In such a system, stakeholders with sometimes very different concerns and agendas interact in the hope of reaching a common goal that cannot be envisioned from the start of the interaction. In order to do so, the stakeholders are expected to share their interests and concerns and use their expertise to work through disagreements. Often, however, they start their interactions based on initial assumptions about each other, which they subsequently test and (re)shape according to the information gathered during intergroup interactions. In turn, the emerging cognitive structures (social networks in minds) will further influence the network centrality of the stakeholders. The structure of the collaboration network is thus subject to constant change, in line with the within and between group dynamics.

Our paper also points towards the relevance of using a social network analysis in the exploration of MPSs dynamics. Modern analytical approaches allow the integration of various network perceptions in aggregated social networks that capture the relational landscape of MPSs. Next to the benefit of using these approaches in research, one could envisage dynamic social network visualization tools used by facilitators of MPSs. MPS stakeholders are often trapped in the social dynamics of these systems and process interventions that may rely on social network visualization tools, which are needed in order to help the system overcome the hurdles of conflict and relational tensions and optimize the collaborative efforts.

Next to the contributions, our study has limitations as well. We have used a behavioral simulation, with a specific decision task and our results may not generalize to other tasks, inter-organizational settings, or MPSs. In order to ensure generalizability, our results need to be replicated in other settings and using other evaluation methods as well. Each of the simulations contained six stakeholder groups; therefore, the size of the social network was rather small, a fact that could have restricted the variance in our centrality measures. Moreover, our network analytic approach aggregated individual relational perceptions (or expectations at Time 1) within groups, and in doing so, we obtained a more accurate image of network centrality. However, we cannot claim that our measure of network centrality was based on objective, measurement-error free data. Rather we claim that due to the aggregation method used, we rely on intersubjective aggregation as an indicator of collaboration. Future research could extend these network approaches in MPSs and use more objective network indicators. Finally, our results focused on collaboration network centrality and no definite claims can be derived about the success of such collaborative relations. High betweenness and closeness may eventually be detrimental for the relational dynamics in multiparty collaboration. Stakeholders with high betweenness may act strategically and display dysfunctional inter-group behavior by filtering and distorting the information shared among the other stakeholders. Moreover, MPSs in which closeness is very high may eventually display collusive dynamics with negative implications for the outcomes of such MPSs (Gray and Schruijer 2010). Future research could investigate more directly the association between the two centrality indicators and collaborative goal achievement in MPSs.
