Introduction

The adage “Two heads are better than one” emphasizes the value of cooperative teamwork, often leading to superior performance compared to individual effort (Lu et al., 2010; Kistruck et al., 2016). However, this effect may not be robust with changing decision-making contexts and interpersonal relationships. For example, while cooperation can enhance intra-group cohesion, it may also lead to increased mutual dishonest activities, resulting in collaborative dishonesty (Leib et al., 2021). Additionally, team members may experience social loafing, where the perceived diffusion of responsibility leads to reduced individual effort (Karau and Williams, 1993; Pearsall et al., 2010). In contrast, competition has been shown to positively influence innovation and overall performance (Xie et al., 2023). Empirical studies suggest that competition within teams is associated with individuals’ achievement goals, motivating them to perform better (Kilduff et al., 2010; Heidemeier and Bittner, 2012). These findings indicate that cooperative and competitive goal structures can organize participants’ behavior in decision-making, but their respective impacts on performance in complex interpersonal contexts remain unclear.

Social interaction, including cooperation and competition, serves as the primary channel for humans to interact with society, depicting an interaction function with other individuals for information exchange, mutual minds and emotional states shaping, and even brain activity for differing intentions and goals (Astolfi et al., 2011; Hari and Kujala, 2009). Social interaction encompasses social exchange, decision-making, and prosocial behaviors through verbal or nonverbal communication, playing a crucial role in both personal daily life and society development (Wang et al., 2018). Liu and Pelowski (2014) proposed a three-dimensional model to clarify types of social interaction based on people’s interaction patterns and contents, including goal structure, task structure, and interaction structure. They notably distinguished goal structure into cooperative and competitive interactions. Specifically, goal structure engages people’s intentions and behaviors, either promoting goal attainment or hindering another’s goal achievement (Deutsch 1949). However, the influencing pathway cannot be explained simply by goal-directed engagement, as other potential factors may impact people’s performance under goals. For instance, team identification, psychological safety, emotion contagion, and interpersonal brain synchronization affect collective behaviors within teams (Edmondson, 1999; Czeszumski et al., 2020; Reinero et al., 2021). Furthermore, social interaction in a two-person context is highly affected by each individual’s personality (Hari and Kujala, 2009). However, these contextual factors in decision-making and mutual personal traits are overlooked by previous research on social interaction structures. Consequently, the underlying processes influencing this relationship remain largely unknown. This raises the need for further studies integrating dynamics of interaction, contextual factors, and psychological personality into a research framework to explain human social interacting behaviors under different goals.

While the social interaction perspective provides a foundational understanding, social interdependence theory offers a more dynamic framework for examining the interplay between contextual factors, interpersonal interactions, and psychological traits under different goal structures. According to social interdependence theory, the patterns of social interaction are shaped by the goal structure, which, in turn, influences individual engagement and decision-making. A cooperative goal promotes a promotive relationship, characterized by positive interdependence, whereas a competitive goal fosters an obstructive relationship, associated with negative interdependence. Additionally, in the absence of interdependence, individuals perceive no shared goal in social interaction, and their goal attainment becomes independent of others’ performance (Johnson, 2003; Hwong et al., 1993; Johnson et al., 1991). These studies provide a theoretical framework for understanding individuals’ behavioral responses and decision-making dynamics in social interactions under varying goal structures. Previous research has demonstrated the asymmetric effects of shifts in goal structure—from cooperative to competitive, and vice versa—on team decision-making performance, including accuracy and speed (Johnson et al., 2006). Specifically, a shift from cooperation to competition is defined as a “downhill shift,” while the reverse transition is referred to as an “uphill shift,” highlighting the differential impact of these shifts on performance (Johnson et al., 2006). These effects are further influenced by contextual factors such as outcome allocation, personality, and information certainty (Beersma et al., 2013; Gerpott et al., 2018). Thus, interpersonal dependence emerges as a critical feature of social interaction, shaped by the goal structure and significantly impacting task performance and decision-making (Courtright et al., 2015; Ring et al., 2022). Integrating social interdependence theory into this research provides valuable insights into how individuals within a team coordinate under cooperative versus competitive goal structures. Furthermore, individual traits and contextual factors within the team are associated with social interdependence and may serve as influential determinants in the relationship between goal structure and task performance. However, these factors remain underexplored and warrant further investigation.

Collectively, this study aims to investigate the effect of goal structure on task performance in two-person decision-making, with a particular focus on the roles of personal traits, trait homogeneity within teams, and how shifts in goal structure influence relationships. By integrating insights from both social interaction perspectives and social interdependence theory, we offer a comprehensive framework for understanding how goal structures affect decision-making in dyadic contexts. To address these research questions, we conducted three sub-studies. According to social interdependence theory, social interaction patterns are shaped by the presence of individual (non-common) goals, as well as common cooperative and competitive goals, with individuals behaving differently in each of these contexts (Johnson, 2003). In study 1, we proposed the following hypotheses:

Hypothesis 1a: Participants will coordinate differently under a common goal and no-common-goal conditions in two-person decision-making.

Hypothesis 1b: Under the motivation of a common goal, participants in a relatively inferior position tend to make riskier decisions in the later phase compared to their earlier decisions, whereas those in a relatively superior position show less of this shift.

In study 2, we adopted a two-person version of the Iowa gambling task (IGT) to investigate how participants’ decision-making propensities vary between uncertain and risky environments. Participants engaged with partners within pairs in either cooperative or competitive relationships while performing the task. Drawing from prior studies, we note that the IGT involves two phases: an initial ambiguous phase followed by a later risky phase (Buelow and Barnhart, 2017). Thus, we hypothesized the following:

Hypothesis 2a: Participants’ IGT scores will significantly differ between the early phase and the late phase, with a lower score in the early phase.

Hypothesis 2b: Participants will exhibit different behaviors under the two goal conditions, with higher scores in the cooperative goal condition compared to the competitive goal condition.

Study 3 was designed to investigate the effect of participants’ personality, social value orientation (SVO), SVO type homogeneity within a dyadic pair, and the shift in goal structure on the relationship between goal structure and task performance. Previous studies have shown that the SVO influences cooperative behavior and group cooperative climate within groups (Bogaert et al., 2008; Bogaert et al., 2012), and that individuals with similar SVO types influence individual-level cooperation in tasks such as gambling (Zhang et al., 2022c). Based on this, we hypothesized the following for Study 3:

Hypothesis 3a: Participants with different SVO types will exhibit varying sensitivities and decision-making tendencies in the context of cooperative and competitive goals, with pro-social individuals showing higher sensitivity in the cooperative goal context.

Hypothesis 3b: The homogeneity of SVO types within a dyadic pair and the shift in goal structure will moderate the relationship between goal structure and decision-making performance.

The current study focuses on choice behaviors in two-person decision-making contexts. We extend the concept of social interaction from hyperscanning and two-person neuroscience studies, a burgeoning area in cognitive neuroscience (Hari and Kujala, 2009; Liu and Pelowski, 2014). By integrating social interdependence theory, we introduce a new perspective into the study of goal-directed task performance. In two-person contexts, interdependence is greater than in individual-group relationships, making it more sensitive to dynamic shifts in goal structures and risk factors. We expect that behavioral methods will complement previous physiological imaging findings, providing ecologically valid evidence and supporting conclusions that reflect real-world decision-making processes.

Study 1

Study 1 aimed to investigate how common goals in social interaction influence on two-person decision-making. The experiment adopted a modified two-person Balloon Analogue Risk Task (BART), a widely applied dynamic sequential decision-making task (Lejuez et al., 2002). We used a 2 × 2 within-subjects design, with goal structure (common goal vs. no-common goal) and outcome feedback (positive vs. negative) as within-subjects factors.

Methods

Participants

A total of 86 college students (42 women, Mage = 21.55 ± 2.16) participated in the experiment (Table 1 provides the sociodemographic characteristics of three sub-studies). Power analysis indicated that this sample size would achieve 0.99 power, with the following parameters: an expected effect size of at least 0.25 (f), an alpha level of 0.05, a default within-subjects correlation of 0.5, and a non-sphericity correction (ƹ) of 1. Participants were right-handed with normal or corrected vision and reported no history of psychiatric disorders. They were randomly assigned into 43 dyadic pairs of the same gender, unacquainted with their partner. Before the formal experimental procedure, all participants gave their written informed consent and were informed of their right to withdraw from participation. Before the formal procedure, they were informed that they could obtain 20 Chinese yuan (about $2.74) for attendance and an additional bonus of up to 10 yuan (about $1.37) according to their task performance.

Table 1 The sociodemographic characteristics of three sub-studies.
Full size table

Experimental procedure

Before the formal procedure, participants had 20 practice trials to familiarize themselves with the task rule. During the formal BART paradigm, individuals were instructed to press the “1” continuously to pump up a virtual balloon displayed on the screen. The balloon expanded with each press, and the monetary reward in the temporary account increased accordingly. Specifically, the first pump won 1 cent (¥ 0.01), and each subsequent pump was twice the previous amount. Participants obtained up to ¥ 1.08 per balloon, after which a new balloon appeared. Participant received positive feedback if they stopped pumping and cashed out the current balloon’s reward by pressing “5” on the keyboard, adding to their cumulative earnings. However, the balloon could explode after any pump, resulting in a loss of money from the account. The exact probability of the explosion was unknown to participants. For each balloon, the probability of bursting after the first pump was set at 1/128. If the balloon did not explode, the probability increased incrementally with each pump: 1/127 after the second pump, 1/126 after the third, and so on, up to a maximum of 128 pumps, at which the probability of bursting reached 1/1 (See Fig. 1). The average breaking point for each balloon was 64 pumps, which was considered the optimal stopping point to obtain the maximum gain (Campbell et al., 2013). Based on previous studies (Lejuez et al., 2002; Zhang et al., 2022a, 2022b), we averaged the number of balloon trials except trials that balloons exploded (the adjusted values). These results, referred to as BART scores, are considered a reliable measure of risk-taking behaviors.

Fig. 1
figure 1

Experimental procedures of the BART in study 1.

Full size image

In the no-common-goal condition, participants were informed they had to complete the balloon task independently to earn more scores, which were proportionally converted to additional winnings. They could only view their live scores during the procedure, and the final scores of both members of a pair \({T}_{1}\) and \({T}_{2}\), would be revealed to each other. The total reward was the sum of both people’s scores (\(S={T}_{1}+{T}_{2}\)).

However, in the common goal condition, participants worked in the same pair and were informed their common goal was to achieve a total winning (\({T}_{{otal}}={T}_{1}^{{\prime} }+{T}_{2}^{{\prime} }\)) exceeding the sum (\(S\)) obtained in the no-common goal condition. Specifically, if \({T}_{{otal}}\ge S\), the pair achieved the common goal and received their earnings individual earnings. However, if \({T}_{{otal}} < S\), the pair failed to achieve the common goal, and each participant received half of the total winnings (\(\,{T}_{{otal}}/2\)) in the current condition. For example, Participant A earned 60 yuan (\({T}_{1}\)) and Participant B earned 40 yuan (\({T}_{2}\)), for a total reward \({of}\)100 yuan (\(S={T}_{1}+{T}_{2}\)) in the no-common goal condition; The common goal of A and B was to earn a total reward not <100 yuan in the common goal condition. Their individual winning, total reward, and whether they reached the common goal would be revealed to both participants by the experimenter. The rules of winning and losing (balloon bursting) on BART were consistent across conditions. Each participant completed 30 balloon tasks in each condition, lasting 10–15 min.

Given that studies, such as those by Xu et al. (2013), have demonstrated the BART to be highly stable and minimally affected by learning effects, the no-common goal condition was presented to participants first, followed by the common goal condition, to minimize the potential influence of the common goal condition on the no-common goal condition. After both rounds of tasks, all participants rated two subjective descriptions to evaluate the manipulation effect of the decision-making context. Specifically, they assessed the degree the following descriptions fit them on a 7-point Likert scale (1 = “Least fits myself”; 7 = “Most fits myself”), including “In the no-common goal condition, I pumped up balloons to achieve the goal and obtain more gains” (Description 1), and “In the common goal condition, I pumped up balloons to achieve the goal and obtain more gains” (Description 2).

Results

Manipulation examination

A total of 84 valid participants were asked to evaluate two subjective descriptions to examine the manipulation effect of the common goal context. Paired-sample t-test analysis showed participants rated lower scores on Description 1 (M = 5.18, SE = 0.30) compared to Description 2 (M = 6.19, SE = 0.22, t(83) = 2.35, p = 0.027, Cohen’s d = 0.68). The rating results suggested that participants’ goals revealed a significant difference between the common goal and no-common goal conditions, and they had more subjective motivation for higher rewards under common goals in a two-person decision-making context. This reflected the guiding influence of shared goals on decision-making under uncertainty. In addition, the result revealed the valid manipulation effect of a common goal.

BART scores

In this study, the average number of pumps for unexploded balloons was calculated as the BART performance score under two conditions (see Table 2). Previous studies show the BART score serves as a promising indicator for measuring the degree of risk-taking behaviors in the real world (Lejuez et al., 2002; Zhang et al., 2022a, 2022b). A higher BART score is strongly associated with optimal decision-making, characterized by choices closer to the optimal stopping point, and with higher monetary rewards, reflecting better task performance. (Steiner and Frey, 2021; Campbell et al., 2013). Paired-sample t-test analysis revealed a significant difference in BART score under two conditions (See Fig. 2). Specifically, the BART score in the no-common goal condition (M = 44.81, SE = 2.16) was lower than that in the common goal condition (M = 49.64, SE = 1.78, t(83) = −2.69, p = 0.010). The results showed that participants with a common goal tended to show better performance in decision-making compared to individual decision-making contexts without a common goal.

Table 2 Summary of the main results from three-sub studies.
Full size table
Fig. 2: Results of study1.
figure 2

A The BART scores in the no-common goal condition and common goal condition; B The BART scores under two conditions for participants in a relatively inferior position; C The BART scores under two conditions for participants in a relatively superior position. *p < 0.050, **p < 0.010.

Full size image

In addition, we conducted further exploratory analysis to investigate how participants adjust their choice preferences after the appearance of a common goal in a dynamic two-person decision-making context. Specifically, we categorized participants’ relative position within a dyadic pair based on their BART performance scores in the no-common goal condition: if a participant’s score \({T}_{1}\) was less than their partner’s score \({T}_{2}\) (\({T}_{1} < {T}_{2})\), the participant was considered to be in a relatively inferior position, making them more likely to fail in achieving the shared goal if they maintained the same decision-making tendency in the common goal condition. Conversely,if \({T}_{1}\ge {T}_{2}\), the participant was in a relatively advantageous position, it increased the likelihood of achieving the common goal and earning higher rewards.

The paired-sample t-test analysis revealed a significant difference in BART score for participants in different relative positions under two conditions. For participants who received a lower score and obtained a relatively inferior position in the no-common goal condition, the BART score in the no-common goal condition (M = 36.00, SE = 2.16) was lower than that in the common goal condition (M = 46.33, SE = 2.27, t(41) = −3.62, p = 0.002). The results showed that participants in the relatively inferior position tended to make riskier decisions (i.e., trying to pump the balloon more times) in the later session for goal accomplishment and higher reward, and choice propensity was adjusted by the participant’s common goal, reflecting a modulation effect of common goal context in two-person decision-making. On the other hand, for participants who received a higher score and obtained a relatively superior position, we observed no significant difference between the BART score in the no-common goal condition (M = 52.95, SE = 2.59) and that in the common goal condition (M = 53.62, SE = 2.60, t(41) = 0.47, p = 0.646).

Discussion

In Study 1, we employed the Balloon Analogue Risk Task (BART), a dynamic sequential decision-making task, to examine choice preferences in the presence or absence of a shared goal under uncertainty. The results revealed higher BART scores in the common goal condition compared to the individual decision-making condition, where participants approached the optimal stopping point more closer in the former. Specifically, compared to participants in the individual, no-common-goal context, those in the common goal condition adjusted their choice propensities, demonstrating the influence of a shared goal within the dyadic pair. The findings were consistent with prior studies, which found that people who make decisions within a common goal tend to outperform those working under individual, no-common-goal conditions. This effect has been observed not only in group settings but also in situations involving the presence of unfamiliar individuals or cues suggesting the involvement of others (Cottrell et al., 1968; Polman and Wu, 2020; Ladley et al., 2015).

Furthermore, we classified participants based on their relative positions within the dyadic pair, following the BART performance scores in the no-common-goal condition. The results indicated that participants in the relatively inferior position were more likely to make riskier decisions in the later phase of the task in pursuit for higher rewards. This suggests a modulation effect of the shared goal on decision-making within the dyadic context. Consistent with previous research using the BART paradigm, the common goal facilitated more coordinate decision-making in dyads compared to the individual, no-common goal condition (Wang et al., 2019). Additionally, the common goal appeared to motivate participants to adjust their choice behaviors after evaluating prior outcomes, highlighting the significant role of goal structure in shaping decision-making dynamics. In sum, Study 1 demonstrates that a shared goal in a dyadic context enhances decision-making coordination and encourages strategic adjustments, underscoring the impact of goal structures on sequential decision-making under uncertainty.

Study 2

Based on findings from Study 1 using BART, the aim of Study 2 was to investigate the effect of goal structure in social interaction on two-person decision-making. Specifically, the goal structure included cooperative and competitive goals. Participants interacted with partners in a dyadic pair and completed their tasks in a cooperative and competitive relationship.

Study 2 utilized the Iowa gambling task (IGT), a well-established paradigm for measuring choice behavior related to real-world risk behavior (Cui et al., 2013; Bechara et al., 2005). Previous studies showed the IGT and BART could assess different psychological properties of decision-making and reflect nonoverlapping processes (Buelow et al., 2015; Upton et al., 2011). Specifically, the IGT was regarded as an effective assessment for evaluating how participants make decisions during changing contexts, from decision-making under ambiguity to decision-making under risk (Brand et al., 2007; Buelow and Barnhart, 2017). Therefore, we adopted the IGT to examine the effect of cooperative and competitive goals on two-person decision-making during the transition from ambiguous to risky contexts.

Methods

Participants

A total of 50 college students (26 women, Mage = 20.34 ± 3.51) participated in Study 2. A power analysis indicated that a sample size of 50 participants would provide 0.99 power, with the following parameters: an expected effect size of at least 0.25 (f), an alpha level of 0.05, a default within-subjects correlation of 0.5, and a non-sphericity correction (ƹ) of 1. Participants were right-handed with normal or corrected vision and reported no history of psychiatric disorders. They were randomly assigned to 23 dyadic pairs of the same gender, where each participant was unacquainted with his partner. 2 pairs (4 participants) were excluded due to technical problems. Before the formal procedure, they were informed they could obtain 20 Chinese yuan (about $2.74) for attendance and an additional bonus of up to 10 yuan (about $1.37) based on task performance.

Experiment procedure

Initially, participants learned they would work together with a partner in dyads and had 3 min to read experiment instructions. During the formal task, four card desks (A, B, C, D) were presented on the screen. Participants chose one card per trial, with four decks presented each time. Each card carried different money amounts. Specifically, desks A and B offered immediate large gains but also resulted in large penalties after ten or more choices, which were two disadvantageous desks. Participants could get a 100-yuan immediate reward from decks A and B, but lose 1250 yuan (large punishment) after ten turns, for a 250-yuan net loss. Conversely, desks C and D offered immediate small winnings and small penalties, yielding small gains over ten or more choices (advantageous) (Manes et al., 2002; Brand et al., 2007). At the beginning of each block, the positions of the cards were not fixed, preventing participants from obtaining information about the locations of the advantageous and disadvantageous decks from the previous task round. Participants received a 50-yuan immediate small reward from C and D, incurring a 1250-yuan (large punishment) after ten turns, for a 250-yuan net win. Participants were instructed to obtain as many winnings as possible by selecting one card once from the four decks until the program stopped them, determining the actual performance bonus.

Participants had two conditions of experimental task to accomplish: the cooperative goal condition and the competitive goal condition. In the cooperative goal condition, participants were instructed to select cards and accrue more winnings with their partners. Participants in each pair completed the task individually, accumulating higher rewards in their temporary accounts. They finally earned the mean value of both temporary accounts. Accordingly, participants’ monetary earnings depended on shared rewards within a pair. They were informed to strive to increase common benefits, relating to actual bonuses. In the cooperative goal condition, each participant had 100 trials, lasting about 20 min. In the end, the pair was informed of personal accounts and common rewards by the experimenter.

In contrast, in the competitive goal condition, participants were informed that they were engaged in a competitive relationship with a partner within a pair. This session aimed to attain higher winnings by selecting cards so that they could outperform their competitor. The regulation of reward allocation was consistent with the zero-sum rule (winner-takes-all), and the winner of the two-person pair would take away the total of both participants’ winnings, leaving the other partner with nothing in this condition. Therefore, participants were told that the obtained earnings in their accounts were associated with their task bonus, motivating them to earn higher rewards to defeat the opponent. Similarly, in the competitive goal condition, each participant had 100 trials to accomplish, which lasted about 20 min. At the end of the competitive session, the two participants of a pair were informed personal account and the winner’s earnings by the experimenter. Before the formal task, individuals underwent 20 practice trials to become familiar with the task rules.

Goal structures served as the within-subject variable, comprising cooperative goal conditions and competitive goal conditions, with the order of goal treatment counterbalanced in a formal procedure. Half of all pairs finished the task with a cooperative goal condition first and then a competitive task, while the other half of pairs vice versa. The entire task lasted 40 to 45 min, and they had no chance to communicate with their partner in any manner until the end of the procedure. Participants had no time limit, but they were also instructed to select and flip one card from the four desks as soon as possible.

Results

Consistent with previous studies, we calculated the net score as an indicator of IGT performance:

$${\rm{N}}{\rm{et\; score}}={Choice\; numbers}\left(C+D\right)-{Choice\; numbers}(A+B)$$

Specifically, we subtracted choices from the advantageous decks (C + D) from those in the disadvantageous decks (A + B) (Marquez-Ramos et al., 2023; Bechara et al., 2005).

A 2 × 2 repeated-measures ANOVAs was conducted with goal structure (cooperative vs. competitive) and decision-making stage (ambiguity vs. risk decision-making stage) in SPSS 21.0 (SPSS Inc., Chicago, IL). According to previous studies, we calculated the mean net score in each condition as the dependent variable (Marquez-Ramos et al., 2023), and goal structure (cooperative vs. competitive), decision-making stage (ambiguity vs. risk decision-making stage) as independent variables. Specifically, the ambiguity stage consisted of 40 trials (Trials 1–40), while the final 60 trials (Trials 41–100) were considered the risk decision-making stage (Brand et al., 2007; Buelow and Barnhart, 2017). The ANOVA results showed a significant main effect of the decision-making stage (F(1, 44) = 6.12, p = 0.017, \({\eta }_{p}^{2}\) = 0.120), with a higher mean net score in the stage of risky decision-making than that in the stage of decision-making under ambiguity. However, we observed no significant main effect of goal structure (F(1, 44) = 1.20, på 0.050). More importantly, we observed a significant interaction effect between the decision-making stage and goal structure (F(1, 44) = 10.87, p = 0.002, \({\eta }_{p}^{2}\) = 0.195) (see Table 2). The follow-up contrasts showed that during the stage of decision-making under ambiguity, there was no significant difference in the mean net score of participants between cooperative goal (M = 8.74, SE = 1.63) and competitive goal (M = 11.67, SE = 2.20, F(1, 44) = 1.16, på 0.050). However, for the stage of decision-making under risk, the participant’s mean net score under the cooperative goal (M = 20.35, SE = 3.29) was higher than that under the competitive goal (M = 9.74, SE = 4.20, F(1, 44) = 4.37, p = 0.042, \({\eta }_{p}^{2}\) = 0.088). The results indicated that compared with the early phase of decision-making under ambiguity, participants with competitive goals were more prone to choose risky options in pursuit of larger rewards during the latter stage of the IGT procedure.

Discussion

Building on the findings from Study 1, Study 2 utilized the Iowa gambling task (IGT), in which dyads made decisions under competitive and cooperative goal structures, respectively. The results revealed that, compared to the early stage, participants in the cooperative goal structure exhibited greater risk aversion in the late stage of decision-making. However, this effect was not observed in the competitive goal structure, partly supporting Hypothesis 2a. Prior research suggests that the two phases of IGT capture a shift in choice propensity from an ambiguous to a risky context (Buelow and Barnhart, 2017). In the later stage, participants were more likely to make riskier decisions based on rational analysis for larger rewards, having gained more experience from earlier trials and accumulated sufficient information about the consequences of their choices. This involved investing greater cognitive resources and utilizing working memory to track choice and outcomes, core components of executive functions, particularly cold executive functions (Brand et al., 2007; Colautti et al., 2022; Marquez-Ramos et al., 2023). However, no significant effects were observed in the early, ambiguous phase, highlighting a distinction between the two phases. In the early stage, the rules of winning and losing were not clearly defined, which meant that participants lacked sufficient information to estimate long-term gains or probabilities. As a result, they relied on intuitive decision-making and reversal learning (Kim et al., 2011). This is in line with prior studies (Brand et al., 2008; Kim et al., 2011).

Additionally, participants’ IGT net scores differed significantly between the two goals, with higher scores in the cooperative goal condition, which supports hypothesis 2b. Building on prior findings, Study 2 extended the understanding of two-person decision-making into cooperative and competitive contexts, which lead to promotive and obstructive relationships, respectively. According to social interdependence theory, cooperative and competitive goal structures foster positive and negative interdependence, respectively, which distinctly influence decision-making (Johnson, 2003; Hwong et al., 1993). Extending the social interdependence model into a two-person decision-making, Study 2 findings indicated that participants reacted differently to advantageous and disadvantageous alternatives depending on the goal structure. Specifically, participants in competitive goals were more likely to choose risky options for larger rewards in the risky stage, emphasizing the crucial role of competitive goal context in motivating individuals to outperform their opponents by leveraging environmental risk learning (Schulze and Newell, 2015). In summary, Study 2 highlights the nuanced impact of goal structures on decision-making dynamics in two-person settings, revealing how competitive and cooperative goals shape risk-taking behavior and decision outcomes.

Study 3

The first two sub-studies found people reacted differently under different common goals in two-person social interactions. Study 3 aimed to explore the influence of participants’ personality, social value orientation (SVO), SVO type homogeneity within a two-person pair, and the shift of structure on the relationship between goal structure and task performance. Study 3 maintained an experimental design similar to Study 2, but differed in several modifications. While the IGT paradigm captures two relatively stable stages, ambiguous and risky blocks, the BART paradigm better simulates decision-making in real life. This is because the BART paradigm allows people to exploit uncertain environments dynamically and sequentially, enhancing external validity (Li et al., 2020). In addition, we focused on personality traits and structural characteristics in two-person-making interaction, such as trait homogeneity within a two-person pair and the direction of goal structure shifts, rarely investigated in the existing literature.

Methods

Participants

A total of 64 college students (32 females, Mage = 22.71 ± 2.27 years) participated in the experiment. All participants were right-handed, had normal or corrected-to-normal vision, and reported no history of psychiatric disorders. Prior to the experiment, participants completed the Social Value Orientation (SVO) Slider Measure (Murphy et al., 2011). This sample size (n = 64) provided a statistical power of 0.99, based on an expected effect size of 0.25 (f), an alpha level of 0.05, a within-subjects measurement correlation of 0.5, and a non-sphericity correlation (ƹ) of 1. The 6-item slider measure presented a resource allocation scenario where participants had a chance to distribute a certain amount of joint payoffs (self/other allocation). Participants selected the most preferred allocative plan by dragging a point along the distribution line that depicted nine allocative plans for each item. We adopted a widely used Chinese-translated version of the slider measure, with a test-retest stability of 0.76 and a convergent validity of 0.57 (Qi et al., 2020). The SVO Slider angles (\({{\rm{SVO}}}^{^\circ }\)) were then computed as the single index of the participant’s response. According to previous studies, participants with high SVO scores (SVO° > 22.45°) are categorized as prosocial decision-makers, whereas those with low SVO scores (SVO° < 22.45°) are categorized as proself decision-makers (Zhang et al., 2022c; Qi et al., 2020). Participants were randomly assigned into 32 dyadic pairs with the same gender, and each participant within a pair was unacquainted with their partner. One pair (2 participants) was excluded because they failed to accomplish the entire formal procedure.

In addition, the present study aimed to investigate whether a participant’s choice propensity under goal structures was impacted by the congruence of their partner’s SVO type within a pair. Therefore, according to the homogeneity of SVO types within each dyadic pair, we classified dyadic pairs into homogeneous pairs (both participants were either prosocial or proself decision-makers), and heterogeneous pairs (consisting of one prosocial and one proself decision-maker). The homogeneity of SVO types within a two-person pair served as a between-subjects variable in the subsequent exploratory statistical analysis. Participants were unaware of their SVO types or the homogeneity of pairs before the formal experimental procedure. Before the formal procedure, they were also informed that they could obtain 20 Chinese yuan (about $2.74) for attendance and an additional bonus of up to 10 yuan (about $1.37) according to their task performance.

Experiment procedure

The experimental paradigm in Study 3 was identical to that of Study 1, employing the Balloon Analogue Risk Task (BART). Participants were required to press the button to inflate a virtual balloon in each trial for a higher reward (positive feedback). However, the balloon could explode after pumping and lead to monetary loss from the temporary bank (negative feedback). The exact probability of balloon explosion remained unknown to the participants. Consistent with the paradigm in Study 1, the probability of balloon explosion in each trial after the first pump was 1/128, and became 1/127 after the next pump, until reaching the maximum 128th pump.

Furthermore, there were many changes in the experimental manipulation and reward allocation rules. Goal structures, including a cooperative goal and competitive goal, acted as the within-subject variable, with the order of manipulation counter-balanced in a formal procedure. Half of the pairs completed the task with a cooperative goal first and then a competitive task, while the other half of the pairs did the opposite.

Participants completed two experimental task conditions: the cooperative goal condition and the competitive goal condition. In the cooperative goal condition, participants were informed that the objective was to inflate the balloon and maximize rewards through collaboration with their partner. Participants in each pair finished the balloon task individually, accruing higher rewards in their temporary accounts, and they received the mean value of the two temporary accounts at last. Therefore, the collective rewards of a pair determined participants’ personal monetary earnings, and they were asked to strive to increase common benefits, which directly impacted their actual earnings. In the cooperative goal condition, each participant had 30 balloon tasks to complete, lasting 10 to 15 min. At the end of the session, the two participants of a pair were informed of personal accounts and collective rewards by the experimenter.

In contrast, for the competitive goal condition, participants were told they were competing with their partner in a pair, and this session aimed to inflate the balloon and earn more rewards to surpass the competitor. Reward allocation followed the zero-sum rule (winner-takes-all), and the winner of the pair would obtain both participants’ rewards, leaving the loser with nothing. Participants were told rewards accrued in their account related to actual earnings, motivating them to inflate the balloon for higher rewards to defeat their opponent. In this condition, each participant had 30 balloon tasks, lasting 10–15 min. At the end of the competitive task, the two participants of a pair were informed of their personal accounts and the winner’s earnings by the experimenter.

For both conditions, after each trial, participants could observe the outcome and real-time variation in their temporary account, but had no chance to communicate with their partner. They had no time limit, but they were also told to make decisions promptly.

Results

Results of the SVO slider measure

We calculated the SVO Slider angles (\({{\rm{SVO}}}^{^\circ }\)) as an indicator of participants’ social value orientation based on previous studies. From participants’ responses to six primary items of the SVO Slider Measure, we classified 30 participants with higher SVO scores (SVO° > 22.45°) as pro-social decision-makers, whereas 32 participants with low SVO scores (SVO° < 22.45°) as proself decision-makers.An independent-samples t-test showed pro-socials’ SVO Slider angles (SVO°) (M = 30.67, SE = 2.02) were significantly higher than pro-socials’ \({{\rm{SVO}}}^{^\circ }\)(M = 4.71, SE = 1.92, t(60) = 9.31, p < 0.001). The results indicated proself participants and pro-socials differed significantly in choice propensities during resource allocation within a team. It also revealed a valid classification of SVO type based on participants’ SVO scores.

BART scores

A 2 × 2 repeated-measures ANOVA was conducted with social value orientation (pro-self vs. pro-social) and goal structure (cooperative vs. competitive) in SPSS 21.0 (SPSS Inc., Chicago, IL). Similar to Study 1, the average number of pumps for unexploded balloons was used as the BART score to measure task performance in the two conditions. The ANOVA results showed a significant interaction effect between social value orientation and goal structure (F(1, 60) = 5.42, p = 0.023, \({\eta }_{p}^{2}\) = 0.084). We observed no significant main effect of goal structure and social value orientation, all på 0.050 (see Table 2). The follow-up contrasts showed that for pro-social pairs, the participant’s BART score was higher under cooperative goal condition (M = 52.22, SE = 3.11) than under competitive goal condition (M = 45.06, SE = 3.10, F(1, 60) = 5.58, p = 0.021, \({\eta }_{p}^{2}\) = 0.086). However, there was no significant difference between the cooperative goal condition (M = 44.28, SE = 3.27) and the competitive goal condition (M = 47.35, SE = 3.26) for pro-self pairs, på 0.050.

Exploratory analysis

We classified two-person pairs into homogeneous and heterogeneous pairs according to their homogeneity of SVO types within a pair. We conducted an exploratory statistical analysis to investigate whether the participant’s choice propensity under competitive and cooperative goal structures was modulated by the homogeneity of SVO types. A 2 × 2 repeated-measures ANOVA was conducted with homogeneity of SVO types within a pair (homogeneous pair vs. heterogeneous pair) and goal structure (cooperative vs. competitive) in SPSS 21.0 (SPSS Inc., Chicago, IL). Homogeneity of SVO types within a two-person pair served as a between-subjects variable, and goal structure served as a within-subject variable, with BART score as the dependent variable.

The ANOVA results showed a significant interaction effect between the homogeneity of SVO types within a pair and the goal structure (F(1, 60) = 5.15, p = 0.027, \({\eta }_{p}^{2}\) = 0.080). We observed no significant main effect of goal structure and social value orientation, all på 0.05. The follow-up contrasts showed that for homogeneous pairs, the participant’s BART score was higher in the cooperative goal condition (M = 50.83, SE = 3.27) than in the competitive goal condition (M = 43.47, SE = 3.17, F(1, 60) = 5.52, p = 0.022, \({\eta }_{p}^{2}\) = 0.086). However, there was no significant difference between the cooperative goal condition (M = 46.13, SE = 3.21) and the competitive goal condition (M = 48.74, SE = 3.12) for the heterogeneous pair, på 0.050.

Additionally, we examined the impact of the transfer direction of goal structure on task performance under cooperative and competitive goals. The transfer direction of goal structure served as a between-subjects variable, comprising participants who completed blocks in the order of cooperative goal to competitive goal (Direction 1) and the reverse order from competitive goal to cooperative goal (Direction 2). A 2 × 2 repeated-measures ANOVA was conducted with the transfer direction of the goal structure (Direction 1 vs. Direction 2) and goal structure (cooperative vs. competitive) as factors.

The ANOVA results revealed a significant main effect of goal structure (F(1, 60) = 4.48, p = 0.039, \({\eta }_{p}^{2}\) = 0.071). We also observed a significant interaction effect between the transfer direction of the goal structure and the goal structure (F(1, 60) = 17.01, p < 0.001, \({\eta }_{p}^{2}\) = 0.224). No other significant effect, på 0.050. Follow-up contrasts showed that for Direction 2 pairs, the participant’s BART score was higher in the cooperative goal condition (M = 55.35, SE = 3.59) than in the competitive goal condition (M = 42.335, SE = 3.59, F(1, 60) = 15.662, p < 0.001, \({\eta }_{p}^{2}\) = 0.209). However, no significant difference between the cooperative goal condition (M = 44.26, SE = 3.21) and the competitive goal condition (M = 48.45, SE = 2.81) for Direction 1 pairs, på 0.050.

Regarding Hypothesis 3b, a moderation effect model was tested using PROCESS V3.3. The goal structure was designed as the independent variable, and the homogeneity of SVO types within a pair was set as the moderating variable. Both were dichotomous variables, and BART scores were continuous dependent variables. The statistical significance of the hypothesized moderation effect was determined by the 95% confidence interval (CI) with 5000 bootstrapping resamples, a standard parameter setting consistent with our previous study (Zhao et al., 2023). The results showed that the moderating model was established. Specifically, the effect of goal structure on task performance was significantly moderated for homogeneous dyads (β = 1.97, SE = 6.38, t = 1.56, p = 0.01, 95% CI = [4.74, 22.62]). Similarly, we also set the transfer direction of the goal structure as the moderating variable, which was a dichotomous variable. The statistical significance of the hypothesized moderation effect was determined by the 95% confidence interval (CI) with 5000 bootstrapping resamples. The results revealed that the moderating model was established. Specifically, the effect of goal structure on task performance was significantly moderated when decision-making was from a competitive goal to a cooperative goal (Direction 2) (β = 1.30, SE = 5.09, t = 2.55, p = 0.010, 95% CI = [2.92, 23.08]). These results showed that characteristics of goal structure in social interaction, such as within-dyad homogeneity of their social value orientations, and transfer direction of goal structure, moderated the relationship between the goal structure and task performance.

Discussion

The aim of Study 3 was to investigate the role of participants’ social value orientation (SVO), the homogeneity of SVO within a two-person dyad, and the direction of shift between goal structures in a dynamic decision-making environment. To achieve this, we used the modified two-person version of the BART paradigm to investigate the psychological mechanism of two-person decision-making in different goal structures. Specifically, we examined the effect of SVO homogeneity within the dyad and the direction of the shift in goal structure. In sum, the results revealed that prosocial dyads exhibited higher BART scores in cooperative goal conditions compared to competitive conditions. However, this effect was not observed in proself dyads, indicating an interaction between SVO homogeneity and goal structure that supported Hypothesis 3a. Previous research suggests that SVO types—such as individualist, competitor, and cooperator—serve as indicators of personality traits and influence individual and group decision-making preferences within a social context (Bogaert et al., 2008; Bogaert et al., 2012; Qi et al., 2018). Our findings provide empirical evidence for these perspectives and extend the implications of SVO measurement to sequential dynamic decision-making contexts. Additionally, two exploratory analyses were conducted to test Hypothesis 3b, focusing on the mechanism of structural characteristics in two-person interaction, including trait homogeneity and the direction of goal structure shift. These analyses revealed that the homogeneity of social value orientations within the dyad, as well as the direction of the shift in goal structure, moderated the relationship between the goal structure and task performance.

General discussion

The effects of goal structure on two-person decision-making

This study investigated the effect of a cooperative versus competitive goal structure on task performance in two-person decision-making. From the social interdependence perspective, cooperative and competitive goals elicit promotive and obstructive interdependence patterns, respectively, which in turn influence subsequent choice preferences. Through three sub-studies, we explored whether cooperative or competitive goals lead to more effective decision-making patterns when individuals interact with a partner in uncertain and dynamic situations. In study 1, we examined the impact of a common goal on two-person decision-making, finding that individuals under a shared goal outperformed those with an individual goal. Building on this, Study 2 examined how participants in dyads responded to decision-making tasks under competitive and cooperative goal structures. The results revealed that individuals were more likely to avoid risk in the risky phase compared with the ambiguous phase under cooperative conditions, while no such effect was observed under competitive conditions, suggesting an asymmetric impact of goal on choice preference. Study 3 extended the research by exploring the roles of participants’ social value orientation (SVO), SVO-type homogeneity, and the dynamic shifts in goal structure on task performance. The results showed that prosocial dyads achieved higher BART scores in the cooperative goal condition compared to the competitive goal condition. However, no such effect was observed for proself dyads, highlighting an interaction between SVO-type homogeneity and goal structure. Collectively, these studies demonstrate that task performance in decision-making is influenced not only by the goal structure but also by the personality homogeneity within dyads and the dynamic characteristics of goal shifts. Specifically, building on findings from Zhang et al. (2022c), we found that similarity in SVO influenced individual-level cooperation. We categorized SVO homogeneity within dyads as either homogeneous or heterogeneous, revealing a significant impact on task performance at the dyadic level. This suggests that the personality configuration within a dyad plays a crucial role in two-person and team decision-making processes. The finding was consistent with our neuroimaging study demonstrating SVO homogeneity within a dyad influenced subsequent cooperation in team decision-making (Zhao et al., 2023). Specifically, the neuroimaging results revealed that the SVO homogeneity moderates the effect of interpersonal relationships on decision-making propensity via inter-brain synchronization (IBS) in the prefrontal cortex (PFC). fNIRS-based hyperscanning results partially supported this hypothesis, showing that SVO homogeneity within the dyad affected subsequent cooperative performance. In sum, the present study provides behavioral evidence and extends the experimental framework for examining decision-making within a cooperative-competitive goal structure, shedding light on the complex interactions between personality, goal structure, and task performance in two-person decision-making contexts.

Our findings make substantial contributions to the understanding of two-person choice behaviors by examining the dynamic variations in social interdependence and uncertainty within a risky environment. First, the modulation effects on task performance revealed asymmetric choice preference between cooperative and competitive goal structures, alongside the asymmetric influence of within-dyad personality homogeneity and shifts in goal structure direction. These findings provide robust support for social interdependence and structural contingency theories (Johnson et al., 2006). Unlike static contexts, the contingency theory extends social interdependence theory by suggesting that decision-makers must adapt to changing interaction structures to effective effective performance in coordination and team behaviors (Beersma et al., 2002; Johnson et al., 2006). This includes shifts between cooperative and competitive goals and the direction of these transitions. Our study extended this perspective and introduced a new dimension—personality trait and trait homogeneity within a dyad—into the social interdependence relationship and the structural contingency model. Second, we further explored dynamic changes in decision-making strategies, particularly with respect to relative position shifts within a two-person team. In Study 1, we classified participants’ relative positions within a dyadic based on their performance scores on the BART under the no-common-goal condition. We then examined whether participants would adjust their risky propensity in subsequent trials when placed in a relatively inferior or superior position. The results showed that participants in inferior positions tended to take riskier decisions for higher rewards in later trials, with their choice propensity modulated by the shared goal. This suggests that the common goal played a significant role in shaping decision-making behavior within a dyadic context. These findings provide additional empirical support for social interdependence theory, which traditionally emphasizes unilateral interdependence but overlooks the impact of dynamic relative position shifts on decision-making. Moreover, our findings enrich the structural contingency model by highlighting the importance of relative position in the dynamic decision-making process. Beersma et al. (2002) developed a contingency model using a speed-accuracy approach to mitigate the potential negative impact of relative performance within a team. They concluded that relative performance position is a crucial factor influencing the effect of goal structure on team decision-making. Building on this, we further explored the influence of relative performance position on decision-making adjustments in a sequential decision-making environment.

Importantly, the present study employed two distinct decision-making paradigms: the BART and the IGT. Although both paradigms simulate real-life choice environments, they engage different cognitive processes and mental experiences, representing non-overlapping decision-making mechanisms. Moreover, risk estimation and reversal learning may differ significantly between the two tasks. Thus, findings from both paradigms further elucidate the modulation of cooperative versus competitive goal structure on two-person decision-making performance and strengthen the cross-task validity of our results. This is consistent with our recent neuroimaging research, which provided behavioral evidence for these effects (Zhao et al., 2023).

Implications for management practice

This research provides valuable insights for improving decision-making performance in management practice and has practical implications for team building and management from an adaptive perspective. Consistent with recent studies (Schulze and Newell, 2015; Heidemeier and Bittner, 2012), our findings highlight the positive impact of competitive performance goals on motivating and coordinating team members within a team and organization. Specifically, the results from Study 2 indicate that competitive goals encourage decision-makers to engage with uncertainty and risky information, fostering a risk-taking propensity aimed at outperforming opponents. Consequently, leaders can design team dynamics to cultivate a moderately competitive atmosphere, which enhances coordination, engagement, and overall team adaptability (Priest et al., 2002). For instance, team leaders often implement reward structures, including monetary incentives and penalties, to motivate team members. By adopting a cooperative-competitive goal structure, leaders can more effectively coordinate team members, leading to heightened motivation with greater precision and lower costs.

On the other hand, personality traits and team-level trait homogeneity can significantly influence team performance. While a body of research underscored the importance of diversity in the workplace (Homan and Greer, 2013; Wang et al., 2019), the findings from Study 3 suggest that homogenous two-person teams may enhance collaborative decision-making. This supports prior research, which proposes that homogeneous teams foster greater intra-team cohesion and perform better in cooperative situations, whereas diverse teams may experience reduced cohesion, leading to conflicts (Crust, 2020), especially in obstructive interdependent relationships. To maintain competitive advantage, organizations can apply insights from our research by paying more attention to the personality traits and homogeneity of team members. By fostering trait alignment, organizations can enhance cohesiveness, improve person-team fit, and ultimately strengthen organizational performance (De Cooman et al., 2016; Zhao et al., 2021).

Limitations and future directions

There are several limitations and directions for future studies that should be considered. First, although the current sample size in each study exceeds the expected minimum calculated using G*Power software, the relatively small sample size may limit the generalizability of our conclusions, particularly with respect to the exploratory analyses. For example, in Study 1, participants were classified into two groups: relatively inferior and relatively superior, with the sample size for the exploratory analysis being nearly half that of the sample size used to test Hypothesis 1a. Additionally, as the majority of participants were university students without practical team experience, this may further limit the external validity of the findings. Future studies could address these limitations by including more diverse samples to enhance the generalizability and external validity of the results.

Second, regarding the experimental paradigm and design, we conducted exploratory analyses to test Hypothesis 3b, which revealed that homogeneity in SVO (Social Value Orientation) types and shifts in goal structures moderated the relationship between goal structure and task performance. However, these conclusions were based on grouping participants according to dimensions such as personality homogeneity and the direction of the goal structure shift, rather than manipulating experimental treatments and comparing task performance across conditions. This approach aligns with previous research (Johnson et al., 2006). Future studies could examine dynamic decision-making scenarios and transitions between goal structures, designing experimental conditions that more accurately simulate real-world environments.

Furthermore, in Study 1, consistent with findings from Xu et al. (2013) showing that the Balloon Analogue Risk Task (BART) is highly stable and minimally influenced by learning effects, the no-common goal condition was presented first, followed by the common goal condition, to minimize potential interference. While previous research indicates that BART exhibits fewer learning effects compared to tasks like the Delay Discounting Task (DDT), the possibility of learning effects cannot be entirely ruled out. Therefore, future studies should incorporate more sophisticated experimental designs to better mitigate the potential impact of learning effects.

Finally, there is ongoing debate regarding the cognitive processing mechanisms underlying the experimental tasks used in this study. For the Iowa Gambling Task (IGT), we divided the task into two phases: decision-making under ambiguity (Trials 1–40) and decision-making under risk (Trials 41–100), based on previous research (Brand et al., 2007; Buelow et al., 2015; Buelow and Barnhart, 2017). To ensure that this division was suitable for the current study, we conducted an additional analysis to investigate whether potential learning effects differed between the two orders of goal structure presentation. Specifically, participants either completed the task with a cooperative goal condition first, followed by a competitive condition, or the order was reversed. The results indicated that the order of goal structure presentation did not influence task performance during the IGT procedure (see the Supplementary file for detailed analysis). However, some studies suggest that the transition between these phases is not distinct, but rather the result of a gradual and subjective learning process (Colautti et al., 2022). To address this inconsistency, future research should employ multidimensional research methods, including neuroimaging and computational modeling. Additionally, with respect to the BART, De Groot and Thurik (2018) proposed that the task measures decision-making under uncertainty rather than under risk, a view that contrasts with recent neuroimaging evidence showing reliable brain activation patterns during risk-taking behaviors in the BART (Li et al., 2020). Future research should focus on distinguishing between decision-making under uncertainty and risk, and refine the presentation of decision-making information in experimental tasks.

Conclusion

In summary, this study explored the impact of cooperative-competitive goal structures on two-person decision-making performance. Through three sub-studies, we developed a theoretical framework demonstrating that two-person decision-making is influenced by the interplay between goal structures, personality homogeneity, and the dynamics of goal structure shifts. Our findings suggest that goal structures significantly shape social interactions by establishing distinct interdependence relationships, which, in turn, influence decision-making propensity and task performance in adaptive decision-making scenarios. These results underscore the importance of understanding goal structures in shaping collaborative and competitive dynamics, offering valuable insights for motivating team members and promoting organizational cohesiveness.