Performance and affects in group problem-solving

The sample used in the study consisted of 114 undergraduate and graduate students at public and private institutions in the city of Salvador, Bahia (Brazil). Of this total, 60 were allocated to the group condition and 54 to the individual condition; 61% were female, the mean age was 24 years, and 83% had yet to complete their higher education, predominantly in the social and humanities areas (65%),with 48% majoring in psychology. Convenience sampling was the ion method employed. In each session, four or five participants were allocated into the two experimental conditions. All participants gave their formal authorization for participation by a Statement of Free and Informed Consent.

This is an experimental study in which the independent variable was the modality of accomplishing the task proposed (individual vs. group). Individual condition is the control group since it is aimed to assess the group condition. The dependent variables were the affective states, performance in the game (task), and the conflict perception related to teamwork. The experiment consisted of completing a task based on the computer game X-Construction Lite , whose goal was to build bridges with a specific structure to allow the passage of a train, without collapsing 2 2

The game offers pieces in the shape of beams that are available on the screen for the participant to in order to build the bridge. The quality of the bridge depends on how the beams are arranged and support the passage of a train.

Analysis of affective states identified pronounced asymmetries and curtoses for the depressed, angry, and unhappy states. This result guided the decision to consider only the frustrated and worried states for negative affects. Aiming to keep a balance between positive and negative affects, only the satisfied and optimistic positive states were retained, as better erpoints to the negative affective states considered.

On comparing affective states before and after the task, significant differences were identified for positive affects, t (113) = 2.335, p < .05, r = .21, and negative ones, t (113) = 8.26, p < .001, r = .61, although only the latter showed a large effect, acing for 38% of the total variance. Before the experiment, the positive affective states in the individual ( M = 4.57, SD = 1.20) and in the group ( M = 4.78, SD = 1.30) condition did not differ, t (112) = -.856, p = .39, as was expected. After the task, a significant difference was obtained, with a medium effect size, t (112) = -3.651, p < .05, r = .33, in which those who performed the task in a group ( M = 4.83, SD = 1.50) experienced more intense positive affective states compared to those who performed individually ( M = 3.84, SD = 1.37).

Similarly, it was found that the negative affective states of participants in the individual condition ( M = 3.53, SD = 1.28) and in groups (M = 3.27, SD = 1.24) did not differ before the task, t (112) = 1.094, p = .28, but differed afterwards, t(95,714) = -3.462, p < .05, r = .33, with a medium effect size. Accordingly, participants who performed the task individually ( M = 2.79, SD = 1.25) had higher negative affects scores in relation to those in the group condition ( M = 2.08, SD = .90). Thus, the following hypotheses were corroborated: hypothesis 1, which predicted differences between the affective states before and after the task, hypothesis 1a and hypothesis 1b, which postulated that the participants in the group condition would higher positive affect and lower negative affect scores in relation to those in the individual condition, with the group's performance being higher.

We evaluated the dimensionality of conflict perception, regarding group work, using principal axis analysis with oblique rotation, which revealed, in the moments before and after the task, respectively, a two-factor solution ( Friedman, Tidd, Currall, & Tsai, 2000; Jehn, 1994, 1997 ) and a one-factor solution ( De Dreu & Weingart, 2003; Martins & Puente-Palacios, 2010 ). The two-factor solution classifies the conflicts in terms of tasks and relationships. To allow for comparability before and after the task, the one-factor solution was chosen. The explained variance in post-task teamwork conflict perception was 52.8%. The factor loadings of the items ranged from .57 to .86.

The comparison of the conflict perception, between the experimental conditions, showed that both cases differed before, t (112) = 3.008, p < .05, r = .27, and after, t (110.53) = 3.863, p < .05, r = .34, the task. Before the task, participants in the group condition ( M = 2.40, SD = .50) had a lower perception of team conflicts when compared to those in the individual condition ( M = 2.68, SD = .47). After the task, these results held, both for individuals ( M = 2.67, SD = .46) and for groups (M = 2.29, SD = .58), which was clearly expected only for those who performed the task individually. These findings corroborate in part the second hypothesis of the study, which predicted differences in conflict perception, after the task, only for participants in the group condition.

The first indicator of the performance result measurement was general performance, calculated based on the game level and the score achieved. This indicator showed a significant difference, with a large effect, between the experimental conditions, t (101.525) = -6.672, p < .05, r = .55, in which the average score in the team condition ( M = 724.43, SD = 232.58) was 84% higher than in the individual condition ( M = 393.48, SD = 290.15), corroborating hypothesis 3.

The second indicator of the performance result ranked performance into three categories (31% low, 37% medium, and 32% high), balancing the frequency distribution (simulating a normal distribution). On comparing the expected frequency with the one observed, a significant difference was verified for both individual performance (? 2 = 18.11, p < .05) and for that of the group (? 2 = 21.7, p < .05). Fig. 1 shows the better performance of the participants in the team condition. The participants appear to be aware of this result, as there was a significant difference in the perception of satisfaction with performance, with a large effect, t (107.852) = -6.861, p < .001, r = .55, when comparing the individual condition ( M = 1.81, SD = .65) with that of the team ( M = 2.80, SD = .88), considering the scale of 1 ( none) through 4 (very much).

Figure 1

Performance in the task by experimental condition.

Regarding the problem-solving process, there was no significant difference, t (112) = –1.202, p = .23, between groups (M = 2.70, SD = .53) and individuals ( M = 2.58, SD = .50) in the overall score, which did not corroborate Hypothesis 4. Only the items related to the identification of strategies, t (112) = -1.714, p < .05, r = .16, and information sharing, t (112) = -5.119, p < .001, r = .44, showed significant differences, with small and medium effects respectively.

Correlations analyses between the perception of the process, satisfaction with performance, and the result itself revealed, in the individual condition, a correlation only between the identification of strategies stage and satisfaction with performance; in the group condition, satisfaction was significantly correlated with all of the process items, except the one relating to planning. The correlation between performance (result) and the resolution process showed that organizing information was the only stage with a positive correlation for both experimental conditions, which makes sense given the nature of the task, requiring a very clear visual gestalt. Those results are reported in Table 1.

Upon controlling the effect of the problem-solving process evaluation in the relation between satisfaction and performance, the reduction was lower in the group condition than in the individual condition. Indeed, in the latter case, more correlations were found between performance and the stages of the resolution process than in the group condition.

In the task evaluation, the items: like the task, t (89.234) = -2.508, p < .001, r = .26, satisfaction with performance, t (107.842) = -6.861, p < .001, r = .55, and holding attention during the game, t (96.223) = -2.254, p < .001, r = .22, were favorable to groups. Although only satisfaction presented a large effect, the results indicated an alignment between the items. In the group condition, the task was assessed as quite pleasurable (83%), consistent with the low post-task conflict perception, as team performance was related to the interaction among its members (88%).

In analyzing how much the conflict perception, developed during the task, affected group performance, no correlation was found between the constructs ( r = .003, p = .49), precluding regression analysis (hypotheses 5, 5a and 5b). Path analysis was used to test the mediational role of affective states between the experimental condition (individual vs. group) and performance (Hypothesis 6). The d absence of multicollinearity was addressed by the correlations between the criterion variable, performance, and the predictor variables: experimental condition ( r = .54, p < .001), positive affects ( r = .38, p < .001), and negative affects ( r = -.31, p < .001), and by the analysis of standardized residuals that placed less than 5% of the cases outside the range of ±2; casewise diagnostics (Centered Leverage Value, Cook's Distance, and Mahalanobis Distance) suggested the absence of undue influence on the model. The assumptions of homoscedasticity and linearity were addressed by observing the graph of the standardized predicted values versus the standardized residuals from the regression.

As a first step in the path analysis, we tested whether experimental condition would exert an influence on the affects experienced. We performed a simple linear regression for each of the affects, using the enter method, the experimental condition explained 11% of the variance of the positive affects, R = .36, adjusted R 2 = .11, F (1,112) = 13.332, p < .001, and 9% of the negative affects, R = .32, adjusted R 2 = .09, F (1,112) = 12.390, p < .05. Then, we performed a multiple regression analysis between the predictor and criterion variables, opting for the stepwise regression model. The Durbin-Watson statistic was 1.35, an acceptable value, indicating the independence of errors.

The results showed that the experimental condition explained 28% (adjusted R 2 ) of the variation in performance, a value that rose to 32% when the model aggregated positive affects and excluded negative ones. The association between criterion and explanatory variables was moderate ( R = .58). Table 2 shows the standardized regression coefficients, which indicated that the experimental condition contributed more (47%) to performance than did positive affects (22%). In view of these results, H6 was rejected.

We also evaluated, using path analysis, the mediational role of the affective states in the relationship between the experimental condition and conflict perception (Hypothesis 7). Similarly, there was a lack of multicollinearity between the variables, and of undue influence in the model. Conflict perception correlated negatively with experimental condition( r = -0.34, p < .001) and with positive affects ( r = -.17, p < .05), and positively with negative affects ( r = .19, p < .05).

The first stage of the path analysis (to test if the experimental condition has an impact on each of the affects) was already done in Hypothesis 6. The next step was the multiple regression analysis between the predictor and criterion variables, choosing again the stepwise model. Experimental condition was the only predictor, explaining 11% (adjusted R 2 ) of the variation in conflict perception, reflecting the weak association between the variables ( R = .34). The standardized regression coefficient indicated that the experimental condition contributed 34% to conflict perception. The exclusion of positive and negative affects from the model does not corroborate the mediation relationship (Hypothesis 7).

Consistent with the hypotheses of this study, group performance was superior to the individual's, and was related more to positive than to negative affective states, indicating a more positive perception by these participants about their own performance. Group performance, however, did not translate into a better problem-solving process, when compared to the individual condition. In fact, there was a greater number of correlations between the problem-solving stages and performance in the individual condition. This result-process dichotomy suggests that the problem-solving stages can contribute heterogeneously to performance. The only stage that was correlated to performance for the two experimental conditions was that of organizing information, which may indicate that it has a key role in problem-solving. Although the study sample was comprised of college students, this work group is a regular methodological practice in universities, which allows finding evidence on the performance of teams in general.

As expected, the task had the capacity to activate affective states, but in a lower intensity than predicted, which may have been influenced by the artificial condition in performing the task (no direct consequences for the participants); a real situation might have a greater potential to induce emotions of higher intensity ( Marston, Hart, Hileman, & Faunce, 1984 ). The disjunctive character of the task (Steiner, cited by Mcgrath, 1984 ) may have led the group to choose its most skilled member(s) to execute the task (social loafing), reducing the motivation and the mobilization of affective states of the other participants. The lack of task interdependence may also have contributed to a lower emotional intensity. The irrelevance of the tasks to participants’ personal goals also minimized the direct impact of the interdependence of results on the affects experienced. When the task was chosen for this study, it was supposed that the attraction of computer games would be enough to trigger a strong emotional activation, minimizing effects of artificiality and the design of the task, which actually did not occur.

It is believed that the moderate emotional activation affected conflict perception, which remained unchanged in the group condition. The results seen here corroborate previous studies ( Bodtker & Jameson, 2001; Jones & Bodtker, 2001; Nair, 2008 ), which suggest that emotional ambience is associated with the maintenance of conflict, since the triggers of both are the same. According to Forsyth (2000) , it is, in particular, conflict or intergroup competition that prepares the ground for the emergence and intensification of negative emotions. Competition was not part of the task design and, in the case of conflict, the group dynamic did not change the initial perception, which was already low. Furthermore, the fact that the group interaction was only to perform a single task may also have been an influence, since groups whose relationships have no continuity tend toward a lower susceptibility in the onset of conflicts ( Jehn & Mannix, 2001).

The levels of affective state activation and conflict perception evaluated post-task did not allow us to prove the impact of conflict perception on performance and the mediation of affective states in the relationships between the experimental condition and these variables. To overcome the main limitation of the study, the activation of affective states, we suggest the use of priming and manipulation checks to assess the effective induction of the affective states.

We can point out several directions for future studies. There is an identified need to the effects of dispositional affects in the evaluation of performance beyond the momentary states considered in this study. This line of research is based on the argument that an individual's characteristics can have an influence on behaviors and outcomes over time. Moreover, it is necessary to explore whether dispositional affects would be triggers of more transient affective states that would have impacts on performance ( Staw & Barsade, 1993).

Research that seeks to understand the complex role of the emotions in the moment of conflict, and in its resolution afterwards, will expand the knowledge about the intricate relations between the constructs. We propose, additionally, investigating the role of the positive emotions, addressing management (regulation) and the expression of affective experience. Finally, the use of complementary methods, such as observation, could provide additional objectivity in assessments of emotions in teams or organizations ( Huy, 1999).

In closing, we suggest two questions for future studies: What is the role of each of the problem-solving process stages in the effective performance of the group? Could emotional intensity be activated more in longitudinal studies that consider the performance of several tasks, in order to better test its mediation in group performance? These are just a few possibilities for transforming the hypotheses presented here on research questions even more complex and relevant.