Trust establishment framework between SDN controller and applications

1.   Introduction

Extensive research shows that, under certain circumstances, groups can be far more productive when they use group support systems (GSS) than when they do not [see 8, 9, for overview of GSS Research]. Such productivity increases may increase the likelihood that organizations will survive and thrive. However, increases in group productivity are not always accompanied by increases in meeting satisfaction and research shows that people who find their GSS experience unsatisfying may not use GSS for subsequent tasks ^[19], thereby losing the benefits it can provide to their organization and to society. Meeting satisfaction is an affective arousal with a positive valance on the part of a participant toward a meeting -a good feeling about the meeting.

Many GSS researchers have collected data on meeting satisfaction along with data on group productivity. In some studies, treatments that produced higher levels of productivity resulted in higher levels of meeting satisfaction as well ^{[7],[11],[20]}. In other studies, treatments that produced higher levels of productivity resulted in lower levels of meeting satisfaction ^[5],[14], or non-significant differences in meeting satisfaction ^{[6],[10],[11],[12],[17],[21],[23],[24]}.

Conflicting research results on meeting satisfaction in GSS literature have never been fully explained. An explanatory theory of meeting satisfaction may illuminate the underlying causes of seemingly conflicting research results, and might make it possible for researchers and practitioners to develop systems and methodologies that increase group efficiency and group effectiveness, without decreasing meeting satisfaction. This paper derives Satisfaction Attainment Theory (SAT), a causal model of meeting satisfaction. It then presents a survey instrument for measuring the constructs in the model, and argues the statistical validity of the instrument. Next, it analyzes the results of a field study that tests the propositions of SAT. Finally, the paper discusses the implications of the findings for research and practice, and proposes directions for follow-on research.

2.   Satisfaction Attainment Theory (SAT)

Satisfaction Attainment Theory is a causal model of meeting satisfaction. SAT is based on four assumptions.

If the assumptions of SAT hold, then the satisfaction (affective arousal with a positive valance) one felt after a meeting would be a function of the perception that, balancing conflicting and mutually exclusive goals, the value of ones goals increased, or the likelihood of their success increased because of the meeting. Meetings that produce positive PNGA should also produce high levels of meeting satisfaction and meetings that produce negative PNGA should also produce low levels of meeting satisfaction.

There are at least two aspects of a meeting with which a person could feel satisfaction: (1) the meeting outcomes and (2) the process by which the outcomes were attained. Satisfaction with meeting outcome (SO) is defined as an affective arousal with a positive valance on the part of a participant with respect to that which was created or achieved in a meeting. Satisfaction with meeting process (SP) is defined as an affective arousal with a positive valance on the part of a participant with respect to the procedures and tools used in a meeting.

Based on the assumptions of SAT, we propose the following:

SAT suggests some possible explanations for the seemingly conflicting results in the exploratory meeting satisfaction data that many GSS researchers have collected along with their measures of productivity. Part of the conflict may spring from the design of experimental tasks. If meeting satisfaction is caused by PNGA, then participants with no stake in the experimental task would not be likely to express high meeting satisfaction, regardless of the outcome. For example, a university student participating in a brainstorming task on the topic, “What useful things can you do with an old spark plug?” might not experience positive PNGA regardless of how many creative ideas were generated. Another student in the same task might report high meeting satisfaction because, by participating, she received extra credit in a required course, and so was able to raise her grade (positive PNGA with respect to the goal of academic success). Such an outcome would be consistent with SAT, but may have little to do with experimental treatments. The failure of experimental tasks to impact PNGA may explain why the majority of meeting satisfaction findings in GSS laboratory research have been non-significant.

SAT suggests that another part of the conflict in meeting satisfaction results may have been caused by a confounding of SO and SP. If exploratory data were collected with instruments that lumped SO and SP in a single satisfaction variable, it is possible that experimental treatments had differential effects on SO and SP, yielding aggregate results that were difficult to explain.

Therefore, a rigorous test of the propositions of SAT requires a measurement instrument that clearly distinguished between SO and SP and an experimental task where participants would be likely to have a stake in the outcome, and therefore also in the process. The next section of this paper describes the development and validation of a questionnaire instrument to measure the constructs of SAT. The following section presents an empirical test of the SAT model.

3.   Measuring Meeting Satisfaction

To test SAT, we required a general purpose instrument to measure PNGA, SO, and SP. Green and Taber ^[13] published a robust and reliable instrument for measuring solution satisfaction (an instance of SO) and decision scheme satisfaction (an instance of SP), which was used in a number of studies [e.g., ^{[1],[2],[15],[16]}. However, the questions on their instrument are only useful for measuring meetings where the outcome is a decision. Further, many of the items in the Green and Taber ^[13] instrument focus on a different meaning of the word, “satisfaction” than does SAT. The difference between the two perspectives can be illustrated with these two statements:

The first statement is a judgment of requirements fulfilled. The second expresses an emotional response. Most of the Green and Taber instrument investigates judgments of requirements fulfilled, asking whether certain goals were attained (e.g. efficiency, coordination, understanding). To test SAT, we needed an instrument to measure affective responses to meeting outcomes, affective responses to meeting processes, and PNGA. Toward that end we developed and refined a set of fifteen questions, five for each of the three constructs in SAT (Appendix).

Because SAT frames satisfaction as an affective arousal, questions about SP and SO explicitly ask questions about feelings towards meeting process or meeting outcomes. Because SAT frames the cause of satisfaction as PNGA, questions about PNGA could only be answered positively if the meeting increased the value of goals or increased the probability of goal attainment.

Having developed the questions in English, we translated them into Dutch because we had a rare and valuable opportunity to collect responses from working professionals in The Netherlands from meetings in which they had a high vested interest. The instrument was translated into Dutch by a native speaker. To validate the translation, it was translated back to English from Dutch by a professional translator. The meanings of the reverse translated questions corresponded well with the original English wordings. We then collected responses from 156 working professionals in the Netherlands using the Dutch version of the instrument. Out of 156 questionnaires, five were discarded from the analysis because at least one of the items measuring PNGA, SO or SP were not completed. Of the remaining 151, there were 117 males and thirty-one females (three did not indicate their gender on the survey). The average age of the participants was 40.6 years (sd 10.0) and they had an average of 16.7 years (sd 10.3) working experience. There were eleven groups and group size ranged from seven to twenty-five with an average of 13.7 (sd 4.7). All groups used GroupSystems by GroupSystems.com in their meetings.

The data was first tested for discriminant and construct validity. The results of exploratory factor analysis are reported in Table 1. The result supports a three-factor model in which all items that measure the same construct load heavily on a single factor. Further, no item loads heavily on a factor that includes items from another construct. The Cronbach's a statistics indicated high reliability, as all were 0.94 or higher (Table 1).

A confirmatory factor analysis using AMOS 4.01 was then conducted. The factor loadings are presented in Figure 2 and the corresponding fit measures are presented in Table 2. The $\chi^{2}$ test was significant, but the $\chi^{2}/{\rm df}$ ratio was 1.92, well below the upper threshold of 5.00 recommended by Wheaton et al. ^[22]. The fit measures suggested adequate fit as well: GFI=0.88, AGFI=0.83, CFI=0.97, and NFI=0.95. The root mean squared error of approximation (RMSEA) was 0.077, which is considered adequate as well ^[3]. Thus, the instrument was successfully validated using both exploratory and confirmatory factor analysis.

4.   Validation of SAT

The three propositions of SAT, presented above, were developed into hypotheses in order to test SAT. The hypotheses correspond directly to the propositions as follows:

1. Participants reporting higher levels of PNGA will report higher levels of SO.
2. Participants reporting higher levels of PNGA will report higher levels of SP.
3. Participants reporting higher levels of SO will report higher levels of SP.

SEM was then used to test the path diagram depicted in Figure 1 along with the three hypothesized path coefficients. The fit measures were identical to those of the confirmatory factor analysis (Table 2) and thus the overall fit of the model was deemed adequate. The standardized parameter estimates were then examined for hypothesis testing and all three hypotheses were supported (Table 3). The parameter from PNGA to SO (hypothesis 1) was significant at ${\rm p}<0.000$ and explained 72.3 percent of the variance in SO. The parameter from PNGA to SP (hypothesis 2) was significant at ${\rm p}<0.000$ and explained 41.0 percent of the variance in SP. The parameter from SO to SP (hypothesis 3) was significant at ${\rm p}<0.0$ and explained only 5.8 percent of the variance in SP. Thus, the overall model and the three hypothesized relationships were supported by the analysis.

Figure 2.Figure 2.

The strongest relationship in the model was between PNGA and SO. This is not unexpected because the outcomes of meetings are likely to directly impact goal attainment. If the outcomes are favorable for an individual, meaning positive PNGA, then SO should manifest as well. However, if the outcomes are unfavorable, meaning a negative PNGA, then SO should be low as well. Meetings that fail to produce a meaningful outcome may lower PNGA because the goal to optimize one's resources effectively is thwarted. Thus, to the extent that the outcomes of a meeting are inline with an individual's salient goals, a strong relationship between PNGA and SO should be expected in meetings in which the participants have a vested interest in the outcome, such as the case here.

The next strongest relationship was between PNGA and SP. This relationship may be strengthened further in instances where individuals believe that the tools and procedures used in the course of a meeting could be used again in the future to further increase PNGA. Finally, the relationship between SO and SP was by far the weakest of the three as SO explained only a small portion of the variance in SP. This is to be expected given that PNGA accounts for a large portion of the variance in SP and that the single goal represented by the relationship between SO and SP is only one of multiple salient goals. Nevertheless, the finding does reveal that distinct portions of the variance in SP are explained seperately by PNGA and SO.

5.   Future Research on Meeting Satisfaction

The findings of this study have a number of implications for future research examining the meeting satisfaction construct. First, it is important to adequately define meeting satisfaction and identify its dimensionality. We have defined meeting satisfaction as an affective arousal with a positive valance on the part of a participant towards a meeting. However, other researchers may choose to define meeting satisfaction differently, such as the degree to which a meeting has fulfilled certain requirements. Failure to provide a clear definition of meeting satisfaction reduces the degree to which research on meeting satisfaction can be generalized.

Second, if meeting satisfaction is defined in terms of affective arousal, then it is important to also distinguish between satisfaction with meeting outcomes and satisfaction with meeting process. It is entirely possible that certain collaborative technologies could produce high levels of one dimension, such as satisfaction with meeting outcomes, without producing high levels of the other dimension, such as satisfaction with meeting process.

Third, it is important to theorize about PNGA when investigating meeting satisfaction. For example, interventions and technology that increase the salience of goals that are favorably impacted by a meeting would tend to increase PNGA which would, in turn, increase SO and SP. However, students participating in laboratory experiments may find their PNGA largely unaffected by the variance in experimental outcome. In such cases, researchers may be in a better position to draw conclusions about other constructs, such as group productivity, than meeting satisfaction. Experimental attempts to manipulate meeting satisfaction should directly examine the impact of the experiment on PNGA.

We encourage researchers to adopt these aforementioned strategies when investigating meeting satisfaction in the lab or the field. Such action may improve the consistency that GSS researchers are able to accurately hypothesize about the influence of various applications and experimental conditions on meeting satisfaction.

6.   Conclusion

Meeting satisfaction is an important measure of collaboration technology effectiveness. In this paper we have derived Satisfaction Attainment Theory (SAT), which suggests that PNGA causes SO and SP and that SO causes SP independent of PNGA. An English and Dutch version of a questionnaire to measure PNGA, SO, and SP is presented in Appendix A. The Dutch version was validated with responses from 151 working Dutch professionals in The Netherlands. SAT was also validated using the same data set. SAT should prove useful to researchers interested in the meeting satisfaction construct and methods for applying SAT to future research are presented.