Competition Between Groups And Individuals During Electronic Brainstorming The Quality of the Best Ideas

(1)

Competition Between Groups And Individuals During

Electronic Brainstorming

The Quality of the Best Ideas

Master thesis (8/18/2012)

Wouter Arkink (s2076446) University of Groningen

Faculty of Economy and Business

Department of Innovation Management and Strategy

Supervisors

(2)

W. Arkink – University of Groningen 2

Abstract

This study reports on an experiment that tests the effect of competition on group- and individually-structured Electronic Brainstorming. The study takes an innovative perspective by measuring the performance in terms of the quality of the best five ideas (Girotra et al., 2010). Previous research measured performance in terms of the number of ideas or the average quality of ideas (Amabile and Gryskiewicz, 1987; Diehl and Stroebe, 1987; Dennis, Alan and Valacich, 1994). Innovators are however not interested in the amount of ideas or the average quality of ideas. They are interested in the best ideas (Dahan and Mendelson, 2001; Girotra et al. 2010). Building on extreme value theory, the quality of the best ideas is influenced by the average idea quality or the number of ideas or the variance in the quality of ideas (Dahan and Mendelson, 2001). An increase in one of these three variables leads to a higher average quality of the best ideas (Girotra et al., 2010). The results of the study indicate that competition has a positive effect on the quality of the best ideas generated by both groups and individuals. Furthermore competition was found to have a larger (positive) effect on the quality of the best ideas produced by groups than on the quality of the best ideas produced by individuals. With stimulation of competition, groups even outperformed individuals in terms of the quality of the best ideas. This indicates that contrary to findings of Diehl and Stroebe (1987), there is a reason to work in groups. The increase in performance of competing groups can be explained by an increase in within-group collaboration (Erey et al., 1993; Dennis and Valecich, 1993; Clydesdale, 2006).

Keywords: Strategy, innovation, brainstorming, idea generation, performance, groups, individuals, competition

(3)

1. Introduction

Companies that do not innovate will not exploit changes and are unable to properly adapt to these changes, resulting ultimately in termination (Cooper, 2006). Furthermore innovation significantly increases the long-term performance of companies (Ancona and Caldwell, 1987). Innovation is therefore of critical importance to companies. Drucker defines innovation as “the means by which entrepreneurs exploit change as an opportunity for a different business or a different service” (Drucker, 1985, p. 17). Innovation is often described as the successful implementation of creative ideas (Amabile and Gryskiewicz, 1987; Brazeal and Herbert, 1999; Christensen and Raynor, 2003; Berkun, 2007), some even state that innovation begins with creative ideas (Amabile, Conti, Coon, Lazenby and Herron, 1996). Innovation is almost always the result of a selection process in which the most promising opportunities from a larger pool of generated opportunities are chosen (Girotra, Terwiesch and Ulrich, 2010). A producer of consumer goods for example considers numerous alternatives from which it selects only a few to investigate further (Terwiesch and Loch, 2004).

Two commonly used approaches by which the process of idea generation is organized are group- and individual idea generation (Sutton and Hargadon, 1996). Idea generation in groups, also referred to as team structures, is best described as: “A group of autonomous stakeholders of a problem domain engage in an interactive process, using shared rules, norms, and structures, to act or decide on issues related to that domain” (Wood and Gray, 1991, p.146,). Individual idea generation is characterized by independent stakeholders who generate ideas related to a certain issues (Osborne, 1957; Diehl and Stroebe, 1987).

Sometimes these groups or individuals participate in generating ideas in the light of a competition (Bullinger Neyer, Rass and Moeslein, 2010). Competitions are widely used by companies in the idea generating process (Bullinger et al., 2010). Competition can be defined as: “To ask groups or individuals of competing users to submit solutions to a given task within a given timeframe” (Piller and Walcher, 2006, p. 310).

(6)

al., 1994; Shalley and Oldham, 1997). This lead to the believe that there is no reason to work in groups (Diehl and Stroebe, 1987). Girotra et al. (2010), however recently showed that for innovative purposes, idea generation in groups has the potential to outperform individual idea generation. Instead of measuring the average quality or the number of ideas, they measured the quality of the best ideas. They found that the best ideas produced by groups were of marginally significant higher quality than the best ideas produced by individuals (Girotra et al., 2010).

Based on extreme value theory, the quality of the best ideas is influence by the average quality of ideas, the number of ideas and the variance in the quality of (Dahan and Mendelson, 2001). One can explain differences in the quality of the best ideas by investigating these three process variables (Dahan and Mendelson, 2001; Girotra et al., 2010).

Previous research on competition also took the number of ideas or the average quality as metric of performance (Erey, Bornstein and Galili, 1993; Shalley and Oldham, 1997). Shalley and Oldham (1997) found that competition lead to an increase in the number of ideas. While for example Erey et al. (1993) found that between-group competition lead to increased average idea quality.

Literature shows several gaps. First off, no research was found that investigates the effect of competition on the quality of the best ideas. The number of ideas or the average quality were mostly taken as performance metric (Eery et al., 1993; Shalley and Oldham, 1997). These metrics however are not suitable when measuring innovative performance (Dahan and Mendelson, 2001). Furthermore, no existing literature was found that compares the performance of competing groups and competing individuals (in Electronic Brainstorming). Finally most research on competition as well as groups and individuals in respect to idea generation use pen and paper (Amabile and Gryskiewicz; 1987; Dennis et al., 1994; Girotra et al., 2010), while companies increasingly organize computer-mediated idea generation (Valacich, Dennis and Connolly, 1994; Nijstad, Stroebe and Lodewijkx, 2003).

(7)

and individual idea generation. The focus is on the use of idea generation for innovative perspectives, therefore the quality of the best ideas is the main unit of analysis. By building on extreme value theory, differences in the quality of the best ideas can be explained by differences in average quality, number of ideas or variance in idea quality (Dahan and Mendelson, 2001; Girotra et al., 2010).

Due to time and resource restraints, regarding the amount of participants needed and the time-frame of a master’s thesis, the two forms of competition that are most commonly associated with positive effects on creativity and innovation are investigated: Competition between groups and competition between individuals (Amabile, 1996; Erey et al., 1993; Shalley and Oldham, 1997). A second reason to investigate these two forms is that managers are more inclined to implement changes when performance is expected to increase (Mayne, 2004). Therefore the forms of competition that are associated with positive effects on creativity and innovation, and thus increased performance, are more likely to be used by managers.

This study finds that the best 5 ideas generated by competing groups are of significant higher quality than the best 5 ideas produced non-competing groups. Competing individuals also outperformed non-competing individuals, indicating an overall positive effect on competition. Furthermore, results indicate that an interaction effect exists between competition and group/individual idea generation. Competition has a greater positive effect on the quality of the best ideas produced by groups. Without competition the best 5 ideas generated by groups were not of significant different quality than the best 5 ideas produced by individuals. When competition is introduced however, groups did outperform individuals.

(8)

2. Literature review

Innovation, the successful implementation of creative ideas in an organization, is of critical importance to companies (Amabile et al., 1987; Brazeal and Herbert, 1999; Christensen and Raynor, 2003; Berkun, 2007). But these creative ideas do not just come into existence; they originate in creativity (Amabile., 1996). Creativity is described as the ability to produce ideas that are novel and useful (Amabile, 1996, Mumford and Gustafson, 1988, Oldham and Cummings 1996, Woodman, Sawyer and Griffin. 1993). In this sense creativity is a starting point for innovation. Often organizations are interested in solving certain problems. One can accomplish this by employing creative problem solving. In which numerous ideas are generated, from which a few are selected for further research. This process of creative problem solving is a three-step-process consisting of: Idea generation, idea evaluation and idea selection (Terwiesch and Ulrich, 2009). The goal of idea generation is to generate a pool of ideas, which can be evaluated and possibly implemented (Zaltman Duncan and Holbek., 1973; Van de Ven, 1986; Amabile and Cheek, 1988). There are multiple types of idea generation methods, among them are: Morphological analysis, K-J method, transformational methods and applied imagination (Zwicky, 1969; Hogarth, 1980; Shah, 1998; Osborn 1979). The best known type of idea generation however is brainstorming, which is first mentioned by Osborn (1953). Its name is derived from the fact that one “uses the brain to storm a problem” (Osborn, 1957, p.80). Brainstorming focuses on the amount of ideas produced, under the assumption that quantity should breed quality (Osborn, 1957). Brainstorming tries to increase the amount of ideas generated, which should lead to higher quality (Osborn, 1957; Dahan and Mendelson, 2001). The method can be applied by both groups and individuals (Osborn, 1953).

2.1 Group- and individual idea generation

(9)

statement of Osborn (1953) a lot of laboratory experiments (e.g., Lamm and Trommsdorff, 1973; Diehl and Stroebe, 1987; Nunamaker, Dennis, Valacich, Vogel and Brown, 1991; Mullen, Johnson and Salas, 1991; Barki and Pinsonneault, 2001; DeRosa, Smith, and Hantula, 2007; Paulus and Brown, 2007; Stroebe, Nijstad, and Rietzschel, 2010) were done in order to research productivity of group- and individual idea generation. These laboratory experiments were however not able to verify the claim of Osborn (1953) concerning the amount of ideas produced by groups, researchers found the opposite to be true; groups even generated fewer ideas than individuals (Diehl and Stroebe, 1987; Paulus and Dzindolet, 1993). Because no difference in quality was found, Diehl and Stroebe (1987) even went as far as to state that there was no reason to work in teams .

2.1.1 Factors influencing idea generation in groups

In traditional brainstorming literature several factors were found in relation to these results: Production blocking, evaluation apprehension, free riding and social loafing are factors that hinder the productivity of groups (Taylor et al., 1958; Dennis and Williams, 2003; Heslin, 2009). Production blocking occurs when participants reduce each other’s productivity, because only one person can speak at a time (Diehl and Stroebe, 1987). Second, the fear of individuals to inform the group about their ideas due to fear of negative reactions can lead to evaluation apprehension (Diehl and Stroebe, 1987). Free riding occurs when participants rely on the efforts of other participants when working on a task. A final drawback to team-based idea generation is “social loafing”, which occurs when individuals that are part of team, put less effort in a task than when they work on the task individually (Karau and Williams, 1993).

(10)

leads to more room for both failure and cooperative success (Sutton and Hargadon, 1996; Goldenberg, Lehmann and Mazursky, 2001). Recent research by Girotra et al. (2010) suggests that the latter is true; collaboration leads to more diverse ideas. This is expected to be due to the uncertainty attached to incongruous knowledge components and more potential for creative recombination due to access to a higher amount of (diverse) ideas (Fleming and Sorenson, 2001; Taylor and Greve, 2006; Girotra et al., 2010).

2.1.2 Electronic Brainstorming and idea generation in groups

Earlier research in the field of idea generation was mostly performed offline, using pen and paper (e.g. Diehl and Stroebe, 1987; Fleming and Singh, 2007; Girotra et al., 2010). The introduction of computers and the internet, however, opened up the possibility of Electronic Brainstormings (EBS), whereby participants work together via a computer and/or the Internet (Gallupe, Cooper, Grise and Bastianutti., 1994). Research suggests that EBS reduces some of the negative effects of group brainstorming (Gallupe et al., 1994; Valacich et al., 1994; Nijstad et al., 2003). The participants are for example able to collaborate on tasks with less influence of production blocking since they do not have to wait for others to finish speaking (Gallupe, Bastianutti and Cooper, 1991; Dennis and Valacich, 1993). Participants can take part in EBS anonymously, which reduces evaluation apprehension (Nunamaker et al., 1991; Dennis and Williams, 2003). Social loafing, on the other hand, might increase because of the anonymity (Derosa et al., 2007).

(11)

2.3 Idea generation and competition

EBS also made it easier to organize large idea generating competitions. Via the internet, participants from all over the world can take part in these competitions (Terwiesch and Ulrich, 2009). These contests are also called “innovation contests”. An innovation contest is a particular type of crowdsourcing (Terwiesch and Xu 2008; Terwiesch and Ulrich 2009), which can be defined as: “A (web based) competition of innovators who use their skills, experience and creativity to provide solutions for a particular contest challenge defined by an organizer” (Bullinger et al., 2010, p.291).

Both can be used as an open innovation tool, whereby one not only looks at internal ideas but also at external ideas (Chesbrough, 2003; Piller and Walcher, 2006; Neyer, Bullinger and Moeslein, 2009). Innovation contests can be seen as a competitive setting in which participants can feel independent and do not feel strong restraints while they are challenged and at the same time can obtain valuable comments from other contestants (Bullinger et al., 2010). It is expected that these phenomena increase creativity and innovativeness (Oldham and Cummings, 1996; Tierney and Farmer, 2002). Amongst others, the promised reward is used as a way to stimulate competition between groups or individuals (Lazear and Rosen, 1981). Although innovation contests are gaining in popularity (Bullinger et al., 2010), the effect of competition on the quality of idea generation is not clear-cut (Li and Vanhaverbeke, 2009).

2.3.1 Competition and group- and individual idea generation

(12)

and Oldham, 1997). Amabile (1996) associates competition inside an organization with undesirable effects on creativity, while competition among organizations is related with desirable effects. At team-level, Erey et al. (1993) confirm these result; they found that competition within teams is detrimental to creativity, while competition between teams has a positive effect. Concerning competition between individuals Shalley and Oldham (1997) state that competing individuals produce a higher amount of creative ideas than individuals who are not competing. As discussed above, literature suggests that between-group- and between-individual competition most likely have positive effects on the idea generation process. This study will focus on these two forms, which is shown in table 1; on the right side the non-competitive counterparts are shown. For the rest of this paper “competition” will refer to the two forms presented in table 1.

Table 1

Between-Group and Between-Individual Competition

Competition No competition Groups (collaboration) Intergroup competition, within-group collaboration Within-group collaboration Individuals (no collaboration) Individuals competing against each other Individuals working on a task

Developed based on: (Erey et al., 1993; Amabile, 1996; Shalley and Oldham, 1996)

2.4 Measuring the performance of idea generation

(13)

average quality of these ideas, but to look at the whole probability distribution of the generated ideas (Dahan and Mendelson, 2001). This means that the number of ideas, average quality and the variance in the quality of ideas should be taken into account (Dahan and Mendelson, 2001; Girotra et al., 2010). Girotra et al., (2010) were one of the first to apply such a research design to (offline) idea generation, they found that although groups had a lower average quality of ideas and a lower number of ideas, their best idea(s) were better than those produced by the same amount of individuals. This is caused by the higher variance in the quality of ideas produced by collaborating groups (Girotra et al., 2010).

Quality of ideas is often measured based on (subjective) ratings. These are mostly assigned by experts in the field, often a second expert rated the ideas to verify reliability (Diehl and Stroebe, 1987). Ratings are often based on some sort of Likert scale (Diehl and Stroebe, 1987). While sometimes only one construct was used to measure idea quality, more often several constructs were used. Novelty, feasibility, specificity and relevance were found to be the most used constructs (Dean, Hender, Rodgers and Santanen, 2006). Novelty is often defined as unique or rare ideas (MacCrimmon and Wagner, 1994). Feasibility is associated with the ease of implementation (Kristensson, Gustafsson and Archer, 2004). Specificity is related to how well defined and detailed the ideas is (Dean et al., 2006). Finally relevance refers to the usefulness, for example its financial potential (Dean et al., 2006). There is little consensus on how to report these constructs. Some report them separately (Diehl and Stroebe, 1987), while others combine them into one overall quality score (Gallupe et al., 1992; MacCrimmon and Wagner, 1994).

2.4.1 Extreme value theory

(14)

selected subset of ideas. As companies are normally interested in the best idea(s), one can assume that the best rated idea(s) is/are in this subset (Dahan and Mendelson, 2001; Girotra et al., 2010). Research on idea generation should therefore focus on the best ideas instead of on the average quality or the amount of ideas. This is where extreme value theory comes in; this theory suggests that to statistically estimate the quality of a subset from a larger sample, it is necessary to understand the distribution of the quality of the best idea. The quality of the best idea(s) depends on the quality of the pool of ideas from which the subset is taken (Girotra et al., 2010). The quality of this initial pool of ideas will increase or decrease if one or more of the following three process variables increase or decrease; (1) mean quality of the ideas, (2) number of diverse ideas and/or (3) variance in the quality of the ideas (Terwiesch and Ulrich, 2009; Girotra et al., 2010). Increasing one of these variables, while holding the other two constant, will lead to a higher quality of the selected subset of ideas (for scientific proof of this statement see: Girotra et al., 2010). This theory is graphically presented in figure 1.

Figure 1

Extreme Value Theory, the Quality of the Best Ideas

Figure 1: Adapted from: Girotra et al., 2010

(15)

2.5 Effects of between-individual- and between-group competition

on EBS

According to previous research, adding a competitive element to idea generating groups or individuals was found to have several effects on the performance of the idea generation process. This section will present hypotheses based on these effects.

2.5.1 General effects of competition

(16)

motivation, competition is also associated with less risk-taking, which has a negative effect on creativity (Amabile, 1982). Evaluation apprehension and thinking about rewards too much were found to have a negative effect on the number of ideas as well as the quality of the ideas (Diehl and Stroebe, 1987). Abra (1993), states that competition is enjoyable and that competition can be a powerful foundation of creativity. Finally, the potential of evaluation attached to a competition is expected to decrease social loafing (Karau and Williams, 1993).

2.5.2 Effects of competition related specifically to groups

Within group collaboration is expected to increase when a competitive element is added (Erev et al., 1993; Bornstein and Erev, 1994). As a result of stronger within group collaboration members are expected to share more of their own ideas which increases building on each other’s ideas (Dennis and Valecich, 1993; Clydesdale, 2006). As previously discussed, this might lead to more similar ideas (Kavadias and Sommer, 2009). Recent research performed by Girotra et al. (2010) however suggests that the uncertainty attached to incongruous knowledge components and more potential for creative recombination due to access to more (diverse) ideas leads to more diverse ideas (Diehl and Stroebe, 1987). Erev et al. (1993), found that competition decreased free-riding, which lowered the productivity losses in groups and increased the incentive to produce outstanding ideas.

2.6 Hypothesis development

The literature described above shows several gaps. Often the number of ideas or the average quality were taken as performance metric. Innovators, however, are interested in the best ideas (Dahan and Mendelson, 2001). Furthermore little research was found about competition during idea generation (Erey et al., 1993; Shalley and Oldham, 1997), to date no research on the effect of competition took the quality of the best idea as performance metric. Also, no comparison was found in literature between competing groups and competing individuals. Finally, research on idea generation mostly uses pen and paper, while computer-mediated idea generation is gaining popularity (Valacich, Dennis and Connolly, 1994; Nijstad, Stroebe and Lodewijkx, 2003).

(17)

makes the study more interesting for innovators, as only these ideas will be taken into further research (Dahan and Mendelson, 2001; Girotra et al., 2010). The design of the experiment allows for comparison between non-competing- groups and individuals and competing- groups and individuals. Furthermore it aims to establish if there is a difference in the effect of competition on groups opposed to individuals. Finally the experiment is performed online to increase generalizability to more modern forms of idea generation (e.g. crowdsourcing and innovation contests).

This section will describe how competition and organizational structure (group/individual) are expected to influence the three process variables presented in figure 1. In the next sections the effects of competition found in the literature are linked to the three process variables that influence the quality of the best idea (see figure 1). Each variable will be linked to competition between individuals, competition between groups and finally the expected difference between (competing) groups and (competing) individuals will be discussed.

2.6.1 Variance in the quality of ideas

The variance in the quality of the ideas will influence the quality of the best ideas (Dahan and Mendelson, 2001). Based on extreme value theory an increase of the variance in idea quality is expected to lead to a higher quality of the best ideas (Girotra et al., 2010). This section will describe the expected effects of competition on the variance in the quality of ideas generated by groups and individuals.

2.6.1.1 Variance: Individuals vs. competing individuals

(18)

competition can work as an incentive to come up with novel ideas (Toubia, 2005; Leimeister et al., 2009). No research was identified in terms of the relative size of these phenomena or their combined effect on the variance in idea quality. To explore whether these phenomena cause a difference in the variance of the idea quality between competing- and non-competing individuals, the following hypothesis is proposed:

H1: The variance in the quality of ideas generated by competing individuals is different than the variance in the quality of ideas generated by non-competing individuals.

2.6.1.2 Variance: Groups vs. competing groups

Adding a competitive element to collaborating groups is likely to increase within-group collaboration (e.g. Erey et al., 1993; Bornstein and Erev, 1994). Idea-sharing is expected to increase when within-group collaboration increases (Dennis and Valecich, 1993; Clydesdale, 2006), which would increase the potential for creative recombination and the danger of incongruous knowledge compartments (Fleming and Sorenson, 2001; Taylor and Greve, 2006). Girotra et al. (2010) found that collaborating in a group indeed leads to more diverse ideas. This would suggest that competing groups would have an even higher variance in the quality of ideas than non-competing groups, this leads to the following hypothesis:

H2: The variance in the quality of ideas generated by competing groups is higher than the variance in the quality of ideas generated by non-competing groups.

2.6.1.3 Variance: groups vs. individuals

(19)

H3a: The variance in the quality of ideas generated by non-competing groups is higher than the variance in the quality of ideas generated by non-competing individuals.

H3b: The variance in the quality of ideas generated by competing groups is higher than the variance in the quality of ideas generated by competing individuals.

2.6.2 Average quality of the ideas

The average quality of ideas will influence the quality of the best ideas (Dahan and Mendelson, 2001). Based on extreme value theory an increase in the average quality of ideas is expected to lead to a higher quality of the best ideas (Girotra et al., 2010). This section will describe the expected effects of competition on the average quality of ideas generated by groups and individuals.

2.6.2.1 Average quality: Individuals vs. competing individuals

As previously discussed, depending on how the competition is understood, extrinsic and/or intrinsic motivation might change (Chikszentmihalyi, 1990). Competition may serve as an incentive to come up with novel ideas (Toubia, 2005; Leimeister et al., 2009), while evaluation apprehension and thinking about rewards can have a negative effect on the quality of the ideas (Diehl and Stroebe, 1987). Recognition and rewards attached to a competition were also found to have positive effects on the average quality and number of ideas (Amabile and Gryskiewicz, 1987; Dennis et al., 1994; Amabile et al., 1996; Amabile, 1997). No previous literature was identified that explains the relative size of these phenomena or their combined effect on the average quality of ideas. To explore whether these phenomena cause a difference in the average quality between competing- and non-competing individuals, the following hypothesis will be tested:

H4: The average quality of ideas generated by competing individuals differs from the average quality of ideas generated by non-competing individual.

2.6.2.2 Average quality: Groups vs. competing groups

(20)

(Amabile, 1988; Clydesdale, 2006). Furthermore competition is expected to decrease free-riding in groups. Free-riding decreases the incentive to produce outstanding ideas (Erev et al., 1993). These two phenomena lead to the following hypothesis:

H5: The average quality of ideas generated by competing groups is higher than the average quality of ideas generated by non-competing groups.

2.6.2.3 Average quality: groups vs. individuals

Research on the average quality of groups gave equivocal results, often no difference in quality was found (Diehl and Stroebe, 1987). Furthermore, Girotra et al (2010) recently found no significant difference in the quality of ideas generated by non-competing groups compared to non-competing individuals. Therefore, the following hypothesis is proposed:

H6a: The average quality of ideas generated by non-competing groups is different from the average quality of ideas generated by non-competing individuals.

Average quality of groups might be decreased due to free-riding (Erev et al., 1993). Competition however decreases free-riding (Erev et al., 1993). Furthermore idea-sharing is expected to increase in competing groups, which is expected to increase average quality (Amabile, 1988; Clydesdale, 2006). No clear indications for increased average quality of ideas generated by individuals were found (Amabile, 1983, Ambile et al., 1986; Amabile and Cheek, 1988; Amabile, 1990; Chikszentmihalyi, 1990; Abra, 1993). Based on this the following hypotheses is proposed:

H6b: The average quality of ideas generated by competing groups is higher than the average quality of ideas generated by competing individuals.

2.6.3 Number of ideas

(21)

2.6.3.1 Number of ideas: Individuals vs. competing individuals

Competition increases the incentives to come up with novel ideas (Toubia, 2005), which is beneficial for both groups and individuals. Anticipation of negative valuation of ideas, and thinking about rewards too much, were found to have a negative effect on the number of ideas (Diehl and Stroebe, 1987). Anticipation of negative valuation of ideas might be decreased because of the anonymity in EBS. Recognition and rewards in itself do have positive effects (Amabile and Gryskiewicz, 1987; Amabile et al., 1996). Also the potential of evaluation attached to a competition is expected to decrease social loafing (Karau and Williams, 1993). Although some negative effects of competition (Diehl and Stroebe, 1987) on the number of ideas generated by individuals were found, a great deal of positive effects were found (Amabile and Gryskiewicz, 1987; Amabile et al., 1996; Toubia, 2005). It is expected that these positive effects outweigh the negative effects, this leads to the following hypothesis:

H7: The number of ideas generated by competing individuals is higher than the number of ideas generated by non-competing individuals.

2.6.3.2 Number of ideas: Groups vs. competing groups

The phenomena mentioned above also account for competing groups (Amabile and Gryskiewicz, 1987; Diehl and Stroebe, 1987; Karau and Williams, 1993; Amabile et al., 1996; Toubia, 2005). Therefore just as for competing individuals it is expected that competing groups produce more ideas than non-competing groups. Furthermore, within-group collaboration is expected to increase when groups compete with each other (e.g. Erey et al., 1993; Bornstein and Erev, 1994), which leads to increased idea-sharing (Dennis and Valecich, 1993; Clydesdale, 2006). This leads to the following hypothesis:

(22)

2.6.3.3 Number of ideas: groups vs. individuals

Numerous researchers suggest that groups produce fewer ideas than individuals (Diehl and Stroebe, 1987). Production blocking, free-riding, evaluation apprehension and social loafing all are detrimental to the number of ideas produced by groups (Taylor et al., 1958; Dennis and Williams, 2003; Heslin, 2009). EBS is however often performed anonymous, which might lead to a reduced effect of evaluation apprehension (Dennis and Valacich, 1993; Gallupe et al., 1991). Social loafing on the other hand is expected to increase due to anonymity. Although some of these effects are reduced (Dennis and Valecich, 1993; Gallupe et al., 1991), it is expected that EBS groups still produce less ideas than EBS individuals. The previous two sections propose that competition increases the amount of ideas produced by both individuals and groups. Both propositions are based on the same effects. Assuming that there is no difference in the power of these effects between competing groups and competing individuals; the following hypotheses are proposed:

H9a: The number of ideas generated by non-competing groups is lower than the number of ideas generated by non-competing individuals.

(23)

3. Method

This research analyzes the effectiveness of (between- group/individual) competition and group work in EBS. To do this, the following four conditions (also presented in table 1) are researched: A: Non-competing individuals: participants receive a challenge which they will work on individually; competition will not be stimulated.

B: Competing individuals: participants will receive a challenge which they will work on individually; competition will be stimulated.

C: Non-competing groups: participants will receive a challenge which they will work on in groups; competition will not be stimulated.

D: Competing groups: participants will receive a challenge which they will work on in groups; competition will be stimulated.

The performance of these conditions is measured based on their impact on the number of ideas, average quality, variance in the quality of the ideas and finally the quality of the best ideas. A within 2x2x2 (competition x group/individual x problem) factorial within subject design is employed to make data collection more efficient. Furthermore learning and order effects can be controlled for with this design (Greenwald, 1976). Each participant took part in two conditions, one of these had a competitive element, the other had no competitive element. Also one of the two comprised group work, while the other comprised individual work. This leads to two possible combinations of conditions; A+D and B+C. These two combinations were presented in both possible orders to control for order and learning effects. The design can be found in Appendix (table 2). Groups of 3 were randomly assigned to one of the eight clusters to control for effects known to be related to creativity, like age, experience with idea generation and gender.

3.1 Participants

(24)

5 minute presentation during three classes (of a total of approximately 85 students). The rest was recruited with the help of an email to 86 students. The given presentation and the email that was sent are attached in the appendix. If students agreed to participate they were sent a confirmation email with additional details (see Appendix). The participants were told that two participants would receive a cash prize. No details were revealed about how those two participants were chosen until the experiment started. The response rate was 41.8%, which is higher than expected, given the time of the year (exam periods). All participants were between 19 and 28 years old (average age: 22) and all had the Dutch nationality. From the 72 participants 47 were male (65%). The experiment as well as the presentation and emails were in Dutch.

Give details how you identified the participants and why they could be representative – take account of methodological bias where possible and how this was managed.

3.2 The treatments

Participants were randomly assigned to one of the 8 clusters presented in table 2 (Appendix). Each cluster consisted of groups of three participants. All eight clusters were run three times. Multiple clusters were run at the same time; up to 12 participants took part in the experiment at the same time, in the same computer room. They were unable to watch each other’s computer screens but were able to see each other.

(25)

Participants were asked to generate solutions to one of the two problems stated above. They were told that they could write down any solution and that they had 12 minutes to write down solutions. Finally they were told that they could not talk out loud with each other, should not use the internet or facebook and that they should not worry about spelling. Google docs, an online writing tool, was used to present the problem as well as the purpose of the session. An example of the problems presented in each of the four conditions as well as the explanation of the purpose of that session is presented in the appendix.

In the group-structured conditions were additionally told that they can chat with each other by using the chat function of Google docs, which was situated next to the problem. They were also told not to disclose their identity. Each group member had a participant number, which was used to indicate which participant wrote what solution and was also used a chat nickname.

During one of the two conditions a competitive element was added. When competition was added, the participants were additionally told that the ideas would be counted and rated. Furthermore they were told that the best individual (or group) would win a cash prize of 50 Euro (Lazear and Rosen, 1981). Groups were told that they would win the cash prize as a group. The formulation of all the treatments can be found in the appendix.

The solutions were automatically saved after 12 minutes and participants were told that the first part of the experiment was done. In between the two treatments participants were asked to fill-in a short questionnaire regarding their age, education, gender and experience in brainstorming. When everybody filled-in the questionnaire the second treatment began. The problem they would work on was different than the first session (see above). When the experiment was finished participants were debriefed and given something to drink.

(26)

To control for the learning effect participant worked on a different problem in each condition. Half of them worked first on problem one and then on problem two, while the other half first worked on problem two and then on problem one (see below). This set-up makes sure that order-effects concerned with the problems are controlled for. The two problems are described below:

Problem 1: “You have been retained by a manufacturer of sports and fitness products to identify new product concepts for the student market. The manufacturer is interested in any product that might be sold to students in a sporting goods retailer (e.g., City Sports, Bike Line, EMS). The manufacturer is particularly interested in products likely to be appealing to students. These products might be solutions to unmet needs or improved solutions to existing needs”. (Girotra et al., 2010, p.15)

Problem 2: “You have been retained by a manufacturer of dorm and apartment products to identify new product concepts for the student market. The manufacturer is interested in any product that might be sold to students in a home-products retailer (e.g., IKEA, Bed Bath and Beyond, Pottery Barn). The manufacturer is particularly interested in products likely to be appealing to students. These products might be solutions to unmet needs or improved solutions to existing needs”. (Girotra et al., 2010, p.15)

3.4 Rating the ideas

Only non-redundant ideas (per group per problem duplicates were removed) were taken into the rating process (Diehl and Stroebe, 1987; Girotra et al., 2010). A total of 661 non-redundant ideas were generated by the 72 participants. The number of ideas is measured by counting the amount of ideas per conditions.

(27)

2006). In this case relevance concerns the financial potential of an idea. Ratings were provided based on a 5-point likert-scale (1=very low quality in terms of the rated construct, 5=very high quality in terms of the rated construct) (McKelvie, 1978; Diehl and Stroebe, 1987; Girotra et al., 2010). Each rater rated all the ideas independently on one construct at a time,so that they could focus on just that construct. To increase objectivity, they were not able to see what condition the ideas derived from. To prevent a decrease in rating quality because of exhaustion, rating was separated in several sessions,

To assess inter-rater agreement the two-way random consistency Intraclass Correlation Coefficient (ICC) was calculated for each construct. These were .675 for novelty; .581 for feasibility; .572 for specificity and .569 for relevance. These indicate fair to good agreement following the criteria of Cichetti and Sparrow (1981). Furthermore Diehl and Stroebe (1987) indicated that raters can be considered in agreement when the scores are no further than 1 point apart. Agreement was .86 for novelty; .88 for feasibility; .91 for specificity and .89 for relevance. An average per construct was calculated by averaging the scores of the two raters (Diehl and Stroebe, 1987). High quality ideas should score high on every dimension. Adding the four constructs or averaging the scores would allow for compensation of low scores on one construct (Dean et al., 2006). An idea that has the following score: 5-1-5-5 (novelty, feasibility, specificity and relevance) would be equally good compared to an idea that scores 4-4-4-4 (novelty, feasibility, specificity and relevance). The former idea, although scoring high on three constructs, would be very difficult to implement due to the low feasibility score. The latter does not have this problem and scores reasonably high on all constructs. This difference would not be noticed when adding or averaging the four constructs. To overcome this problem the scores on the four constructs are multiplied with each other to create one overall quality score per idea (Straus and McGrath, 1994).

(28)

idea has an overall quality score of 20 and the average overall quality of all the ideas generated by the specific group of participants is 30, the variance of that idea will be 102 =100.

(29)

4. Results

This section will describe the results of the hypothesis described in section 2.5 as well as an analyses of the quality of the top-5 ideas. To compare two sample means, often a one-way Anova test is used. One of the assumptions of this test is that the data has a normal distribution. The Shapira-wilk test showed that the overall quality rating of ideas as well as the variance in the quality of ideas turned out to be non-normal distributed W(661)=.202, p<.001. Using a series of Q-Q plots (Wilk and Gnanadesikan, 1968) it was found that the overall quality and the variance both came closest to a gamma distribution (figures 2 and 3, Appendix). To compare the mean quality, the variance and the quality of the top-5 ideas between conditions, a Generalized Linear Model (GZLM) was used. This model allows testing of non-normal distributed data (McCullagh and Nelder, 1989). Dummy variables were created for competition (1=competition, 0=no competition), group (1=group, 0=individual) and the problem (1=problem 1, 0=problem 2). The gamma with identity link model was employed for both the variance and the overall quality. Graphical analysis of the standardized deviance residuals shows that this type of model has a good fit. The standardized residuals histogram of both the average and the variance in idea quality show a normal curve (figures 4 and 5, Appendix). Furthermore, deviance values between 0.91 and 1.2 suggest that the data provides a good fit to the chosen GZLM model (McCullagh and Nelder, 1989).1 To compare the average quality and variance in the quality, the estimated marginal means were used (least significant differences adjustment).

The number of ideas was found to be normally distributed using a Shapiro-Wilk test, W(48)=.975, P=.377. 2 Furthermore Levenes tests suggests that homogeneity of variances can be assumed during all three tests (individuals: F(20)=1.83, p=.174, groups: F(20)=.191, p=.901 competing groups vs. competing individuals: F(20)=.179, p=.910). The independence assumptions of parametric tests were not violated. Comparison between ideas produced by the same set of participants will not be made. None of the assumption of parametric tests, such as the

1_{The analysis described was performed on the whole dataset. Analysis on each of the subsamples}

(conditions) lead to the same conclusions, gamma with log identity is the best type of model.

2_{Again, the analysis described concerns the whole dataset. Analyses on each subsample (conditions) lead to}

(30)

one-way Anova are violated, therefore a univariate one-way anova is used to compare differences in productivity.

(31)

Table 4

Results, Comparing Competitive with Non-competitive EBS.

Statistic compared N EMM for competition (SD) EMM for no competition (SD) Difference of EMM: Competition-no competition Wald chi-square (df=1) F (df=1) P-value two tailed Within team-varianceI Groups 265 9262.38 (1231.84) 3912.97 (503.53) 5349.41*** 16.26*** <.001 Individuals 397 3417.55 (383.11) 2664.33 (289.82) 753.22 2.38 0.246

Average quality ratingI

Groups 265 92.94 (8.20) 50.87 (4.38) 42.07*** 20.50*** <.001 Individuals 397 61.73 (3.97) 43.97 (2.79) 17.76*** 13.38*** <.001 Number of ideasII Groups 24 10.58 (1.527) 11.42 (1.527) -0,83 0.15 0.704 Individuals 24 16.42 (1,464) 16.67 (1,464) -0,25 0.015 0.905

Quality of top-5 ideasI

Groups 117 160.65 (14.64) 90.95 (8.01) 68,70*** 17.41*** <.001 Individuals 120 126.24 (9.17) 97.90 (8.66) 28.34* 5.97* 0,03

(32)

Table 5

Results, Comparing Individual and Group EBS

Statistic compared N EMM for groups (SD) EMM for individuals (SD) Difference of EMM: groups- individuals Wald chi-square (df=1) F (df=1) P-value two tailed Within team-varianceI No competition 337 3912.97 (503.53) 2664.33 (289.82) 1248.64* 4.62* 0.032 Competition 324 9262.38 (1231.84) 3417.55 (383.11) 5844.83*** 19.08*** <.001

Average quality ratingI

No competition 337 50.87 (4.32) 43.97 (3.04) 6.09 1.71 0.191 Competition 324 92.94 (8.20) 61.73 (3.97) 31.21*** 14.01*** <.001

Average number of ideasII

No competition 24 11.42 (1.27) 16.67 (1.27) 5.25** 8.57** 0.008 Competition 24 10.58 (1.69) 16.42 (1.69) -5,84* 5.94* 0.024

Quality of top-5 ideasI

No competition 120 90.95 (8.01) 97.90 (8.66) -6,95 0.347 0.556 Competition 120 160.65 (12.00) 126.24 (9.19) 34.41* 5.19* 0.046

* Significant at <.05 level, ** significant at <.01 level, significant at <.001 level. I: Idea rating is the unit of analysis. II: The number of non-redundant ideas is the unit of analysis. EMM are the Estimated Marginal Means after taking into account the other control variables.

4.1 Variance in the quality of ideas

(33)

group) in one specific condition (Girotra et al.,2010). To measure the variance, a dependent variable was constructed that is the squared difference of the quality of an idea and the average quality of ideas generated by a specific team of three participants (working individually or as a group) in single condition. A gamma identity link GLZM was employed for all three (variance in idea quality) comparisons.

4.1.1 Variance: Non-competing individuals vs. competing individuals

No significant difference was found in terms of the variance in idea quality between competing and non-competing individuals, Wald X2 (1)=2.38, p=.246. There is thus not enough evidence to state that there is a difference in the variance in idea quality between competing individuals and non-competing individuals. H1 is therefore rejected.

4.1.2 Variance: Non-competing groups vs. competing groups

For competing groups, a significant difference in the variance in the quality of ideas was found (compared to competing individuals), Wald X2 (1)=16.26, p<.001. This result suggests that, competition between groups has an effect on the variance in the quality of the ideas. More specifically, as expected, the variance of competing groups is higher than non-competing groups. H2 is thus accepted. The big mean difference (5349.41) between the two estimated marginal means suggest a relatively big effect.

4.1.3 Variance: Groups vs. individuals

A significant difference in the variance in idea quality between competing groups and non-competing individuals was found, Wald X2 (1)=4.62, p=.032. This suggests that working in a group has an effect on the variance in idea quality. In particular, non-competing groups produce more diverse ideas than non-competing individuals. H3a is therefore accepted.

(34)

groups and competing individuals in terms of variance in idea quality (5844.83) is relatively large, which suggests a relatively large effect.

4.2 Average quality

This section will describe the effects on the average quality. Comparisons are made between individuals and competing individuals, groups and competing groups and groups and individuals. A gamma identity link GLZM was employed for all three (variance in idea quality) comparisons.

4.2.1 Non-competing individuals vs. competing individuals

A significant difference in average idea quality between competing- and non-competing individuals was found, Wald X2(1)=13.38, p<.001; H4 is thus accepted. This result suggests that competition between individuals has an effect on the average idea quality of ideas. More specifically the results indicate that the average quality of ideas generated by competing individuals is significantly higher than the average quality of idea generated by non-competing individuals.

4.2.2 Non-competing groups vs. competing groups

Also at the group level a significant difference in average idea quality was found, Wald X2 (1)=20.50, p<.001. This result suggests that competition between groups has an effect on the average idea quality of ideas. More precisely, as expected, competing groups produced ideas of significantly higher average quality than non-competing groups. Which leads to accepting H5. The average quality of ideas generated by competing groups was 82.7% higher than the average quality of ideas generated by non-competing groups, this suggests a relatively large effect.

4.2.3 Groups vs. individuals

No significant difference between non-competing groups and non-competing individuals was found, Wald X2 (1)=1.71, p=.191. There is thus not enough evidence that non-competing groups produce ideas of different average quality than non-competing individuals; H6a is rejected.

(35)

accepted. The data suggests that when competition is introduced, working in a group has an effect on the average quality of ideas. Specifically, as expected, competing groups generate ideas of significant higher quality than competing individuals.

4.3 Number of ideas

This section will describe the effects on the number of non-redundant ideas generated. Comparisons are made between individuals and competing individuals, groups and competing groups and groups and individuals. One-way Anova tests were used for all three comparisons.

4.3.1 Non-competing individuals vs. competing individuals

No significant difference was found in terms of the number of ideas generated between competing individuals and non-competing individuals, F (1)=0.015, p=0.905. H7 is thus rejected; there is thus not enough evidence to state that there is a difference in the number of ideas generated between competing individuals and non-competing individuals.

4.3.2 Non-competing groups vs. competing groups

Also no significant difference in the number of ideas were found between competing groups and non-competing groups F (1)=0.15, p=.704. Which leads to rejection of H8. There is not enough evidence to state that there is a difference in the number of ideas generated between competing groups and non-competing groups.

4.3.3 Groups vs. individuals

A significant difference in the number of ideas generated between non-competing groups and non-competing individuals was found, Wald X2 (1)= 8.57, p=.008. H9a is thus accepted. This suggests that working in a group has an effect on the number of ideas produced. More precisely, collaborating individuals generate in the same amount of time fewer ideas than individuals.

(36)

number of ideas. More specifically, as expected the number of ideas generated by competing groups is significantly lower than the number of ideas produced by competing individuals.

4.4 Average quality of the top-5 ideas

This section will examine differences in the average quality of the top-5 ideas between conditions. Per condition, per team of three participants (groups and individuals), the best 5 ideas (based on the overall quality score) were incorporated into the analyses. In the following paragraphs the results of this analyses will be discussed regarding individuals, groups and groups vs. individuals. All analyses are performed with a GLZM with gamma identity link model at an alpha level of .05.

4.4.1 Individuals

Based on extreme value theory, it is expected that the increase in average idea quality of ideas when individuals were in competition leads to an increase in the average quality of the best 5 ideas (Coles, 2001). Therefore the following hypothesis is proposed and tested:

H10: The average quality of the best 5 ideas generated by competing individuals is higher than the average quality of the best 5 ideas generated by non-competing individuals.

As expected, a significant difference was found in the average quality of the top-5 ideas between competing individuals and non-competing individuals, Wald X2 (1)=5,97, p=.030. This suggests that competition has an effect on the average quality of top-5 ideas produced by individuals. More precisely, the top-5 ideas produced by competing individuals are of higher average quality than those produced by non-competing individuals.Based on these results H10 is thus accepted.

4.4.2 Groups

(37)

H11: The average quality of the best 5 ideas generated by competing groups is higher than the average quality of the best 5 ideas generated by non-competing groups.

As expected, the difference in the average quality of top-5 ideas between competing- and non-competing groups found to be significant, Wald X2 (1)=17.41, p<.001. This suggests that competition has an effect on the average quality of top-5 ideas generated by groups. In particular, the best 5 ideas produced by competing groups are of higher average quality than the best 5 ideas produced by non-competing groups. H11 is thus accepted.

4.4.3 Groups vs. individuals

Non-competing groups generated less ideas that were had higher variance in idea quality compared to non-competing individuals. The average quality of ideas was not significantly different. The effect on the quality of the best 5 ideas depends on how the differences in variance in idea quality and number of ideas come together (Coles, 2001). To test whether these differences cause a difference in the quality in the best 5 ideas, the following hypothesis will be tested:

H12: The average quality of the best 5 ideas generated by non-competing individuals differs from the average quality of the best 5 ideas generated by non-competing groups.

No significant difference in the quality of the best 5 ideas was found when comparing non-competing groups with non-competing individuals Wald X2 (1)= 0.347, p=.566. There is thus not enough evidence to state that the top-5 ideas generated by groups are of different average quality than the top-5 ideas generated by individuals. H12 is thus rejected.

(38)

H13: The average quality of the best 5 ideas generated by competing individuals differs from the

average quality of the best 5 ideas generated by competing groups.

When competition was introduced to groups and individuals, a significant difference in the average quality of top 5 ideas was found, Wald X2 (1)=5.19, p=.046. Collaborating in a group seems to have an effect on the average quality of the best ideas that are produced in competition. This leads to accepting H13. More precisely, the average quality of the 5 best ideas generated by competing groups is higher than the average quality of the 5 best ideas produced by competing individuals.

In figure 6 a profile plot of all four conditions is presented. On top of the findings discussed above, the profile plot suggests an interaction effect between competition and groups/individuals. Competition seems to have a bigger effect on groups than on individuals. A GLZM with a identity link was used to analyze the interaction. The type of problem was again controlled for by entering it as a factor. A significant interaction effect indeed was found, Wald X2 (1)= 4.109, p=.043. When competition is introduced, the increase in average quality of the top-5 ideas produced by groups is higher than the increase in average quality of the top-5 ideas produced by individuals.

Figure 6

(39)

5. Discussion

This section will discuss the results of the experiment. First differences in performance between non-competing individuals and non-competing individuals will be discussed. Followed by the differences between competing groups and non-competing groups. Finally the differences between groups and individuals will be discussed.

5.1 Individuals vs. competing individuals

This section will discuss the results regarding differences in performance between non-competing individuals and non-competing individuals. Table 6 shows an overview of these results.

Table 6

Effect of Competition on Individual EBS

Statistic Effect Statistic Effect

Variance in idea quality =

Average quality + Average quality top 5 ideas +

Number of ideas =

+:Significant increase when individuals compete, =: no significant difference when individuals compete, -: significant decrease when individuals compete

(40)

The increase in average quality suggests that the incentive to come up with good ideas, recognition and rewards were more dominant than the negative effects related to competition and average quality (i.e. evaluation apprehension) (Amabile and Gryskiewicz, 1987; Dennis et al., 1994; Amabile et al., 1996; Amabile, 1997; Toubia, 2005; Leimeister et al., 2009). These effects thus seem to be the main drivers of the increase in the quality of the best ideas (Dahan and Mendelson, 2001).

Previous research suggests that the incentive to come up with good ideas, recognition and rewards are also related to an increase in the number of ideas (Diehl and Stroebe, 1987; Dennis et al., 1994; Toubia, 2005; Leimeister et al., 2009). The number of ideas, however, did not show a significant difference when competition was introduced to individual EBS. An explanation for this finding can be that competing individuals internally filtered out ideas of lesser quality. It may thus be that individuals indeed thought of more ideas, but did not wrote them all down.

5.2 Groups vs. competing groups

This section will discuss the results regarding differences in performance between groups and competing groups. Table 7 shows an overview of these results.

Table7

Effect of Competition on Group EBS

Variance in idea quality +

Number of ideas =

+:Significant increase when groups compete with each other, =: no significant difference when groups compete with each other, -: significant decrease when groups compete with each other

(41)

did not show a significant increase. Just as for competing individuals it might be that individuals working in a competing group internally filtered out ideas of lower quality. An additional explanation can be that some ideas were shared via the chat window, but were not written down as a solution. This would explain the increase in variance in quality and average quality, while the number of ideas did not change significantly. This study revealed how competition between groups influences the three process variables that in its turn influence the quality of the best ideas (Dahan and Mendelson, 2001). It shows that increases in the variance in idea quality and the average quality are the main drivers of an increase in the quality of the best ideas (Coles, 2001; Dahan and Mendelson, 2001). It further shows that for innovative purposes, group-structured EBS benefits from a competitive element.

5.3 Groups vs. individuals

This section will discuss the results regarding performance differences between groups and individuals. First non-competing groups and individuals will be discussed, followed by competing groups and individuals. Table 8 shows an overview of the results regarding non-competing groups and individuals. Table 9 shows the results for non-competing groups and individuals.

5.3.1 Non-competing groups vs. non competing individuals

This section will compare the performance of non-competing groups and individuals. Results concerning the non-competing conditions are presented in the table below.

Table 8

Effect of Group Work on non-competitive EBS

Average quality = Average quality top 5 ideas =

Number of ideas -

(42)

Fleming (2007) and Girotra et al. (2010) found that groups produce more diverse ideas. This was supported by this research. Findings concerning the average quality and number of ideas were also in line with previous research; non-competing groups produced significantly fewer ideas that were not of significant different average quality (Diehl and Stroebe, 1987; Dennis and Williams, 2003; Heslin, 2009; Girotra et al., 2010). Furthermore, the average quality of the best ideas produced by non-competing groups was not significantly different from the average quality of ideas produced by non-competing individuals. This is a different finding compared to a study of Girotra et al. (2010). They found a marginally significant increase for non-competing groups. It is however difficult to compare results since this research was performed anonymously. Anonymity might have increased free-riding and thereby lowered the number of ideas as well as the variance in idea quality (Dennis and Valacich, 1993; Gallupe, et al., 1991). Especially groups suffer from a decrease in the number of ideas due to idea building, which could explain the weaker performance of non-competing groups (Girotra et al., 2010). This study shows that during non-competitive EBS there is little reason (for innovators) to perform group-structured EBS . Only when the number of ideas produced are irrelevant, there could be an advantage to non-competitive group-work (Dahan and Mendelson, 2001). In that case the higher variance in idea quality would likely lead to an increase in the quality of the best ideas.

5.3.2 Competing groups vs. competing individuals

This section will compare the performance of competing groups and individuals. Results concerning competing groups and competing individuals are presented in the table below.

Table 9

Effect of Group Work on Competitive EBS

Number of ideas -

Competition Between Groups And Individuals During Electronic Brainstorming The Quality of the Best Ideas