Incentives and creativity in groups ú - PDF

Incentives and creativity in groups ú Joachim Ramm, Sigve Tjøtta and Gaute Torsvik June, 2013 Abstract In his best selling book Drive Dan Pink hammers the idea that money incentives will restrain individual

Please download to get full document.

View again

of 16
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.


Publish on:

Views: 32 | Pages: 16

Extension: PDF | Download: 0

Incentives and creativity in groups ú Joachim Ramm, Sigve Tjøtta and Gaute Torsvik June, 2013 Abstract In his best selling book Drive Dan Pink hammers the idea that money incentives will restrain individual creativity and hamper performance in jobs requiring out of the box thinking. This paper reports from an experiment designed to test if the perceived negative incentive e ect on creative problem solving stretches over to situations where a group of individuals work together to find a solution. We do not find a negative impact of incentives on group performance. As a comparison we ran the same experiment (the candle problem) with and without incentives for individuals as well. Incentives did not reduce performance there either. Comparing individuals with groups we find that team-work seems to facilitate creative problems solving. JEL Codes: J24, J31, O31, M11. Keywords: Incentives, innovation, creativity. ú We thank Kjell Arne Brekke and seminar participants at the Choice Lab seminar (NHH) and at the Department of Economics, University of Bergen, for helpful comments. We thank Sebastian Skancke and Eirik Strømland for excellent research assistance. Minestry of Local Governemnet and Regional Development University of Bergen, University of Bergen, 1 1 Introduction Some tasks can be solved in a mechanic way by exploiting existing knowledge and by following well rehearsed routines and practices. Insight problems, on the other hand, require a more novel and explorative approach. Finding a solution often entails a reframing and restructuring of the task. One has to look at the problem from an unusual angle - think outside the box - to make progress. How does monetary rewards a ect our ability and e ort to solve such insight problems? In his best selling book Drive Dan Pink (Pink (2010)) hammers the idea that money incentives impairs individual creativity. He gets his ammunition from experiments done in psychology and economics. One of the experiment he leans on is the Candle Problem which was invented by Karl Duncker (Duncker (1945)) to study functional fixedness in cognitive reasoning. Duncker showed that it is hard for grown ups to solve problems requiring out of the box thinking as we have acquired a mental block against using objects in an unfamiliar, novel way. The psychologist Sam Glucksberg (Glucksberg (1962)) ran a Candle Problem experiment with monetary rewards and his results indicate that money incentives aggravates the problem of functional fixedness. Inspired by this finding, we wanted to study if money incentives also hampers creativity in groups It is an important question. The rewards associated with finding novel solutions to intricate puzzles can be substantial and a team of workers are often assigned to solve such problems. The challenge and intensity of operating in a group, playing ideas with team-mates, may overturn the mechanisms that transform large stakes to enhanced functional fixedness (less creativity) in an individual setting. Could it be that explorative work is organized in groups precisely because the substantial rewards that are awaiting those who find a solution do not block creativity in teams? To examine this possibility we ran a Candle Problem experiment where two individuals worked together to come up with a solution. We organized one session with monetary rewards to the winners and one session without rewards. The results show no negative e ect of 2 incentives. In fact average performance is slightly better when performance is incentivized, but the di erence is not statistically significant. Encouraged by this finding we wanted to replicate the Glucksberg study, i.e. with individual problem solvers. Based on the previous finding we expected incentives to have a negative e ect on performance. But again, our results show no di erence in performance between incentivized and non-incentivized participants. We think the fact that we cannot reject the null hypothesis warrant some attention as it goes counter to the widespread notion that incentivizing performance impair the kind of creativity needed to solve the Candle Problem. In addition to this null result we do, however, find a significant di erence between individuals and groups. Those working together perform much better in the Candle Problem. Of course just the fact that a group consists of several minds - two in our case - implies that we should expect them to arrive at a solution faster than a single mind. But even adjusting for this, we find a substantial positive group e ect in performance. Our finding indicates that individuals are more creative, better at engaging in out of the box thinking, when working together in a group than when working alone. Despite a booming research on group e ects in decision making we are not aware of any studies examining group e ects on creative problem solving. 1 Hence, we believe this paper provides novel insights into an important question. 2 The research questions We address two questions. How do groups solving insight problems respond to monetary incentives? And irrespective of the extrinsic rewards that are awaiting, are groups better at solving insight problems than individuals? 1 Charness and Sutter (2012) and Kugler et al. (2012) survey recent experimental work on team decisions. Charness and Sutter conclude that teams tend to behave more in line with game theoretic predictions than individuals do. Kugler et al argue that heterogeneity in the levels of cognitive abilities, beliefs and attitudes towards risk among team members is causal for the di erence in decision making by groups and individuals. 3 There are two important di erences between mechanical tasks and insight problems. First, experience is not always an advantage for solving insight problems as functional fixedness - the inability to consider an object or a problem from an unusual angle - may prevent the kind of exploration that is needed to find a solution (Duncker 1945). Another di erence is that individuals usually have a higher level of intrinsic motivation for insight problems than for mechanical problems (Deci 1971). Due to these di erences extrinsic rewards - performance pay - may work di erently on insight problems and mechanical problems. It is uncontroversial - and well documented - that su ciently strong performance incentives elevate the attention and e ort of those solving mechanic tasks. 2 This may not be true for insight problems. In a well known study Edward L. Deci argues that for tasks that are high on intrinsic motivation, extrinsic rewards may lead to over-justification and may crowd out the inner drive to such an extent that the incentives reduces performance Deci (1971). 3 In a study that has received less attention - at least until Dan Pink revitalized it - Sam Glucksberg suggests that extrinsic incentives will aggravate functional fixedness; incentives may prevent exploration and out of the box thinking. To test this hypothesis Glucksberg ran a variant of the Candle Problem. It is this problem we extend by letting groups solve it. In the Candle Problem participants are presented with a candle, a box containing tacks and a book of matches. They are asked to attach the candle to the wall by only using the objects presented so that the candle can burn properly, and that no wax will drip on the table or on the floor. The participants usually have a time limit to find a solution. To solve this problem one has get the insight that the box containing the tacks can be used as a platform for the candle. One literally has to engage in out of the box thinking. See figure 1 for the 2 Performance pay may, however, not always work as intended (by the incentive providers) for more mechanical tasks either, as incentives may lead to task shifting and gaming (Holmstrom and Milgrom, 1991, Oyer, 1998) and may also displace other forms of extrinsic motivations (fairness concerns, social appraisal, etc.) (Gneezy and Rustichini, 2000; Bowles and Polania-Reyes, 2012) 3 Deci ran an experiment and found a pattern in performance consistent with then notion that incentives may have a detrimental e ect on puzzle solving. The interpretation of this result - that it leads to over justification which then crowds out intrinsic motivation - is contested (Fehr and Falk, 2002). There are however several other studies that show a similar e ect, see Frey and Jegen (2001) for an overview. 4 time limit to solve the problem. Figure 1 - The Candle Problem 5 Standard setup Solution Figure 1. The Candle Problem standard setup and solution to the Candle Problem. Note that if the tacks are delivered outside the box, the issue of functional fixedness is not as pressing (when tacks are in the box one has to reinvent the box as a platform). This fact was exploited by Glucksberg when he examined the e ect of economic incentives. In Glucksberg s experiment, participants were divided into two groups. Each group conducted a di erent variation of the Candle Problem, one in which the box was empty and 4 Note that in this depiction of the setup a book of matches is used, in the version conducted in regards the to tacks this thesis were placed a box of on matches the table were ( tacks used. Using outside a box box of matches version), gives and an one additional in which way the to tacks reach were the situated solution; inside the box the containing box ( tacks the matches inside can box also version). be used a platform or container for the candle. The participants in both groups were assigned either to a non-reward condition named 5 low-drive (non- incentivized group), or a reward condition where they could receive $5 for being among the top 25% fastest solvers, or $20 for being the fastest, named the high-drive condition (incentivized group). A failure was defined as being unable to find the correct solution within the time limit of 15 minutes. In the tacks outside box version, the results were as expected; the group with the financial incentive performed better, they had a lower fraction of failures and a shorter solution time. The surprise came in the tacks inside the box version, this time the incentivized group performed significantly worse than the group working without incentives. 5 Our first research question is to examine if the - apparent - negative incentive e ect on human creativity carries over to groups? It seems like an important research question as creativity and novel solutions are often associated with considerable rewards. The rewards can be part of a designed incentives scheme within an organization but that is not always the case. The rewards can also come from higher market value, improved career opportunities and social appraisal. If high stakes impair individual creativity by preventing out of the box thinking it is of considerable interest to examine if this also is the case when a group of individuals interact to solve such problems. Another, perhaps more basic research question is if there is a positive group e ect for these kinds of problems. Suppose you have 10 individuals available and you want to minimize the expected time it takes to solve an insight problem (the Candle Problem for example), should you arrange 5 groups of two individuals or let the 10 persons grope for a solution in isolation? The answer to this problem is not obvious. It has been shown in both strategic and non-strategic settings that it matters whether the decision maker is an individual or a group of individuals. In general, groups appear to be more strategic and goal oriented decision makers. They appear to be better at processing information and they solve general reasoning problems faster than individuals. 4 These findings are suggestive for what we should expect in our case as well, but it has also been shown that groups do not pool information in an e cient way and that some perspectives, some ideas, may be suppressed by the majority or by a subgroup of dominating persons within a group ((Kugler et al., 2012)). Hence it could be that fewer ideas will be explored in a unified group than if the members groped for a solution in isolation. 5 4 Confer Charness and Sutter (2012) and Kugler, Kausel and Kocher (2012) provide recent surveys on how team decisions compare with individual decisions. 5 We have never seen a group version of the Candle Problem (with or without incentives). There are however many other hypothesis that has been examined by using the Candle Problem: Does living abroad increase creativity Maddox and Galinsky (2009): Are older kids more prone to functional fixedness than younger ones, German and Defeyter (2000): Are creative individuals more dishonest Gino et al. (2012) to name a few. 6 3 The Experiments As our goal is to check if money rewards hamper creativity also in groups, we only organized the Candle Problem with stacks inside the box The group session In our first session (May 2012) we recruited 60 undergraduate students from the University of Bergen. We randomly grouped them together two and two and - again randomly - assigned pairs either to the incentivized or the non-incentivized version of the experiment. All participants received 100 NOK upon arrival, 1 NOK is approximately 1/6 US dollar. Those in the incentivized version of the experiment were informed that each member of the pair with the shortest solution time would receive a prize of 1000 NOK, pairs in the second and third place would each get 200 NOK. They where also told that that the total available time was 15 min. Participants in the session without incentives where told that the task was to solve the problem and that time would be measured and that there was a time cap on 15 min. We ended up with 16 incentivized and 14 non-incentivized pairs. After completing the task, the participants were asked if they were familiar with the problem. Those who were are excluded from the analysis. Our analysis is based on 15 incentivized and 10 non-incentivized pairs. 3.2 Individual sessions To contrast our finding in the group session with the outcome when individuals solve the problem in isolation - that is, to replication the Glucksberg study - we requited (a month later, june 2012) 29 students from University of Bergen; 14 students solved an individual incentivized Candle Problem and 15 participated in a session without incentives. We ran 6 Due to the di culty of finding a book of matches - used in the Glucksberg experiment - the participants got a box of matches. 7 another individual session in February We recruited 50 students to participate in the experiment in order to have a sample large enough to check for gender di erences in how participants responded to incentives (in the Glucksberg experiment only men participated). The individual sessions are pooled together when we compare individuals and groups. Participants were randomly assigned to the di erent treatments (incentives or not). The magnitude of the individual rewards were also the same as in the group session. We asked participants in the individual sessions if they were familiar with the Candle Problem. Only 3 confirmed (all in February 2013 sessions), and they are excluded from our analysis which is based on 39 incentivized and 40 non-incentivized individuals 4 Results We compare two performance measures across sessions, success ratios i.e. the fraction of participants that manage to solve the problem, and for the successful, the time they use to solve the problem. Incentives and creativity Our first finding is that monetary incentive does not impact on the two above mentioned performance indicators, neither for individuals nor for groups. The mean solution times together with the success ratios are presented in Table 1. 8 Table 1. Performance with and without incentives Success ratio Fraction (N) Solution Time Mean (Stdv) Individuals Non-incentivized 0.77 (40) 5.38 (3.65) Incentivized 0.72 (39) 6.45 (3.90) All 0.75 (79) 5.89 (3.78) Groups Non-incentivized 1.00 (10) 3.41 (1.55) Incentivized 1.00 (15) 2.97 (2.07) All 1.00 (25) 3.15 (2.69) The success ratio is given by the number of individuals (pairs) that solved the problem within 15 minutes divided by the number that participated in the session. The mean solution time is conditional on having solved the problem. All groups solved the candle light problem within the time limit. Groups that are incentivized solve the problem slightly faster than the non-incentivized groups, but the di erence in average solving time is not significant, according to a Kruskal-Wallis rank test, p =0.36, two sided. Incentivized individuals have a slightly lower success rate than those who perform without money rewards, but, again, the di erence is not significant (a binomial probability test, p =0.49, two sided test). Conditional on having succeeded, individuals who are incentivized need more time to solve the problem than those who are not rewarded. The di erence is relatively small and insignificant (Kruskal-Wallis rank test, p =0.26, two sided). We are not able to reproduce the Glucksberg (1962) results. This could be due to the gender composition of the participants. We use a mix of male and female students, whereas in the Glucksberg experiment only male students solved the problem. To examine this possibility we split our individual sessions into a male and a female sub-sample. There is, however, no significant gender di erence in how individuals respond to incentives. 7 The only 7 Female success ratios are 0.64 and 0.68 in the incentivized and non-incentivized sessions. They are not significantly di erent according to a binomial probability test (p =0.81, two sided). On average incentivized females need more time to solve the problem than those who work without extrinsic rewards, 6.72 minutes compared to 4.81 minutes. The di erence is significant at moderate significance level (Kruskal-Wallis rank test, p = 0.09, two sided). Male success ratios are 0.82 and 0.86 for incentivized and non-incentivized respectively, the di erence is not significant di erent (a binomial probability test, p =0.72, twosided). 9 statistically significant gender di erence we find is that a higher fraction of men manage to solve the problem (a binomial probability test, two-sided, gives p =0.04). Solution times are not significantly di erent for males and females. Group creativity Table 1 indicate that there is a positive group e ect. Groups perform considerable better than individuals, both in terms of success ratios and mean solution times. All groups succeeded in solving the candle light problem compared with individuals success ratios of 0.72 and 0.77 for incentivized and non-incentivized, respectively. The solution time for groups are significantly lower than individuals both for incentivized and non-incentivized (p = and p = 0.08 respectively, Kruskal-Wallis rank tests, one sided test). It is not surprising that a pair of individuals solve an insight problem faster than one individual. A pair solve the problem as soon as one of the members has the insight. So even in the absence of any synergies one would expect teams to outperform individuals in the Candle Problem. In order to check if there is a profound team e ect, we construct pairs from the individual session and compare their performance with the performance in the real group session. The performance of the simulated groups is recorded as follows. First, we draw, without replacement, a random sample of m pairs from the individual sessions. 8 The performance of a simulated pair is given by the best performance of the two individuals comprising a specific pair. If both individuals managed to solve the candle light problem, this constructed group is successful and its solution time is the fastest individual solution time of pair. If only one in the constructed pair managed to solve the probe in the individual session the group is successful and its solution time is the successful partner s time. If neither of the individuals in the c
Related Search
Similar documents
View more...
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks