Where & When Can Open Collaborative Innovation Thrive

+2

No comments posted yet

Comments

Slide 1

What is “open collaborative innovation”? It includes open source (as in software), but also other forms of open collaboration, such as wikipedia and Creative Commons projects.

Slide 3

Jeremy Bentham, whose preserved corpse clothed and seated, has greeted visitors to the college since 1850. Total = 72,500 (40,000 remaining to be transcribed at start of crowdsourcing) In 50 years, transcribed 4800 manuscripts (33.33 p.a.) Crowdsourcing Sep ‘10-May ‘12 transcribed 3,415 manuscripts (171.75 p.a.) Source: Causer, T., Tonra, J., & Wallace, V. (2012). Transcription maximized; expense minimized? crowdsourcing and editing The Collected Works of Jeremy Bentham. Literary and Linguistic Computing. doi: 10.1093/llc/fqs004

Slide 4

The citation count measured appearances of the two words consecutively, as a term. We found a similar pattern when we count appearance of the two words in the same paragraph (but not consecutively). The count was carried out in January 2010, using the Web of Science database, which provide access to current and retrospective bibliographic information, author abstracted, and cited references from over 10,000 leading journals of science, technology, social sciences, arts and humanities and over 100,000 book-based and journal conference proceedings. It provides access to seven databases: Science Citation Index (SCI), Social Sciences Citation Index (SSCI), Arts & Humanities Citation Index (A&HCI), Index Chemicus, Current Chemical Reactions, Conference Proceedings Citation Index: Science and Conference Proceedings Citation Index: Social Science and Humanities

Slide 10

These collaborations share similarities: participants create goods and services of economic value, exchange and reuse each other’s work, labor purposefully with just loose coordination, and permit anyone to contribute and consume.

Slide 11

Open innovation complements/substitute firm-based, but what affects its performance? Important for understanding design of open innovation systems and encouraging it.

Slide 15

Always Contribute Lots Contribute, but only if others do Always Contribute little

Slide 19

Assume: search model rather than a development model. The development is contained within an individual, who must assemble existing pieces or develop own. For simplicity, we don’t represent the coordination of individual modules of work, as this has been done by Baldwin, C. Y., K. B. Clark. 2006. The Architecture of Participation: Does Code Architecture Mitigate Free Riding in the Open Source Development Model? . Management Science 52(7) 1116-1127.

Slide 21

Note that they interact freely, not in a pre-determined, structured way

Slide 24

1. Direct benefits not necessary 2. Cooperators matter, but decreasingly so 3. Altruists are nice, but even general population does well

Slide 25

1a: Over a range of Reciprocator and free-rider levels, a higher ratio of Cooperators leads to better performance.

Slide 26

Changing population ratio can explain almost the entire variance in performance. This is low homogeneity (different needs), perfectly non-rival (no cost to contribute). Point in the back in previous slide. Note non-linear effect of changing percentage of cooperators – even a few make a big difference in performance The min-max points are the result of the characteristics of the other participants: reciprocators vs. free riders. Non-cooperators are a systematic mix of reciprocators and free-Riders

Slide 27

Again – non linear effect. In this case, adding reciprocators increases performance quickly, removing them deteriorates performance quickly. But the effect becomes weak as cooperators join in because cooperators substitute for reciprocators. Robust to free riders to a point, after which there’s a rapid deterioration.

Slide 28

There’s a systematic explanation for these points. 1% cooperators, many reciprocators=40% performance, but as free riders join in, performance deteriorates only when they are overwhelming majority. Injection of cooperators can raise performance quickly, but sensitive to the presence of free-riders.

Slide 29

1a: Over a range of Reciprocator and free-rider levels, a higher ratio of Cooperators leads to better performance.

Slide 31

Performance = % of goals achieved through exchange

Slide 32

1a: Over a range of Reciprocator and free-rider levels, a higher ratio of Cooperators leads to better performance.

Slide 34

High rivalry is detrimental to performance and we know that high similarity in needs is equally bad. Indeed, when we combine the two, the result is very low performance. An intuitive example is food: it’s a highly rival good, and if we all crave the exact same dish, few people will be satisfied which means that overall performance is low. As we move from similar to dissimilar needs, performance improves linearly. Note that rivalry is still high, but as we saw earlier, rivalry matters less when needs are dissimilar. Again, food is high rival, but if each one of us is after a different dish, we can swap, make more people happy, and overall performance increases. Now, some people assume that rivalry is bad for performance. If that’s true, then a decrease in rivalry should lead to increased performance. Let’s see what happens as we decrease rivalry. Well, not much. Performance inches higher, but the difference is negligible. Why is that? Because when needs are not similar, rivalry matters only little. So increasing or decreasing it (as we did here) has a negligible effect on performance. What is the effect of need similarity on performance? Is it critical? To examine that let’s move towards high need similarity. Well, surprise! As you can see, when rivalry is low, need similarity (or dissimilarity) doesn’t matter much by it self. Performance is still robust even with high need similarity, as long as the good is non-rival. The example here is an desired mp3 file – even if we all go for the same song, the good in non-rival, so all of us can end up happily. Finally, notice the shape of the edge that connect this final point to our starting point. It is non-linear, which means that the effect of rivalry drops quickly. these manipulations (need homogeneity and rivalry) were done with randomly sampling from Kurzban-Hauser space.

Slide 35

High rivalry is detrimental to performance and we know that high similarity in needs is equally bad. When we combine the two, performance is very low, not surprisingly. An intuitive example is food: it’s a highly rival good, and if we all crave the exact same dish, few people will be satisfied which means that overall performance is low. As we move from similar to dissimilar needs, performance improves. Even though rivalry is still high, rivalry matters less when needs are dissimilar. Again, food is high rival, but if each one of us is after a different dish, we can swap and trade, make more people happy, and overall performance increases. Now, some people argued that open collaboration thrives because goods are non-rival. If that were true, then a decrease in rivalry should lead to increased performance. Let’s see what happens as we decrease rivalry. Well, not much. Performance inches higher, but the difference is negligible. Why is that? Because when needs are not similar, rivalry matters only little. So increasing or decreasing it has a negligible effect on performance. What is the effect of need similarity on performance? Is it critical? To examine that let’s move towards high need similarity. Well, surprise! As you can see, when rivalry is low, need similarity (or dissimilarity) doesn’t matter much by it self. Performance is still robust even with high need similarity, as long as the good is non-rival. The example here is an desired song or movie – even if we all go for the same song, the good in non-rival, so all of us can end up happily. these manipulations (need homogeneity and rivalry) were done with randomly sampling from Kurzban-Hauser space.

Slide 37

Even a sliver of cooperators can jumpstart a project. Performance is high even with a sliver of cooperators. Decreasing returns to increasing cooperators. Lerner, J. and J. Tirole 2005 "Economic Perspectives on Open Source." In J. Feller, B. Fitzgerald, S. A. Hissam, and K. R. Lakhani (eds.), Perspectives on Free and Open Source Software: 47-78. Cambridge, Massachusetts: The MIT Press. Mockus, A., R. T. Fielding, and J. D. Herbsleb 2005 "Two Case Studies of Open Source Software Development: Apache and Mozilla." In J. Feller, B. Fitzgerald, S. A. Hissam, and K. R. Lakhani (eds.), Perspectives on Free and Open Source Software: 163-209. Cambridge, Massachusetts: The MIT Press.

Slide 38

Open innovation complements/substitute firm-based, but what affects its performance? Important for understanding design of open innovation systems and encouraging it.

Slide 39

What is “open collaborative innovation”? It includes open source (as in software), but also other forms of open collaboration, such as Wikipedia and Creative Commons projects.

Slide 41

Performance = % of goals achieved through exchange

Slide 1

Where Can Open Collaboration Thrive? A Model of Performance Atlanta Competitive Advantage Conference

Slide 2

Relevant Google queries are 1.2-1.5 billion monthly Wikipedia: First result for 56% of searches Britannica: receives < 0.5% or readers

Slide 4

Academic Interest is Palpable

Slide 5

“Open” Beyond Free Software

Slide 6

“Open” Beyond Free Software 641,870 participants

Slide 7

“Open” Beyond Free Software

Slide 8

Key Points Open collaboration has economic, social impact Its performance remains a puzzle: Why? Where? Define phenomenal boundaries Identify elements that affect performance Agent-based model to investigate performance Open collaboration is robust engine for innovation and production Performs well even in unfavorable conditions Likely to continue expanding into new areas

Slide 9

What is Open Collaboration? not only free software

Slide 10

1 von Krogh & von Hippel 2003 3 Shah 2005 5 von Krogh, Spaeth & Lakhani 2003 2 Lee & Cole 2003 4 Mockus, Fielding & Herbsleb 2005 6 Lakhani & von Hippel 2003 Examples Include

Slide 11

Despite Impact, Performance is a Puzzle Differs from firm-based production, innovation (Lee & Cole 2003; von Hippel & von Krogh 2003) How it survives despite massive free-riding/non-contributing users? When expands beyond software? In which environments can it thrive? How to design open innovation systems? What affects performance?

Slide 12

A General Model of Performance Propose a general model of performance based on: Inductive fieldwork in a non-software setting Deductive agent-based simulation to generate propositions Complements literature on Motivation (Roberts, Hann & Slaughter 2006; Shah 2006; von Hippel & von Krogh 2003) Organization (O'Mahony & Ferraro 2007; von Krogh, Spaeth & Lakhani 2003)

Slide 13

The Model what affects performance? Behavior, Goods, Needs

Slide 14

Goods How rival (subtractive) are the goods? Behavior How cooperative are the agents? Needs How similar are their needs?

Slide 15

Different folks, different strokes – of cooperation Fischbacher, Gächter & Fehr, 2001, Economics Letters Kurzban & Hauser, 2005, Proceedings of the National Academy of Sciences USA Ishii & Kurzban, 2008, Human Nature Rustagi, Engel & Kosfeld, 2010, Science How Cooperative are the Agents? Remaining 4% are inconsistent

Slide 16

How Rival are the Goods? To what extent one’s consumption interfere with another’s

Slide 17

How Diverse are their Needs? Participants’ needs can be more or less similar This affects performance, but how? Similarity benefits performance (e.g., Baldwin, Hienerth & von Hippel, 2006) Diversity benefits performance (e.g., Bonaccorsi & Rossi 2003; Ghosh 1998; von Hippel 2005)

Slide 18

The Method Agents interact & exchange

Slide 19

Behavioral Algorithm Each agent has set of resources & set of needs Searches & interacts to fulfill needs, subject to others’ cooperation and rivalry of goods Cooperates, subject to cooperative type If search fails, makes its own Performance How productive is the collective in turning inputs into outputs? Direct benefits: $$$, learning, reputation Division of labor, hierarchy Selection, sorting, communication, sanctions

Slide 20

Single Agent’s Algorithm

Slide 21

Agents Interact Throughout

Slide 22

Network of Exchange Interactions

Slide 23

What Affects Performance? Cooperation - less than you think

Slide 24

General population Direct benefits are not necessary Cooperators matter – decreasingly Altruists are needed, but small core suffices

Slide 25

How Cooperation Matters? Cooperators improve performance Cooperators exhibit decreasing marginal returns

Slide 26

Reciprocators Matter Greatly Max Min What causes the variance?

Slide 27

Performance Robust with Few Cooperators, Many Reciprocators Max Min Many Reciprocators Few Free-Riders Many Free-Riders Few Reciprocators

Slide 28

Many Reciprocators Few Free-Riders Many Free-Riders Few Reciprocators General population

Slide 29

How Cooperation Matters? Cooperators improve performance Cooperators exhibit decreasing marginal returns Cooperators reduce variance Threshold effect of free riding

Slide 30

What Affects Performance? How rival are goods? How similar are needs?

Slide 31

General population Performs well even with rival goods Similarity benefits performance (Baldwin et al, 2006) Diversity benefits performance (von Hippel, 2005 & others) Rivalry and needs interact – nicely

Slide 32

How Rivalry & Needs Matter? Non-Rivalry increases performance Need Homogeneity decreases performance Rivalry and Need Heterogeneity interact

Slide 33

What does it Mean? Implications

Slide 34

Rivalry & Needs Interact to Affect Performance Rivalry Similarity in Needs Performance

Slide 35

Rivalry & Needs Interact to Affect Performance Rivalry Need Similarity Performance General population

Slide 36

Rivalry–Needs compensatory effect Rivalry has non-linear effect on performance 36

Slide 37

Goods Performance benefits from low rivalry, but... High performance possible even with high rivalry, as long as needs are diverse Cooperation Cooperators help, but small core suffices Performance is high even with random sample of human population, free riders and all Needs Dissimilar needs - an advantage, diversity helps But even similar needs can be satisfied

Slide 38

Propositions about Performance How it survives despite massive free-riding/non-contributing users? Free riders matter only in the extreme In which environments can it succeed? How to design open innovation systems? Near non-rival goods, diversity of participant needs, many cooperators or reciprocators If conditions are less than ideal... Some elements can compensate for others!

Slide 39

Where Can Open Collaboration Thrive?

Slide 40

Tak Tack

Slide 41

Cooperation & Needs Interact

Summary: “Open”, as in open source, is now associated with more than free software. It has expanded to describe other instances of similarly open collaborations, including internet enterprises such as Wikipedia, online forums and communities, and also initiatives in medicine, science, and society at large. These collaborations share similarities: participants create goods and services of economic value, exchange and reuse each other’s work, labor purposefully with just loose coordination, and permit anyone to contribute and consume. Despite the economic and social impact of such open collaborations, their performance remains a puzzle. Here we show that open collaboration is a robust engine for innovation and production, one that performs well even in unfavorable conditions. We review multiple instances of open collaboration to define the phenomenal boundaries. We combine extant theory with recent finding on human cooperation to identify elements that determine the performance of open collaboration: the cooperativeness of participants, heterogeneity of their needs, and the degree to which the goods are rival (subtractable). We construct an agent-based model to investigate performance in various circumstances. As we vary conditions while observing outcomes, we find that open collaboration performs robustly even in seemingly inhospitable conditions: when cooperators are a fraction of participants, free riders are present, goods are rival, or diversity is lacking. The results suggest that open collaboration is likely to continue expanding into new areas. We offer advice for leaders of open collaboration, managers, and policy makers.

Tags: open source performance simulation model software wikipedia community economics psychology

URL: