Open data science and algorithm development competitions over a unique avenue for rapid discovery of better computational strategies. We highlight three examples in computational biology and bioinformatics research where the use of competitions has yielded significant performance gains over established algorithms. These include algorithms for antibody clustering, imputing gene expression data, and querying the Connectivity Map (CMap). Performance gains are evaluated quantitatively using realistic, albeit sanitized, data sets. The solutions produced through these competitions are then examined with respect to their utility and the prospects for implementation in the field. We present the decision process and competition design considerations that lead to these successful outcomes as a model for researchers who want to use competitions and non-domain crowds as collaborators to further their research.
This paper presents NASA’s experience using a Center of Excellence (CoE) to scale and sustain an open innovation program as an effective problem-solving tool and includes strategic management recommendations for other organizations based on lessons learned.
This paper defines four phases of implementing an open innovation program: Learn, Pilot, Scale and Sustain. It provides guidance on the time required for each phase and recommendations for how to utilize a CoE to succeed. Recommendations are based upon the experience of NASA’s Human Health and Performance Directorate, and experience at the Laboratory for Innovation Science at Harvard running hundreds of challenges with research and development organizations.
Lessons learned include the importance of grounding innovation initiatives in the business strategy, assessing the portfolio of work to select problems most amenable to solving via crowdsourcing methodology, framing problems that external parties can solve, thinking strategically about early wins, selecting the right platforms, developing criteria for evaluation, and advancing a culture of innovation. Establishing a CoE provides an effective infrastructure to address both technical and cultural issues.
The NASA experience spanned more than seven years from initial learnings about open innovation concepts to the successful scaling and sustaining of an open innovation program; this paper provides recommendations on how to decrease this timeline to three years.
Radiation therapy (RT) is a critical cancer treatment, but the existing radiation oncologist work force does not meet growing global demand. One key physician task in RT planning involves tumor segmentation for targeting, which requires substantial training and is subject to significant interobserver variation.
To determine whether crowd innovation could be used to rapidly produce artificial intelligence (AI) solutions that replicate the accuracy of an expert radiation oncologist in segmenting lung tumors for RT targeting.
We conducted a 10-week, prize-based, online, 3-phase challenge (prizes totaled $55 000). A well-curated data set, including computed tomographic (CT) scans and lung tumor segmentations generated by an expert for clinical care, was used for the contest (CT scans from 461 patients; median 157 images per scan; 77 942 images in total; 8144 images with tumor present). Contestants were provided a training set of 229 CT scans with accompanying expert contours to develop their algorithms and given feedback on their performance throughout the contest, including from the expert clinician.
Main Outcomes and Measures The AI algorithms generated by contestants were automatically scored on an independent data set that was withheld from contestants, and performance ranked using quantitative metrics that evaluated overlap of each algorithm’s automated segmentations with the expert’s segmentations. Performance was further benchmarked against human expert interobserver and intraobserver variation.
A total of 564 contestants from 62 countries registered for this challenge, and 34 (6%) submitted algorithms. The automated segmentations produced by the top 5 AI algorithms, when combined using an ensemble model, had an accuracy (Dice coefficient = 0.79) that was within the benchmark of mean interobserver variation measured between 6 human experts. For phase 1, the top 7 algorithms had average custom segmentation scores (S scores) on the holdout data set ranging from 0.15 to 0.38, and suboptimal performance using relative measures of error. The average S scores for phase 2 increased to 0.53 to 0.57, with a similar improvement in other performance metrics. In phase 3, performance of the top algorithm increased by an additional 9%. Combining the top 5 algorithms from phase 2 and phase 3 using an ensemble model, yielded an additional 9% to 12% improvement in performance with a final S score reaching 0.68.
A combined crowd innovation and AI approach rapidly produced automated algorithms that replicated the skills of a highly trained physician for a critical task in radiation therapy. These AI algorithms could improve cancer care globally by transferring the skills of expert clinicians to under-resourced health care settings.
We study group contests where group sizes are stochastic and unobservable to participants at the time of investment. When the joint distribution of group sizes is symmetric, with expected group size , the symmetric equilibrium aggregate investment is lower than in a symmetric group contest with commonly known fixed group size . A similar result holds for two groups with asymmetric distributions of sizes. For the symmetric case, the reduction in individual and aggregate investment due to group size uncertainty increases with the variance in relative group impacts. When group sizes are independent conditional on a common shock, a stochastic increase in the common shock mitigates the effect of group size uncertainty unless the common and idiosyncratic components of group size are strong complements. Finally, group size uncertainty undermines the robustness of the group size paradox otherwise present in the model.
James “Hondo” Geurts, the Acquisition Executive for U.S. Special Operations Command was in the middle of his Senate confirmation hearing in 2017 to become Assistant Secretary of the Navy for Research, Development and Acquisition. The questions had a common theme: how would Geurts’s experience running an innovative procurement effort for U.S. Special Forces units enable him to change a much larger—and much more rigid—organization like the U.S. Navy? In one of the most secretive parts of the U.S. military, Geurts founded an open platform called SOFWERX to speed the rate of ideas to Navy SEALs, Army Special Forces, and the like. His team even sourced the idea for a hoverboard from a YouTube video. But how should things like SOFWERX and protypes like the EZ-Fly find a place within the Navy writ large?
We experimentally study the effects of sorting and communication in contests between groups of heterogeneous players whose within-group efforts are perfect complements. Contrary to the common wisdom that competitive balance bolsters performance in contests, in this setting theory predicts that aggregate output increases in the variation in abilities between groups, i.e., it is maximized by the most unbalanced sorting of players. However, the data does not support this prediction. In the absence of communication, we find no effect of sorting on aggregate output, while in the presence of within-group communication aggregate output is 33% higher under the balanced sorting as compared to the unbalanced sorting. This reversal of the prediction is in line with the competitive balance heuristic. The results have implications for the design of optimal groups in organizations using relative performance pay.
Scientists typically self-organize into teams, matching with others to collaborate in the production of new knowledge. We present the results of a field experiment conducted at Harvard Medical School to understand the extent to which search costs affect matching among scientific collaborators. We generated exogenous variation in search costs for pairs of potential collaborators by randomly assigning individuals to 90-minute structured information-sharing sessions as part of a grant funding opportunity for biomedical researchers. We estimate that the treatment increases the baseline probability of grant co-application of a given pair of researchers by 75% (increasing the likelihood of a pair collaborating from 0.16 percent to 0.28 percent), with effects higher among those in the same specialization. The findings indicate that matching between scientists is subject to considerable frictions, even in the case of geographically-proximate scientists working in the same institutional context with ample access to common information and funding opportunities.
The purpose of this article is to suggest a (preliminary) taxonomy and research agenda for the topic of “firms, crowds, and innovation” and to provide an introduction to the associated special issue. We specifically discuss how various crowd-related phenomena and practices—for example, crowdsourcing, crowdfunding, user innovation, and peer production—relate to theories of the firm, with particular attention on “sociality” in firms and markets. We first briefly review extant theories of the firm and then discuss three theoretical aspects of sociality related to crowds in the context of strategy, organizations, and innovation: (1) the functions of sociality (sociality as extension of rationality, sociality as sensing and signaling, sociality as matching and identity); (2) the forms of sociality (independent/aggregate and interacting/emergent forms of sociality); and (3) the failures of sociality (misattribution and misapplication). We conclude with an outline of future research directions and introduce the special issue papers and essays.
We report results of a natural field experiment conducted at a medical organization that sought contribution of public goods (i.e., projects for organizational improvement) from its 1200 employees. Offering a prize for winning submissions boosted participation by 85 percent without affecting the quality of the submissions. The effect was consistent across gender and job type. We posit that the allure of a prize, in combination with mission-oriented preferences, drove participation. Using a simple model, we estimate that these preferences explain about a third of the magnitude of the effect. We also find that these results were sensitive to the solicited person’s gender.
Utilizing suggestions from clinicians and administrative staff is associated with process and quality improvement, organizational climate that promotes patient safety, and added capacity for learning. However, realizing improvement through innovative ideas from staff depends on their ability and decision to contribute. We hypothesized that staff perception of whether the organization promotes learning is positively associated with their likelihood to engage in problem solving and speaking up. We conducted our study in a cardiology unit in an academic hospital that hosted an ideation contest that solicited frontline staff to suggest ideas to resolve issues encountered at work. Our primary dependent variable was staff participation in ideation. The independent variables measuring perception of support for learning were collected using the validated 27-item Learning Organization Survey (LOS). To examine the relationships between these variables, we used analysis of variance, logistic regression, and predicted probabilities. We also interviewed 16 contest participants to explain our quantitative results. The study sample consisted of 30% of cardiology unit staff (n=354) that completed the LOS. In total, 72 staff submitted 138 ideas, addressing a range of issues including patient experience, cost of care, workflow, utilization, and access. Figuring out the cost of procedures in the catheterization laboratory and creating a smartphone application that aids patients to navigate through appointments and connect with providers were two of the ideas that won the most number of votes and funding to be implemented in the following year. Participation in ideation was positively associated with staff perception of supportive learning environment. For example, one standard deviation increase in perceived welcome for differences in opinions was associated with a 43% increase in the odds of participating in ideation (OR=1.43, p=0.04) and 55% increase in the odds of suggesting more than one idea (OR=1.55, p=0.09). Experimentation, a practice that supports learning, was negatively associated with ideation (OR=0.36, p=0.02), and leadership that reinforces learning was not associated with ideation. The perception that new ideas are not sufficiently considered or experimented could have motivated staff to participate, as the ideation contest enables experimentation and learning. Interviews with ideation participants revealed that the contest enabled systematic bottom-up contribution to quality improvement, promoted a sense of community, facilitated organizational exchange of ideas, and spread a problem-solving oriented mindset. Enabling frontline staff to feel that their ideas are welcome and that making mistakes is permissible may increase their likelihood to engage in problem solving and speaking up, contributing to organizational improvement.
BACKGROUND: The association of differing genotypes with disease-related phenotypic traits offers great potential to both help identify new therapeutic targets and support stratification of patients who would gain the greatest benefit from specific drug classes. Development of low-cost genotyping and sequencing has made collecting large-scale genotyping data routine in population and therapeutic intervention studies. In addition, a range of new technologies is being used to capture numerous new and complex phenotypic descriptors. As a result, genotype and phenotype datasets have grown exponentially. Genome-wide association studies associate genotypes and phenotypes using methods such as logistic regression. As existing tools for association analysis limit the efficiency by which value can be extracted from increasing volumes of data, there is a pressing need for new software tools that can accelerate association analyses on large genotype-phenotype datasets.
RESULTS: Using open innovation (OI) and contest-based crowdsourcing, the logistic regression analysis in a leading, community-standard genetics software package (PLINK 1.07) was substantially accelerated. OI allowed us to do this in <6 months by providing rapid access to highly skilled programmers with specialized, difficult-to-find skill sets. Through a crowd-based contest a combination of computational, numeric, and algorithmic approaches was identified that accelerated the logistic regression in PLINK 1.07 by 18- to 45-fold. Combining contest-derived logistic regression code with coarse-grained parallelization, multithreading, and associated changes to data initialization code further developed through distributed innovation, we achieved an end-to-end speedup of 591-fold for a data set size of 6678 subjects by 645 863 variants, compared to PLINK 1.07's logistic regression. This represents a reduction in run time from 4.8 hours to 29 seconds. Accelerated logistic regression code developed in this project has been incorporated into the PLINK2 project.
CONCLUSIONS: Using iterative competition-based OI, we have developed a new, faster implementation of logistic regression for genome-wide association studies analysis. We present lessons learned and recommendations on running a successful OI process for bioinformatics.
Tomohiro Ishibashi (Bashi), chief executive officer for B to S, and Julia Foote LeStage, chief innovation officer of Weathernews Inc., were addressing a panel at the HBS Digital Summit on creative uses of big data. They told the summit attendees about how the Sakura (cherry blossoms) Project, where the company asked users in Japan to report about how cherry blossoms were blooming near them day by day, had opened up opportunities for the company's consumer business in Japan. The project ultimately garnered positive publicity and became a foothold to building the company's crowdsourcing weather-forecasting service in Japan. It changed the face of weather forecasting in Japan. Bashi and LeStage wondered whether the experience could be applied to the U.S. market.
Most United States Patent and Trademark Office (USPTO) patent documents contain drawing pages which describe inventions graphically. By convention and by rule, these drawings contain figures and parts that are annotated with numbered labels but not with text. As a result, readers must scan the document to find the description of a given part label. To make progress toward automatic creation of ‘tool-tips’ and hyperlinks from part labels to their associated descriptions, the USPTO hosted a monthlong online competition in which participants developed algorithms to detect figures and diagram part labels. The challenge drew 232 teams of two, of which 70 teams (30 %) submitted solutions. An unusual feature was that each patent was represented by a 300-dpi page scan along with an HTML file containing patent text, allowing integration of text processing and graphics recognition in participant algorithms. The design and performance of the top-5 systems are presented along with a system developed after the competition, illustrating that the winning teams produced near state-of-the-art results under strict time and computation constraints. The first place system used the provided HTML text, obtaining a harmonic mean of recall and precision (F-measure) of 88.57 % for figure region detection, 78.81 % for figure regions with correctly recognized figure titles, and 70.98 % for part label detection and recognition. Data and source code for the top-5 systems are available through the online UCI Machine Learning Repository to support follow-on work by others in the document recognition community.
This paper discusses several challenges in designing field experiments to better understand how organizational and institutional design shapes innovation outcomes and the production of knowledge. We proceed to describe the field experimental research program carried out by our Crowd Innovation Laboratory at Harvard University to clarify how we have attempted to address these research design challenges. This program has simultaneously solved important practical innovation problems for partner organizations, like NASA and Harvard Medical School (HMS), while contributing research advances, particularly in relation to innovation contests and tournaments. We conclude by proceeding to highlight the opportunity for the scholarly community to develop a “science of innovation” that utilized field experiments as means to generate knowledge.
Selecting among alternative projects is a core management task in all innovating organizations. In this paper, we focus on the evaluation of frontier scientific research projects. We argue that the “intellectual distance” between the knowledge embodied in research proposals and an evaluator’s own expertise systematically relates to the evaluations given. To estimate relationships, we designed and executed a grant proposal process at a leading research university in which we randomized the assignment of evaluators and proposals to generate 2,130 evaluator–proposal pairs. We find that evaluators systematically give lower scores to research proposals that are closer to their own areas of expertise and to those that are highly novel. The patterns are consistent with biases associated with boundedly rational evaluation of new ideas. The patterns are inconsistent with intellectual distance simply contributing “noise” or being associated with private interests of evaluators. We discuss implications for policy, managerial intervention, and allocation of resources in the ongoing accumulation of scientific knowledge.
We investigate the factors driving workers’ decisions to generate public goods inside an organization through a randomized solicitation of workplace improvement proposals in a medical center with 1200 employees. We find that pecuniary incentives, such as winning a prize, generate a threefold increase in participation compared to non-pecuniary incentives alone, such as prestige or recognition. Participation is also increased by a solicitation appealing to improving the workplace. However, emphasizing the patient mission of the organization led to countervailing effects on participation. Overall, these results are consistent with workers having multiple underlying motivations to contribute to public goods inside the organization consisting of a combination of pecuniary and altruistic incentives associated with the mission of the organization.
Tournaments are widely used in the economy to organize production and innovation. We study individual data on 2775 contestants in 755 software algorithm development contests with random assignment. The performance response to added contestants varies nonmonotonically across contestants of different abilities, precisely conforming to theoretical predictions. Most participants respond negatively, whereas the highest-skilled contestants respond positively. In counterfactual simulations, we interpret a number of tournament design policies (number of competitors, prize allocation and structure, number of divisions, open entry) and assess their effectiveness in shaping optimal tournament outcomes for a designer.
The last two decades have witnessed an extraordinary growth of new models of managing and organizing the innovation process, which emphasize users over producers. Large parts of the knowledge economy now routinely rely on users, communities, and open innovation approaches to solve important technological and organizational problems. This view of innovation, pioneered by the economist Eric von Hippel, counters the dominant paradigm, which casts the profit-seeking incentives of firms as the main driver of technical change. In a series of influential writings, von Hippel and colleagues found empirical evidence that flatly contradicted the producer-centered model of innovation. Since then, the study of user-driven innovation has continued and expanded, with further empirical exploration of a distributed model of innovation that includes communities and platforms in a variety of contexts and with the development of theory to explain the economic underpinnings of this still emerging paradigm. This volume provides a comprehensive and multidisciplinary view of the field of user and open innovation, reflecting advances in the field over the last several decades.
The contributors—including many colleagues of Eric von Hippel—offer both theoretical and empirical perspectives from such diverse fields as economics, the history of science and technology, law, management, and policy. The empirical contexts for their studies range from household goods to financial services. After discussing the fundamentals of user innovation, the contributors cover communities and innovation; legal aspects of user and community innovation; new roles for user innovators; user interactions with firms; and user innovation in practice, describing experiments, toolkits, and crowdsourcing and crowdfunding.
As of 2013, Havas was the 6th largest global advertising, digital, and communications group in the world. Headquartered in Paris, France, the group was highly decentralized, with semi-independent agencies in more than 100 countries offering a variety of services. The largest unit of Havas was Havas Worldwide, an integrated marketing communications agency headquartered in New York, NY. CEO David Jones was determined to make Havas Worldwide the most future-focused agency in the industry by becoming a leader in digital innovation. The case explores the tensions within the company as David Jones attempts to change the company to compete in an industry undergoing digital transformation. The case uses the example of the acquisition of Victors & Spoils, a crowdsourcing advertising agency, to examine internal reactions.