Zebra Medical Vision

Shane  Greenstein; Sarah  Gulick

Case | HBS Case Collection | September 2018

Zebra Medical Vision

Abstract

An Israeli startup founded in 2014, Zebra Medical Vision developed algorithms that produced diagnoses from X-rays, mammograms, and CT-scans. The algorithms used deep learning and digitized radiology scans to create software that could assist doctors in making diagnoses. By July 2018, Zebra had developed seven algorithms to analyze scans for emphysema, liver density, compression fractures, bone density, brain bleeds, breast cancer, and a calcium score used to detect calcified plaque in coronary arteries. For each scan analyzed, Zebra charged hospitals $1. By 2018 Zebra found itself in a race with its competitors to perfect these algorithms, create software tools, distribute the tools to physician partners, and create a market. Zebra already had several partners in the U.S. and Europe who gave feedback on its development. Management had to answer the following questions: What should they do next at Zebra? Should they work on the accuracy of the already developed algorithms or continue to develop many new tools? If they chose to develop new products, which applications should they address? And how should they go to market?

Keywords: radiology; Machine learning; X-ray; CT scan; medical technology; probability; FDA 510(k); diagnosis; Health Testing and Trials; Technology; Competitive Strategy; Product Development; Commercialization; Decision Choices and Conditions; Israel;

Language: English

Format: Print

24 pages

Educators Purchase

Citation:

Greenstein, Shane, and Sarah Gulick. "Zebra Medical Vision." Harvard Business School Case 619-014, September 2018.

Article | MIS Quarterly | September 2018

Do Experts or Crowd-Based Models Produce More Bias? Evidence from Encyclopædia Britannica and Wikipedia

Shane Greenstein and Feng Zhu

Organizations today can use both crowds and experts to produce knowledge. While prior work compares the accuracy of crowd-produced and expert-produced knowledge, we compare bias in these two models in the context of contested knowledge, which involves subjective, unverifiable, or controversial information. Using data from Encyclopædia Britannica, authored by experts, and Wikipedia, an encyclopedia produced by an online community, we compare the slant and bias of pairs of articles on identical topics of U.S. politics. Our slant measure is less (more) than zero when an article leans towards Democratic (Republican) viewpoints, while bias is the absolute value of the slant. We find that Wikipedia articles are more slanted towards Democratic views than are Britannica articles, as well as more biased. The difference in bias between a pair of articles decreases with more revisions. The bias on a per word basis hardly differs between the sources because Wikipedia articles tend to be longer than Britannica articles. These results highlight the pros and cons of each knowledge production model, help identify the scope of the empirical generalization of prior studies comparing the information quality of the two production models, and offer implications for organizations managing crowd-based knowledge production.

Keywords: online community; Collective Intelligence; wisdom of crowds; bias; Wikipedia; Britannica; knowledge production; Knowledge Sharing; Knowledge Dissemination; Prejudice and Bias;

Citation:

Greenstein, Shane, and Feng Zhu. "Do Experts or Crowd-Based Models Produce More Bias? Evidence from Encyclopædia Britannica and Wikipedia." MIS Quarterly 42, no. 3 (September 2018): 945–959. View Details

Cite View Details Find at Harvard

The study assembles new data to construct a census of worldwide web server use across the globe. We document a large concentration of investment in the United States, and a wide dispersion across scores of countries. We find tens of billions of dollars of unmeasured value in the open source servers. The statistical models show the quality of the country’s network and the country’s technical sophistication are associated with more web servers, and the innovative environment also plays a role. We find less of a role for economic development, property rights and the rule of law. The findings stress that policies for local supply of content depend on policies encouraging advanced networking and a technically sophisticated populace. While the findings highlight the danger for misattribution in growth accounting, the statistical model points towards the potential to proxy for unmeasured servers with statistical methods.

WeWork co-founder Miguel McKelvey was concerned about the culture of his rapidly expanding global venture. In particular, he wanted to ensure that WeWork continued to be a great place to work, both because he cared about WeWork's people and because a better work experience was core to the company’s brand. In the summer of 2017, he had made a presentation to the company at an all-hands meeting. In it, he proposed the creation of a culture operations team, and he announced that he would take the title of chief culture officer to lead it. In this case, which describes the founding and growth of WeWork from 2010, we encounter McKelvey planning for his new role with co-founder Adam Neumann and envisioning a “culture OS” that he and Neumann would use to lead the organization internally—and to offer as a “product” to scaled enterprises externally.

Faculty & Research

Zebra Medical Vision

Abstract

Citation:

About the Author

More from the Author