Harnessing the Power of the Crowd to Increase Capacity for Data Science in the Social Sector (1606.07781v1)

Abstract: We present three case studies of organizations using a data science competition to answer a pressing question. The first is in education where a nonprofit that creates smart school budgets wanted to automatically tag budget line items. The second is in public health, where a low-cost, nonprofit women's health care provider wanted to understand the effect of demographic and behavioral questions on predicting which services a woman would need. The third and final example is in government innovation: using online restaurant reviews from Yelp, competitors built models to forecast which restaurants were most likely to have hygiene violations when visited by health inspectors. Finally, we reflect on the unique benefits of the open, public competition model.

• The paper demonstrates how crowdsourcing competitions effectively address data science challenges in the social sector through innovative applications.
• It showcases three case studies where machine learning techniques achieved over 90% accuracy and up to 50% increased operational productivity.
• The study highlights the role of open innovation in democratizing advanced analytics and bridging skill gaps in public service domains.

Harnessing the Power of the Crowd to Increase Capacity for Data Science in the Social Sector

This paper presents a structured examination of the deployment of crowdsourcing in data science competitions to bolster the operational capacities of social sector organizations. The authors, affiliated with DrivenData, offer a compelling narrative through three distinct case studies: "Box-plots for Education," "Countable Care," and "Keeping it Fresh." Each case elucidates a unique application of data science and machine learning to solve pressing issues across education, healthcare, and public administration.

Case Studies Overview

1. Box-plots for Education: This case centers on Education Resource Strategies (ERS), a non-profit aiming to assist public school districts in optimizing resource allocation. The challenge involved automating the categorization of school budget line-items, a process traditionally reliant on manual labor. The competition's outcome was a machine learning algorithm employing logistic regression, coupled with robust feature engineering, such as n-grams and tf-idf, achieving over 90% accuracy. This innovation yielded significant productivity gains, saving approximately 1,000 hours of labor per employee annually.
2. Countable Care: This paper was orchestrated in collaboration with Planned Parenthood to predict women's health care service needs using demographic and behavioral data from the CDC's NSFG. Competitors faced the challenge of extracting useful insights from incomplete survey data due to branching logic. The competition resulted in a set of ensemble models that outperformed conventional approaches, providing valuable predictive insights into healthcare patterns and delivering the outcomes to the Guttmacher Institute for further application.
3. Keeping it Fresh: Partnering with the City of Boston and Yelp, this case aimed to enhance public health inspection efficiency by predicting restaurant hygiene violations using Yelp review data. The challenge involved leveraging social media insights to forecast potential health risks, optimizing inspection resource allocation. The top-performing models demonstrated a potential 30% to 50% increase in inspection productivity, with plans for field integration and testing by the City of Boston.

Implications and Future Directions

The evidence from these case studies suggests substantial potential for employing open innovation and crowdsourcing mechanisms to improve the efficiency and effectiveness of social sector functionalities. This method democratizes access to advanced data science capabilities, offering non-commercial entities the opportunity to utilize cutting-edge algorithms without prohibitive costs. The competitive aspect fosters creativity and a diversity of approaches, culminating in solutions that might be elusive to any single entity.

The paper implicitly suggests that integrating community-driven data science initiatives could address skill shortages in the social sector. Moreover, these competitions stimulate interest among data scientists towards social impact challenges, creating a valuable bridge between voluntary expertise and social imperatives.

In the broader context of AI development, the success demonstrated in these case studies indicates that similar strategies could be employed across various domains where resource constraints limit the capacity to implement sophisticated data-driven strategies. The momentum from such initiatives could drive further advancements in AI applications, particularly in optimizing complex, real-world systems in resource-constrained environments.

Ultimately, the case studies encapsulate the strategic value of harnessing crowd intelligence to resolve large-scale societal challenges, fostering a symbiotic relationship between data science professionals and the social sector, with promising implications for future developments in AI and machine learning.