Mitigating Decision Biases of Marginalized Populations with Algorithm Assessments
Institute or Center: National Institute on Minority Health and Health Disparities (NIMHD)
Project title: Mitigating Decision Biases of Marginalized Populations with Algorithm Assessments
- Expertise in biomedical informatics and/or computer science, data science, data management, data architecture, statistical modeling and analysis, and artificial intelligence and machine learning (ML) methods and applications
- Experience developing strategies for data ingestion, harmonization, and computation across platforms, including techniques such as random forest, XGB, gaussian naïve bayes, etc.
- Knowledge of various programming languages (e.g., Python, Perl, C/C++, SQL, Java) and analytical tools (e.g., SAS, Hadoop, Apache Spark, Hive, R)
- Demonstrated leadership, management and communication skills, and proven ability to build trust and form alliances with diverse stakeholders across different scientific fields, such as health disparities, big data analytics, statistics, and mathematics
About the position: NIMHD seeks a data scientist to develop and test an algorithmic auditing toolkit and a centralized research collaboration platform.
- The toolkit that will allow algorithm developers and operators, as well as individual NIH and external researchers, to check for, and to mitigate, bias before implementing algorithms.
The platform will
- Store publicly available datasets and effective workflows and program codes that mitigate biases in algorithm development and training.
- Provide a forum for data management discussions that can enhance the quality of the data.
- Foster teamwork among algorithm developers and allow the engagement of non-technical health disparity and minority stakeholders.
The Scholar will also manage the toolkit and platform after implementation and act as a liaison with the platform’s user base of collaborating researchers, organizations, institutions, and community leaders.
About the work: The Scholar will have a unique opportunity to have an integral role in supporting NIMHD’s mission to reduce and encourage elimination of health disparities in our society. By building an algorithmic bias auditing tool and online research collaboration platform, the Scholar will help catalyze meaningful research on how to mitigate algorithmic bias, thus contributing to improving the health outcomes of marginalized populations.
Datasets involved: Platform users will be able to import their own algorithms and datasets into the testing environment to test algorithm performance and receive recommendations from the system in terms of possible strategies to mitigate any identified biases. The platform will store publicly available datasets, as well as effective workflows and program codes that mitigate biases in algorithm development and training, with the goal of not perpetuating health disparities or causing undue harm to marginalized populations.
Why this project matters: This project focuses on mitigating and preventing implicit/explicit biases and unintended consequences associated with the use of automated decision support systems in healthcare. The research collaboration platform has the potential to break down silos related to generating, analyzing, and sharing research data between developers and impacted communities. By broadening the set of stakeholders able to participate in ML fairness efforts, the platform will help include affected communities in the conversation to mitigate biases and will help improve the health outcomes of marginalized populations.
Work Location: Bethesda, MD
Work environment: The Scholar will work in an intellectually stimulating environment in close collaboration with content-expert staff within NIMHD. The Scholar will act as a liaison with collaborating external researchers, organizations, institutions, and community leaders and will report to the senior advisor for data science, management and analyses with technical oversight provided by senior bioinformatics specialists. The Scholar will also work with a team of intramural biomedical informatics and algorithmic auditing experts across the NIH.
To apply to this or other DATA Scholar positions, please see instructions here: datascience.nih.gov/data-scholars-2021.
This page last reviewed on February 22, 2021