Research Pillars

Research activities in 5 interconnected “pillars”

MIDAS focuses on five pillars within its overall effort to support data science and Artificial Intelligence (AI) research. Across these pillars, MIDAS promotes responsible research practices, supports the adoption of new data types and methodologies, and facilitates convergence science. The pillars will change over time: current pillars will be retired or shift focus, and new pillars established. Furthermore, these pillars are interconnected where possible, to maximize the synergies between them and leverage the interdisciplinary MIDAS community and the collaborative spirit U-M is known for.

In addition to supporting research within the pillars, MIDAS fosters connections and collaborations in data science and AI for the entire University, maintains a large umbrella covering all relevant research areas, and supports an inclusive community of researchers. MIDAS also continues to promote cutting-edge methodological development as the foundation for data science and AI.

For more information or to participate in any of the activities, contact: midas-contact@umich.edu

Responsible Research Pillar: Enhancing Scientific and Societal Impact

MIDAS collaborates with our researchers to develop foundational principles, methodology and tools, and deploy such tools for the ethical use of data and algorithms and to improve the reproducibility and replicability of scientific findings. Current activities include:

Responsible Data Science and AI

Overview: As data science and AI become a major force in science and in society, increasingly complex analytical pipelines working with poorly understood data pose significant issues of bias, inclusion and fairness. MIDAS is mobilizing our researchers to promote ethical data science and AI. Our approaches include

raising awareness through research discussions and public events, and enabling the development and dissemination of technical solutions through collaboration with our faculty researchers. An example of such events is the annual Future Leaders Summit. An example of technical solution development is FIDES (Framework for Integrative Data Equity Systems).

Who will benefit: All U-M researchers who conduct related research and those who want to examine the issues of bias, inclusion and fairness in their research.

Coordinators: H. V. Jagadish (Director, MIDAS | Professor, Computer Science and Engineering) and Jing Liu (Executive Director, MIDAS)

Reproducible Research

Overview: MIDAS seeks to establish best practices among the community by complementing technical resources with reproducible methodologies and processes. We seek to build upon the research best practices and tools that our researchers develop to make data-intensive research more reproducible, build a central resource including a collection of methods and tools and a showcase (demonstration) and develop training activities.

Who will benefit: U-M and external researchers who want to make their research methodologies accessible and reproducible will find our resources useful. Researchers who develop tools and processes for reproducible research could amplify their voice through our collaboration.

Coordinators: Jing Liu (Executive Director, MIDAS) and Daniel Alexander (Reproducible Research Specialist)

Data Feminism

Overview: The term “data feminism” was coined by Catherine D’Ignazio and Lauren Klein as the title of their 2020 book, which provides a framework for data scientists to address issues in data justice, equity and fairness through feminism principles and approaches. D’Ignazio and Klein’s keynote speech at the MIDAS 2020 Symposium caused a sensation on campus and many of our researchers responded to the call to action. MIDAS is working with other units to develop research on data justice based on feminism principles, as well as events to gather a critical mass around this theme.

Who will benefit: Researchers whose work is related to data justice and who are interested in seeking collaboration.

Coordinator: H.V. Jagadish (Director, MIDAS | Professor, Computer Science and Engineering) and Jing Liu (Executive Director, MIDAS)

Data For Social Good

Overview: MIDAS researchers and students carry out projects to support the data strategy of government and community partners, including the City of Detroit, the Native American tribal nations and other organizations in Michigan. MIDAS coordinates and defines such projects together with our partners, and connects cutting-edge research and ethical data science approaches for positive societal impact.

Who will benefit: Researchers who would like to increase the impact of their research; trainees who would like to gain real-world experience; government and community partners.

Coordinator: Jing Liu (Executive Director, MIDAS)

Data and AI in Society community forum

Overview: Humans constantly grapple with how new technological advances fit into our moral and ethical framework. As the intensive data use and AI tools become essential in almost every sector, every organization in our society, and every aspect of our daily lives, we all need to think about how we adapt to the co-existence of human and AI, and how these new technologies can support, instead of harm, our values. MIDAS convenes these forums to discuss emerging issues such as AI tools in the classroom and in creative endeavors, and catalyze stronger connections between research and policy.

Who Will Benefit: All data science and AI researchers and users of data and AI products who wonder how such technologies are shaping our work, our lives and our society.

Coordinator: Josh Pasek (Associate Professor, Communications and Media)

Data Pillar: Measuring and Improving Society

Societal transformations have complicated traditional survey methods for data collection, while a plethora of new data sources are creating opportunities to measure human behaviors and the human social condition. MIDAS supports the development and use of data science and AI methods to better understand society through new data types such as text, video, sensor and digital trace data. Current activities include:

Unstructured Data for Social Science

Overview: MIDAS supports the campus community to develop this research theme with a two-pronged approach: research incubation and training. To enable innovative research, MIDAS collaborates with the AI Lab and the Institute for Social Research to offer research connection meetings to connect experts for unstructured data (text, image, video, etc) and domain researchers who plan to address significant research questions using such data, showcase new datasets and plan for research projects and grants. For training, MIDAS organized a series of tutorials designed to introduce Natural Language Processing (NLP) to domain researchers. We are developing training for generative AI at this moment.

Who Will Benefit: Researchers who want to learn the fundamentals of research with unstructured data, researchers who have such data and want to connect with methodology experts, and methodology experts who want to seek domain collaborators to significant research questions.

Coordinators: Beth Uberseder (Research Manager, MIDAS)

Supporting the development of new data and their access

Overview: MIDAS supports and collaborates with our faculty and campus units who develop new data sources and data infrastructure for social science research, and enable the wide adoption of such new resources. To name just a few examples: 1) The Research Data Ecosystem, a major effort of ICPSR with NSF funding to modernize the existing software platform to increase the ability of researchers to safely and securely access, connect, store, and manipulate data. MIDAS is a collaborator for the grant application and the project implementations. 2) MIDAS provided initial funding and data access for Libby Hemphill (faculty member in the School of Information and ICPSR) to develop the Social Media Archive. 3) MIDAS pilot funding for faculty effort to develop new data for social science (such as a large scale data on romantic relationships and digitizing the G.I. Bill record data).

Who will benefit: Researchers and units that develop new data sources and infrastructure.

Coordinators: Jing Liu (Executive Director, MIDAS) and Beth Uberseder (Research Manager, MIDAS)

AI Pillar: AI for Science and Engineering

AI methods are revolutionizing research through being powerful tools for many steps of the research workflow, and through automating and accelerating the entire research workflow. We catalyze creative and transformative applications of AI with the potential to lead to major scientific breakthroughs; and enable a broader U-M research community to adopt AI in imagining, planning, executing, and supporting research applications across a range of science and engineering domains. Current activities include the following:

AI in Science and Engineering postdoc program

Overview: This campus-wide program provides outstanding early-career researchers with intensive training and research experience as they ready themselves for independent research in academia and other sectors. As one of the sites in a new global network of postdoc programs focusing on AI in Science and Engineering, we support postdocs who apply AI methodology to address significant research questions. AI is defined broadly to include machine learning, robotics, Bayesian inference, and simulation. Science and engineering includes mathematical sciences, physical sciences, earth and environmental sciences, basic biological sciences, and engineering.

Who Will Benefit: Postdocs, their faculty mentors, and research collaborators.

Coordinators: H.V. Jagadish (Director, MIDAS | Professor, Computer Science and Engineering), Jing Liu (Executive Director, MIDAS), and William Currie (Professor and Associate Dean for Research and Engagement, SEAS)

Generative AI for research

Overview: Generative AI, such as ChatGPT and Bard, generates novel texts, images and other contents by learning from extensive datasets using machine learning algorithms. By leveraging their ability to learn patterns from existing data, generative AI models can be a valuable component in various research domains, aiding researchers in tasks such as forming hypotheses, building databases, providing experimental design and carrying out analysis, and even writing scientific papers. MIDAS coordinates research discussions, training activities and builds collaboration to enable the use of generative AI to accelerate research and enable new research. The first faculty workshop is offered in July, 2023.

Who Will Benefit: All researchers on campus who would like to explore generative AI methods in research.

Coordinator: Jing Liu (Executive Director, MIDAS)

AI in Science and Engineering for the campus

Overview: Using the postdoc training program as the core, we are developing research and training activities for campus researchers and trainees, to connect AI methodologists and domain researchers, to enable AI skills training for researchers, to build an AI in Science and Engineering research community. The first AI in Science and Engineering summer academy is offered in 2023.

Who Will Benefit: All researchers on campus who would like to apply AI methods to science and engineering research.

Coordinators: Jing Liu (Executive Director, MIDAS) and Ken Reid (Data Scientist, MIDAS)

AI-Driven Research Workflow

Overview: Significant advancements in scientific computing, AI, and the hardware and software research environment are enabling researchers to develop AI-driven research workflows (ARWs): AI, including Generative AI, as components for data processing and analytics, and as tools to design and monitor experiments. Institutions and researchers are capitalizing on this opportunity to significantly accelerate research. MIDAS has organized a mini-symposium and faculty discussions around this theme, and is now building collaboration among campus researchers to enable ARW development and adoption.

Who Will Benefit: All researchers on campus who would like to explore how ARWs could facilitate their research and researchers who are interested in developing ARWs.

Coordinator: Jing Liu (Executive Director, MIDAS)

Emerging Pillar: Cultivating New Strengths

We support team activities in strategic research areas that have the potential to grow into pillars. The focus is on areas that are, or are expected to be, national priorities and / or U-M strengths, and can be significantly boosted with Data Science and AI. Naturally, activities in this pillar are exploratory, and a strong focus will emerge with time. Current activities include the following:

Data Science for Environmental Research

Overview: As research on the environment, climate change and sustainability becomes a national priority, an increasing number of U-M environmental scientists are embracing data science and AI methods. MIDAS has been building collaboration and support for these researchers and our concerted effort lays the foundation for our next pillar.

MIDAS offers a summer academy for environmental scientists to help U-M faculty and research scientists adopt data science and AI techniques to environmental science research and integrate data science into their grant applications. The program also helps foster a U-M research community that will advance the application of data science and AI to research that encompasses environmental, climate, and earth sciences; as well as ecology. We also offer follow-up sessions that focus on developing data science components for grant proposals.

In addition, our Propelling Original Data Science (PODS) pilot funding program has been supporting a number of environmental research projects that use cutting-edge data science and AI methods, including:

Improving the efficiency of energy grids
Improving urban water quality
Better machine learning models to monitor air quality
Supporting decision making for navigating vital waterways in winter
Identifying communities vulnerable to climate change
Developing data systems of fish communities in the Great Lakes
Detecting illicit wildlife trading

Who Will Benefit: All environmental scientists who want to explore data science and AI methods for their research, and methodologists who seek to collaborate with environmental scientists.

Coordinator: Beth Uberseder (Research Manager, MIDAS) and Ken Reid (Data Scientist, MIDAS)

Analytics Pillar: Transforming Health Interventions

Health intervention research and implementation is one of the biggest users, as well as inspirations, of cutting-edge data science and AI methods. MIDAS collaborates with campus partners to enable the adoption of cutting-edge analytics and modeling of complex data to boost U-M’s biomedical and healthcare research. Current activities include the following:

Sequential Decision Making for Health Intervention

Overview: MIDAS enables the application of Sequential Decision Making (SDM) to health intervention and other related research through a methodology working group and training for researchers.

The working group aims to develop research grants to both advance cutting-edge SDM methodologies beyond micro-randomized trials / just-in-time adaptive interventions, and to enable the wide adoption of such methodologies. Through a series of meetings, the working group is identifying a) current and developing SDM methodologies that could be enhanced / extended by U-M researchers, b) identifying healthcare challenges that could be mitigated with the application of these methods, and c) identifying challenges in non-health care domain that could benefit from the same methods (such as industrial engineering).

A workshop, planned for early 2024, will focus on methods for optimizing adaptive interventions. This workshop will produce researchers with a strong grasp of the core components of adaptive interventions and available experimental tools for optimizing them, and produce and disseminate via a durable website a set of recommendations on how to teach / interpret the curriculum on optimizing adaptive interventions.

Who Will Benefit: Both SDM methodologists and researchers who may adopt these methods in their research.

Coordinator: Beth Uberseder (Research Manager, MIDAS)

NSF IUCRC : Center for Data-Driven Drug Development and Treatment Assessment (DATA)

Overview: MIDAS researchers and their U-M collaborators and industry and healthcare partners have established an NSF-funded Center for Data-Driven Drug Development and Treatment Assessment (DATA) for pre-competitive research in drug design, drug treatment assessment, drug repositioning, patient phenotyping, and quantitative pharmacovigilance using novel machine learning (ML) and Artificial Intelligence (AI) techniques. The external partners include healthcare systems, non-profits, technology companies and pharmaceutical manufacturers, ranging in size from startups to multinational organizations.

Who Will Benefit: The Center will be a hub to identify and address the complex and time-consuming nature of drug design and treatment assessment, creating solutions that not only reduce the cost associated with the drug design process, but also those that assess, monitor, and optimize treatment and overall patient health.

Coordinator: Kayvan Najarian (PI, DATA; Associate Director, MIDAS; Professor, Computational Medicine and Bioinformatics) and Ivana Tullett (Managing Director, DATA)

Biomedical Data Science Summer Academy

Overview: The annual summer academy is open to all U-M researchers and trainees, as well as biomedical scientists from the public and private sectors. We especially welcome faculty members who want to grasp the basic concepts and methods of data science, so that they can start building a data science component in their research, and work more effectively with their data science collaborators. Furthermore, we help like-minded researchers get to know each other through this academy and develop collaboration. Participants are introduced to supervised and unsupervised machine learning, including deep learning; determine which data science/artificial intelligence techniques are appropriate for research application; and hear use cases of data science and AI methods in biomedical research. We also offer follow-up sessions that focus on developing data science components for grant proposals.

Who Will Benefit: All biomedical scientists who are interested in learning about incorporating data science into their research, but the content is geared towards junior faculty members and those from the public and private sectors.

Coordinator: Kayvan Najarian (PI, DATA; Associate Director, MIDAS; Professor, Computational Medicine and Bioinformatics) and Ken Reid (Data Scientist, MIDAS)