A Conversation with Dr. Porsche Fischer
How did the M.A. in Educational Statistics and AI develop, and how did you become involved?
The program builds on Michigan State University’s longstanding strength in Measurement and Quantitative Methods, or MQM. The late William Schmidt helped establish that foundation and later worked with the College of Education to develop a master’s program that could bring rigorous quantitative methods together with emerging applications of artificial intelligence.
I joined MSU initially as a curriculum designer for the new program and now serve as its program coordinator. Much of my work has focused on translating the program’s vision into a curriculum that prepares students not only to understand educational statistics and AI, but also to apply them responsibly to real educational problems.
Educational organizations generate enormous amounts of data. Schools and districts have local assessment and administrative data, while states, agencies and researchers work with large-scale datasets such as M-STEP, SAT, ACT, NAEP, TIMSS and PISA. The challenge is not simply having access to data. It is knowing how to ask meaningful questions of it.
For example: What does the evidence tell us about student learning? Where are inequities emerging? Which students may need additional support? How can we evaluate whether an intervention is working? What information should influence curriculum, policy or resource decisions?
The program is designed to give education professionals the statistical, analytical and AI-related tools to investigate questions like these while keeping educational context and human judgment at the center of the process.
What distinguishes this program from a traditional statistics or data science degree?
The M.A. in Educational Statistics and AI is a 30-credit, fully online program that integrates graduate-level statistics, artificial intelligence, educational measurement, ethics and applied research.
One of its distinctive features is that students learn these methods specifically within educational contexts. Statistical or AI techniques are not taught as isolated technical exercises. Students learn to connect methods to the kinds of questions educators, researchers, analysts and policy professionals actually encounter.
We also expose students to multiple ways of working with data. Coding remains valuable, and students may work with languages and tools used in statistical and data science environments. But contemporary analytics also includes visual, low-code and no-code workflows that can make sophisticated methods more accessible to professionals who do not come from traditional programming backgrounds.
For example, a student might construct a workflow that imports data, examines missing values, prepares the dataset, applies an appropriate statistical or predictive method, evaluates the results and develops visualizations for stakeholders.
The important point is that making the technology easier to use does not remove the need for statistical reasoning. Students still need to understand what a method does, whether it is appropriate for the question and data, what assumptions it makes, how well the results hold up under scrutiny and what conclusions the evidence actually supports.
Technology can accelerate parts of the workflow. It cannot replace professional judgment.
How does the program connect what students learn to real-world practice?
Application is embedded throughout the curriculum and culminates in the Educational Statistics and AI Capstone.
The capstone gives students an opportunity to integrate what they have learned in statistics, AI, data analysis, ethics and educational decision-making around an authentic problem. Rather than completing a traditional thesis, students develop an applied project that demonstrates their ability to move from a meaningful educational question to analysis, interpretation and communication.
That gives students something valuable beyond course credit. They leave with evidence of what they can do, whether they are seeking advancement in their current organization, moving into a new role or expanding the analytical work they already perform.
What was your background before joining the Educational Statistics and AI program?
My professional background has consistently been at the intersection of education, program development, evaluation and data.
I hold a Doctor of Human Services degree with a specialization in Advanced Program Evaluation and Data Analytics. Before joining this program, I served as the program evaluator for STEMM-UP at Michigan State University, a federally funded workforce development initiative.
Before coming to MSU, I managed educational programs at Wayne State University’s Center for Health and Community Impact, and earlier I led education and professional development initiatives for a statewide nonprofit organization.
Across those roles, I repeatedly saw the same challenge: organizations often collect substantial amounts of data but need greater capacity to turn that information into useful evidence for decision-making. That experience strongly influences how I approach curriculum design. The goal is not analysis for its own sake. It is analysis that helps people understand a problem, evaluate possible actions and make better-informed decisions.
Who is this program designed for?
We intentionally designed the program for people with a range of professional backgrounds.
That includes teachers, instructional and district leaders, institutional researchers, program evaluators, analysts, policy professionals, researchers and others working with educational data. It can also serve professionals who want to move into educational data science or AI-related roles while developing specialized knowledge of educational settings.
A strong cohort benefits from having people with different forms of expertise. Someone with many years of experience in a school or educational organization may bring deep contextual knowledge. Another student may bring stronger technical experience with data or emerging technologies. Those perspectives can strengthen one another.
That kind of interdisciplinary exchange is particularly important in AI. Technical knowledge alone is not enough to understand an educational problem, and domain expertise alone may not be enough to evaluate increasingly complex analytical technologies. We want students to learn how to bring those forms of knowledge together.
What principles guided the design of the curriculum?
One of the most important ideas underlying the program is that we cannot prepare students for a rapidly changing field by teaching them to memorize a fixed collection of tools.
Artificial intelligence changes too quickly for that.
Instead, the curriculum emphasizes durable knowledge and adaptable ways of thinking. This aligns closely with Cognitive Flexibility Theory, developed by MSU scholar Rand Spiro and colleagues. Cognitive Flexibility Theory focuses on learning in complex and changing domains, where knowledge has to be reorganized and applied differently as new situations arise.
That is an excellent fit for AI.
Students need foundational knowledge, but they also need to understand how concepts fit together, how to examine a problem from multiple perspectives and how to adapt what they know when technologies or circumstances change.
A particular software platform may change. A model may be replaced. A new analytical technique may become available. What should remain is the student’s ability to evaluate the technology, understand the problem, select appropriate methods, interrogate the results and make a reasoned decision.
That is much more durable than training someone to follow a sequence of steps in one particular tool.
Ethics is a significant component of the program. What does responsible AI mean in the context of educational data?
Responsible AI in education extends far beyond questions about whether students should use generative AI for coursework.
When educational data are involved, we may be working with sensitive information about students, teachers, families, schools and communities. Students therefore need to understand privacy, security, appropriate data use and applicable requirements such as FERPA and, when relevant, COPPA, as well as institutional policies and broader principles of responsible data governance.
But responsible AI is also about the quality of the decisions being made.
Students need to ask where data came from, what populations are represented, what may be missing, whether a model performs differently across groups, whether an output is valid for the intended use and what consequences could result from acting on it.
Bias is one important part of that conversation, but it is not the only one. We also examine issues such as fairness, transparency, explainability, accountability, data provenance, privacy, appropriate human oversight and unintended consequences.
Students learn that an AI system producing an answer does not make the answer authoritative. The output has to be examined as evidence. It needs to be validated, interpreted in context and considered alongside other relevant information.
Comparing results across tools or models can sometimes be useful as one form of triangulation, but agreement between AI systems is not the same as establishing validity. Ultimately, responsible use requires people who know how to question the technology as well as use it.
AI is changing extremely quickly. How can a graduate program keep pace?
That is one of the central design challenges of any AI-related program.
Our response is not to build the curriculum around the assumption that the tools we use today will still be dominant several years from now. The courses establish durable concepts and competencies, while individual learning activities, applications, examples and technologies can evolve as the field changes.
We have already seen how quickly that can happen. Tools considered appropriate during the early stages of course development can change substantially or be replaced before the course is offered. That reinforces why adaptability has to be designed into the curriculum from the beginning.
Students should graduate knowing more than how to operate today's technology. They should understand enough about statistics, AI, educational data, ethics and evidence to evaluate the next technology they encounter.
That is where Cognitive Flexibility Theory becomes especially important. We are trying to develop learners who can reconstruct and apply their knowledge as the environment changes rather than becoming dependent on a particular tool or workflow.
What do you ultimately hope students take away from the program?
I want graduates to be able to approach educational problems with both analytical capability and informed judgment.
They should be able to work with complex data, use contemporary statistical and AI methods, communicate findings clearly to different audiences and understand the ethical responsibilities that accompany those capabilities.
Just as important, they should know when to question a result, when additional evidence is needed and when a decision requires human expertise that should not be delegated to an automated system.
The technology will continue to change. The lasting capability is knowing how to learn, evaluate, adapt and make responsible decisions as it does.
The fully online M.A. in Educational Statistics and AI consists of ten three-credit courses, including a culminating applied capstone. The program’s first cohort will begin in Spring 2027, and applications are accepted on a rolling basis.
Learn more about the M.A. in Educational Statistics and AI, attend an upcoming virtual information session, request information or apply through Michigan State University.