Evidence · uncertainty · consequence
Research
My research is transdisciplinary because the questions I care about exceed any one field. I combine philosophy, institutional analysis, and statistical modeling to study what scientific methods claim, how they are validated, and what is at stake when we use them.
Research pillars
Values in science & Bayesian practice
I ask whether Bayesianism, understood as an applied practice of scientific inference, can avoid arguments from inductive risk or remain value-free in the relevant sense.
Scientific integrity & institutional design
I study how institutions police plagiarism, p-hacking, and other academic misconduct—and how enforcement systems can protect inquiry without enabling accusations to be weaponized against scholars because of identity or political viewpoint.
Applied modeling in sports & health
I build expected-outcome and matchup-adjusted models in baseball and football, alongside public-health research where selection, spatial structure, and interpretation are central.
Current directions
Bayesianism, values, and inductive risk
I investigate whether Bayesianism, treated as an applied practice of scientific inference rather than only an ideal theory of graded belief, can avoid arguments from inductive risk or remain value-free in the relevant sense. My forthcoming paper analyzes how choices about ambiguous data, priors and hyperpriors, computational diagnostics, and model comparison can predictably shape posterior claims; ongoing work extends that analysis to the broader statistical and institutional practices through which Bayesian results are produced, checked, and used.
Scientific integrity as an institutional problem
I investigate the epistemic effects of how universities and scholarly communities police plagiarism, p-hacking, and other forms of academic misconduct. The project asks how enforcement systems can protect research integrity while avoiding evidentiary and procedural structures that allow accusations to be weaponized against scholars on the basis of identity or political viewpoint. Planned formal work includes probabilistic and agent-based models of reporting, enforcement, and incentives, paired with an institutional-design analysis. The agent-based modeling is a future stage of the project, not a completed result.
Models for player evaluation and availability
My public sports research examines matchup-adjusted player ratings, expected-outcome models, and residualized outcomes: tools for asking what a player contributed after accounting for the physical and competitive context around an observed result. Current examples include FAIR xwOBA in baseball and Opponent-Adjusted Evaluation of NFL Pass Blocking and Pass Rushing Performance in football.
Ongoing non-public research. In separate professional-football work, I use Bayesian models to study how prior injury histories relate to later career availability. The statistical challenges include out-of-sample prediction and partial pooling across sparse injury categories. The data, organization, model details, and findings remain confidential.
High-dimensional and spatial inference
My statistics thesis studies data-driven control selection in spatial epidemiology, combining double-selection LASSO with spatial controls to investigate county-level COVID-19 mortality. More broadly, I am interested in methods that make high-dimensional applied inference more stable without hiding the assumptions needed to interpret it. Read the thesis case study →
Forthcoming publication
Bayesian Practice and the Persistence of Inductive Risk
Maximilian J. Gebauer. Forthcoming in Philosophy of Science.
Treating Bayesianism as actual statistical practice reveals familiar forms of methodological underdetermination: choices about ambiguous data, priors and hyperpriors, computational diagnostics, and model comparison can all predictably shape posterior claims.
Read the accessible overview →Selected talks, presentations, and invited lectures
Approximating Posterior Distributions with Variational Bayes
A methodological talk on scalable Bayesian approximation and the inferential tradeoffs it introduces.
A Dynamic Rating Scheme for NFL Pass Rushers
A sports-modeling talk on evolving performance estimates and validation over time.
Judgment and Decision-making under Uncertainty
A talk connecting formal uncertainty to the decisions people and institutions must make.
Expected Utility Maximization and Environmental Decision-making
A philosophical analysis of expected utility in high-stakes environmental choices.