Advancing the frontiers of ambiguity aversion theory and its investment applications
You deserve frameworks built on evidence, not elegance alone. Our research programme publishes working papers, replication datasets, and computational tools that let you verify every claim before it enters your decision process. We treat ambiguity as a first-class problem, not an afterthought.
Open replication
Every published result ships with data and code you can run yourself
Peer scrutiny
Working papers circulate through a network of sceptics before they reach you
Live testing
Models face historical market episodes where elegant theories typically break
Featured research on recursive ambiguity, learning under ambiguity, and robust dynamic programming
Recursive ambiguity models address what happens when you face uncertainty today that reshapes the uncertainty you will face tomorrow. Standard single-period frameworks ignore this compounding, producing recommendations that collapse when decisions link sequentially. Our working papers derive the conditions under which recursive multiple-prior preferences remain dynamically consistent, and identify the parametric restrictions that make computation tractable without trivialising the problem.
Learning under ambiguity examines how you update beliefs when you do not fully trust the data-generating process itself. Bayesian updating assumes you know the likelihood function; our research relaxes this, studying how robust learning rules perform when the model class is misspecified. We characterise the long-run behaviour of ambiguity-averse learners and derive the convergence rates that distinguish genuine learning from false confidence.
Robust dynamic programming in portfolio choice builds control policies that perform well across a set of plausible probability distributions rather than optimising against a single estimated model. Our contributions include new algorithms for high-dimensional robust portfolio problems, calibration methods that avoid overfitting to historical accidents, and diagnostic tools that flag when robustness comes at excessive performance cost.
Empirical validation remains the gatekeeper for every theoretical claim. We replicate published results, extend them to new datasets, and publish the failures alongside the successes. Our replication archive includes code, data, and detailed documentation so you can verify, challenge, or extend our findings without contacting us first.
Case studies in model stress
Selected examples demonstrating how ambiguity-aware models performed during periods of market stress, regime change, or policy uncertainty — the moments when standard frameworks typically fail
Tool connections
How our theoretical frameworks connect with existing quantitative toolkits
Stochastic programming
Ambiguity sets derived from our multiple-prior frameworks plug directly into scenario trees, replacing point estimates with distributional robustness constraints that preserve computational tractability.
Bayesian methods
Our learning-under-ambiguity results extend standard Bayesian updating to settings where the likelihood itself is disputed, producing posterior sets rather than single posterior distributions.
Machine learning
Robust portfolio rules provide regularisation principles that prevent overfitting in high-dimensional prediction settings, with explicit ambiguity parameters that cross-validate on out-of-sample uncertainty.
Risk management platforms
Simulation engines
Monte Carlo scenarios from our robust control problems feed directly into proprietary and open-source simulation platforms, generating stress tests that probe model fragility rather than confirming model comfort.
Data pipelines
Replication datasets and code repositories connect to standard data workflows, enabling automated updates of ambiguity parameters as new market information arrives.
Collaborate with us
Join researchers and practitioners advancing ambiguity-aware decision theory