Connor Douglas

Connor Douglas, Joel Persson, Foster Provost

Off-policy evaluation (OPE) estimates the value of a target treatment policy (e.g., a recommender system) using data collected by a different logging policy. It enables high-stakes experimentation without live deployment, yet in practice accuracy depends heavily on the logging policy used to collect data for computing the estimate. We study how to design logging policies that minimize OPE error for given target policies. We characterize a fundamental reward-coverage tradeoff: concentrating probability mass on high-reward actions reduces variance but risks missing signal on actions the target policy may take. We propose a unifying framework for logging policy design and derive optimal policies in canonical informational regimes where the target policy and reward distribution are (i) known, (ii) unknown, and (iii) partially known through priors or noisy estimates at logging time. Our results provide actionable guidance for firms choosing among multiple candidate recommendation systems. We demonstrate the importance of treatment selection when gathering data for OPE, and describe theoretically optimal approaches when this is a firm's primary objective. We also distill practical design principles for selecting logging policies when operational constraints prevent implementing the theoretical optimum.

Connor Douglas, Foster Provost, Arun Sundararajan

Algorithmic agents are used in a variety of competitive decision settings, notably in making pricing decisions in contexts that range from online retail to residential home rentals. Business managers, algorithm designers, legal scholars, and regulators alike are all starting to consider the ramifications of "algorithmic collusion." We study the emergent behavior of multi-armed bandit machine learning algorithms used in situations where agents are competing, but they have no information about the strategic interaction they are engaged in. Using a general-form repeated Prisoner's Dilemma game, agents engage in online learning with no prior model of game structure and no knowledge of competitors' states or actions (e.g., no observation of competing prices). We show that these context-free bandits, with no knowledge of opponents' choices or outcomes, still will consistently learn collusive behavior - what we call "naive collusion." We primarily study this system through an analytical model and examine perturbations to the model through simulations. Our findings have several notable implications for regulators. First, calls to limit algorithms from conditioning on competitors' prices are insufficient to prevent algorithmic collusion. This is a direct result of collusion arising even in the naive setting. Second, symmetry in algorithms can increase collusion potential. This highlights a new, simple mechanism for "hub-and-spoke" algorithmic collusion. A central distributor need not imbue its algorithm with supra-competitive tendencies for apparent collusion to arise; it can simply arise by using certain (common) machine learning algorithms. Finally, we highlight that collusive outcomes depend starkly on the specific algorithm being used, and we highlight market and algorithmic conditions under which it will be unknown a priori whether collusion occurs.

Connor Douglas, Utkucan Balci, Joseph Aylett-Bullock

In this paper, we propose Precision-Informed Semantic Modeling (PRISM), a structured topic modeling framework combining the benefits of rich representations captured by LLMs with the low cost and interpretability of latent semantic clustering methods. PRISM fine-tunes a sentence encoding model using a sparse set of LLM- provided labels on samples drawn from some corpus of interest. We segment this embedding space with thresholded clustering, yielding clusters that separate closely related topics within some narrow domain. Across multiple corpora, PRISM improves topic separability over state-of-the-art local topic models and even over clustering on large, frontier embedding models while requiring only a small number of LLM queries to train. This work contributes to several research streams by providing (i) a student-teacher pipeline to distill sparse LLM supervision into a lightweight model for topic discovery; (ii) an analysis of the efficacy of sampling strategies to improve local geometry for cluster separability; and (iii) an effective approach for web-scale text analysis, enabling researchers and practitioners to track nuanced claims and subtopics online with an interpretable, locally deployable framework.

Connor Douglas, Everett Witt, Mia Bendy et al.

The field of quantitative analytics has transformed the world of sports over the last decade. To date, these analytic approaches are statistical at their core, characterizing what is and what was, while using this information to drive decisions about what to do in the future. However, as we often view team sports, such as soccer, hockey, and baseball, as pairwise win-lose encounters, it seems natural to model these as zero-sum games. We propose such a model for one important class of sports encounters: a baseball at-bat, which is a matchup between a pitcher and a batter. Specifically, we propose a novel model of this encounter as a zero-sum stochastic game, in which the goal of the batter is to get on base, an outcome the pitcher aims to prevent. The value of this game is the on-base percentage (i.e., the probability that the batter gets on base). In principle, this stochastic game can be solved using classical approaches. The main technical challenges lie in predicting the distribution of pitch locations as a function of pitcher intention, predicting the distribution of outcomes if the batter decides to swing at a pitch, and characterizing the level of patience of a particular batter. We address these challenges by proposing novel pitcher and batter representations as well as a novel deep neural network architecture for outcome prediction. Our experiments using Kaggle data from the 2015 to 2018 Major League Baseball seasons demonstrate the efficacy of the proposed approach.