This project develops a statistical toolbox for the approximation of probability distributions that commonly arise in data analysis. The problem of approximating probabilities arise in many tasks of data science: in Bayesian statistics and its many variants, in classical hypothesis testing with p-values, in likelihood-based methods when the model involves latent variables, in models with intractable likelihoods, in the construction of knockoffs for principled variable selection, for example. State-of-the-art methods for such approximations include Markov chain Monte Carlo (MCMC), where a Markov chain is generated in such a way that it converges to the probability of interest as the length of the sequence goes to infinity. This stands at odds with modern developments in computing hardware, which provide an increasing number of parallel processors, but where each process has a stagnating clock speed. Methods that are amenable to parallel computing must emerge to help scientists in all fields to make the most of their data. The project builds upon a framework called Unbiased Markov chain Monte Carlo (UMCMC), in which accuracy improves arbitrarily with the number of parallel runs. Each run involves the generation of coupled Markov chains for a random time horizon. Part 1 develops UMCMC to realize its potential as a comprehensive basis for probabilistic computation on modern hardware. The project includes theoretical analyses of cost and measures of efficiency, and methodological innovations towards adaptive, efficient, robust and convenient computation. Part 2 contributes to the applicability of UMCMC, by conceptualizing the design of coupled Markov transitions, and considering a number of challenging settings: distributions supported on submanifolds and their application in economics, distributions on graphs and their applications in the fight against malaria, and Bayesian nonparametric models for cell type deconvolution from transcriptomics data.

The goal of this project is to determine the impact of greenwashing on investors’ portfolio allocations and asset prices. Specifically, we want to develop methods to estimate the level of greenwashing used by firms. This will enable us to study which firms use greenwashing most (within which sector, ownership structure etc.), and how it affects portfolio allocations and asset prices. Our tools aim to help regulators reduce transition risk when designing new policies, and investors when selecting assets. We will use state-of-the-art methodologies in Artificial Intelligence, such as Natural Language Processing (NLP). This project includes four Work Packages (WP). We start from the research hypothesis that when greenwashing is uncovered, it is most often reported in the newspapers. News may even be the direct source that investors use to update their beliefs on greenwashing and as a result their portfolios. In the first WP, we will measure the extent to which greenwashing practices are discussed in the financial news, and contaminate the discussion on climate risk. In the second WP, we aim to detect greenwashing at individual firms by comparing the contents and tonality of corporate websites, press releases and financial news articles related to a given firm. The underlying idea is that if a firm makes misleading claims, the contents and tonality used by these three communication channels in their coverage of the claims will be different. The third WP will be an industry-oriented paper that will aim to disseminate results to practitioners. The fourth WP will contain tasks related to project management. This project has a strong interdisciplinary component and is at the confluence of three topical subjects: 1) transition to sustainability as a response to climate change, 2) use of AI to deal with big data, and 3) financial portfolio allocation. It is both original and ambitious: original as it builds on a literature in finance that almost completely ignored the possibility of greenwashing, and ambitious because greenwashing is, by nature, difficult to identify and measure. This project will build a toolbox that researchers will be able to use and further develop freely to advance research on climate risk. We will make all our algorithms publicly available after publication, hoping that they can serve as a basis for further research on greenwashing and more generally on the disclosure of misleading information (e.g., related to social issues) by firms.

Virtual forms of interaction at work are gradually replacing the traditional in-person work style across many knowledge-intensive industries, and this change could signal the beginning of a lasting shift in how work is conducted. Despite obvious benefits, this emerging trend also raises an array of complex challenges, namely related to inclusiveness and participation in virtual work environment. Understanding the complexities of inclusive interactions in these digital environments, and what new challenges may arise, is a crucial issue. As a first step, this proposed research program seeks to focus on female participation and engagement on virtual work settings. In doing so, it aims to investigate: how interactions from and towards women are different compared to their male peers in virtual work environments; What are the drivers of (potential) differences? And finally, how digital platform architecture and design opportunities could be exploited to alleviate the barriers to female participation and interaction in these emerging work settings? To address these questions, a large-scale randomized vignette (survey) experiment with actual professional respondents is proposed. The results of the proposed experiment can enhance our understanding from the new virtual form of work and could generate implications for companies on how to arrange and design their virtual work environments.