My research explores how Artificial Intelligence (AI) and Bayesian -and Non-Bayesian- learning reshape entrepreneurship and innovation.
Neoclassical economics assumes decision-makers navigate an exogenous, probabilistic environment—one that exists independently of their actions. But what if the environment is endogenous, actively shaped by agents' decisions? This perspective challenges traditional models and raises a key question: how can probabilistic (Bayesian) thinking and AI, a prediction machine, drive innovation in uncertain conditions?
Through large-scale lab and field experiments, I examine how AI and strategic decision-making intersect to empower entrepreneurs. My work aims at shedding light on how different levels of Bayesian learning and AI can be tools not just for navigating the unknown but for actively reshaping it, offering scholars fresh perspectives on strategy and innovation in the age of intelligent systems.
(Authors' names are listed in alphabetic order)
PUBLICATIONS
Included as Best Paper, AOM 2022, TIM Division
WORKING PAPERS
with Arnaldo Camuffo and Alfonso Gambardella (Under Review at Management Science)
Finalist Best Conference Paper prize, AOM 2023, STR Division
Included as Best Paper, AOM 2023, STR Division
with Annamaria Conti (SSRN working paper)
WORK IN PROGRESS
with Claudia Frosi
A Scientific Approach to Entrepreneurial Decision Making: Large Scale Replication and Extension
(Strategic Management Journal)
This paper runs a large-scale replication of Camuffo et al. (2020), involving 759 firms in four randomized control trials. The larger sample generates novel and more precise insights about the teachability and implications of a scientific approach in entrepreneurship. We observe a positive impact on idea termination and results that are consistent with a non-linear effect on radical pivots, with treated firms running few over no or repeated pivots. We provide a theoretical interpretation of the empirical results: the scientific approach enhances entrepreneurs' efficiency in searching for viable ideas and raises their methodic doubt because, like scientists, they realize that there may be alternative scenarios from the ones that they theorize
Design-Based and Theory-Based Approaches to Strategic Decisions
(Organization Science)
Recent literature highlighted two important dimensions of strategic decisions. On the one hand decision-makers formulate and test theories about future scenarios; on the other hand, they take actions to shape future scenarios. Both dimensions are embedded in what decision makers do. However, this paper develops a unified framework that encompasses these two approaches and disentangles their implications. We provide evidence using a 3-arm randomized control trial conducted in Italy that trained 308 early-stage entrepreneurs randomly allocated to a training that emphasizes the former approach, a training that emphasizes the latter approach, and a control group. We find that both dimensions entail the collection of fewer information to make decisions. However, the design-based dimension implies that entrepreneurs still take actions to change scenarios when they receive unfavorable information, while the theory-based dimension induces entrepreneurs to terminate their projects. The theory-based dimension is associated with greater performance conditional on survival. We conclude that the theory-based dimension is ideal when decision-makers seek high performance, while the design-based dimension ensures that projects earn fair returns and survive in spite of negative information.
Bayesian Entrepreneurship and Unknowns: Experimental Evidence
(Under review Management Science)
We investigate how early-stage entrepreneurs update their beliefs as they become increasingly aware of unknown events—future occurrences they cannot describe. We distinguish among three types: Deep-Bayesian entrepreneurs, who adopt a theory-driven approach; Standard-Bayesian entrepreneurs, who rely on evidence-based updating without theoretical priors (akin to the lean startup approach); and Non-Bayesian entrepreneurs, who eschew probabilistic reasoning and focus on reshaping their environment through action. We conduct two field experiments comparing these three types. Ceteris paribus, we find that Deep-Bayesian entrepreneurs hold higher priors and adjust their beliefs less abruptly than Non- Bayesian entrepreneurs. While they resemble Standard-Bayesian entrepreneurs in priors and belief updating, they exhibit greater flexibility in their updating dynamics. Notably, only Deep-Bayesian entrepreneurs react to unknowns in a manner consistent with Reverse Bayesianism (Karni and Vierø, 2017), expanding the state space to accommodate unknowns while preserving probabilities in the original space. This enables them to integrate unknowns into their strategies without discarding prior theories, mirroring the methodical doubt of scientists when refining their models while preserving core knowledge. Our findings suggest that theoretical rigor enables entrepreneurs to navigate extreme uncer- tainty by incorporating unknowns into their frameworks rather than disrupting them.
The Selective Tailwind Effect of Artificial Intelligence in Entrepreneurship
(available at SSRN)
What role does AI play in entrepreneurial decision-making? We explore this question by exploiting large-scale data on US startups and the random release of Google Analytics 4 (GA4), which introduced AI tools especially beneficial for mobile app developers. Leveraging this shock in a difference-in-differences model, we find that the GA4 release significantly boosted customer acquisition, with the performance gains from AI adoption driven by the upper tail of the treatment effect distribution. These effects are most pronounced for innovative startups led by highly skilled founders. A survey of tech founders and an ad hoc experiment elucidate the mechanism: Entrepreneurs rely on causal theories to extract values from AI-detected data anomalies to generate valuable innovations. Our findings underscore the selective tailwind effect of AI in complementing skilled human capital to generate breakthroughs.
The Devil is in The Tails. Entrepreneurial Uncertainty and Strategic Changes: Evidence from a Field Experiment in India
This study explores the role of theory-based experimentation (TbE) in entrepreneurship, focusing on its impact within contexts of fundamental uncertainty. Conducted over 14 months with early-stage entrepreneurs in Hyderabad, India, during the COVID-19 pandemic, the research delves into the relationship between uncertainty, theory-based decision-making, and entrepreneurial actions such as pivoting. The findings indicate that TbE entrepreneurs demonstrate higher willingness to change theory, registering more strategic changes (pivots) compared to a control group. TbE entrepreneurs exhibit a higher tolerance for uncertainty, valuing both uncertainty in their predictions and the potential for higher rewards. The study suggests that TbE entrepreneurs show a greater openness to uncertain value realizations in their projects. Additionally, the research contributes to entrepreneurship studies by empirically capturing entrepreneurs' perceived uncertainty and substantiating the theoretical debate on the entrepreneurship-uncertainty nexus. It also adds to the growing body of research emphasizing the benefits of theorizing in decision-making processes.