PhD Student · School of Management, Politecnico di Milano
Giacomo Modugno
I study information markets: how changes in market conditions affect the informativeness of signals and, in turn, how receivers interpret and respond to them. I use the diffusion of generative AI as a particularly salient setting in which these changes can be observed.
My work questions the assumption that the effectiveness of a signal is a stable property of the signal itself, and treats it instead as an outcome that evolves with the conditions of the market in which it is observed. I combine archival analysis of real funding markets, laboratory experiments with neurophysiological measures, and artificial markets populated by generative AI agents.
Research
-
Investor Evaluation and the Complexity of Entrepreneurial Communication: Crowdfunding Success in the Age of Generative AI
Abstract
Textual complexity is a central feature of entrepreneurial communication that helps investors infer the quality of entrepreneurs and ventures. Yet the diffusion of Generative-AI makes this inference more fragile. Once complex language can plausibly be produced with AI support, complexity becomes less clearly attributable to the entrepreneur, even when AI use is neither observed nor disclosed. We argue that, while complexity becomes less diagnostic it remains cognitively demanding to process. For this reason, complex texts are more likely to lead investors to disengage from the opportunity, reducing their willingness to invest. We test this argument in crowdfunding using a mixed-method design. In a laboratory experiment combining EEG, pupillometry, and eye-tracking, investors exposed to high complexity campaign texts are more likely to disengage from the investment opportunity and less likely to invest than those exposed to low complexity texts. In an archival study of more than 11,800 campaigns on Kickstarter, we show that textual complexity increased sharply after ChatGPT’s public release and that, following the introduction of ChatGPT, the association between complexity and campaign success became more negative at high levels of complexity. These findings contribute to research on entrepreneurial communication, investor cognition, and resource acquisition by showing that Generative-AI changes the diagnostic value of entrepreneurial texts and, in turn, the cognitive processes through which investors evaluate ventures.
-
A dime a dozen: the consequences of signal diffusion on perceived signal cost and signal effectiveness
Abstract
Resource providers rely on signals to mitigate information asymmetries in crowdfunding. This study posits that greater signal diffusion reduces their effectiveness by lowering perceived signal acquisition costs. However, this negative effect weakens under high information asymmetry and strengthens under moderate information asymmetry. We test these hypotheses through a mixed method approach that combines archival data analysis with a set of controlled experiments. Using a novel dataset of more than 11,800 crowdfunding campaigns from Kickstarter, we show that human capital signals have diffused over time on the platform. Analyzing the effect of these signals on resource attraction we find that signal effectiveness reduced with their diffusion on the market. However, in project categories characterized by severe information asymmetry the loss of effectiveness is lower. A set of controlled experiments, that manipulate signal diffusion, perceived cost, and information asymmetries, and involve participants with crowdfunding experience test the causality of the relation.
-
Generative AI and the informational value of bounded signals
Abstract
Generative AI is changing how receivers interpret signals used to evaluate workers, founders, and ventures. We examine whether receivers’ beliefs about generative AI change the relative informational value of different realizations of the same bounded ordinal signal. We develop a Bayesian model that characterizes receivers’ beliefs about the extent to which AI increases the likelihood of attaining the maximum signal and about how this increase varies across latent-quality types. We identify the conditions under which a high but nonmaximum realization of the signal induces a higher posterior expectation of latent quality than the maximum. This signal reversal may arise whether AI is believed to disproportionately benefit lower-quality senders, affect all sender types equally, or disproportionately benefit higher-quality senders, depending on the perceived ease of attaining the maximum. We test these predictions in controlled experiments involving hiring and investment decisions. Our findings extend signaling theory by showing that generative AI can alter the relative informational value of bounded signals and identifying the conditions under which the highest observable realization ceases to be the most informative.
-
From Novice to Expert: Mapping the Success Journey of Serial Crowdfunders
Abstract
This study investigates how the determinants of crowdfunding success evolve across different stages of experience of serial crowdfunders, defined as entrepreneurs who repeatedly launch campaigns on crowdfunding platforms. While prior research has identified factors such as pledging conditions, project quality, founder credibility, and social capital as key drivers of success, most studies have treated serial crowdfunders as a homogeneous group and adopted a static perspective. Using a dataset of 46,190 campaigns launched on Kickstarter by 16,242 serial crowdfunders between 2009 and 2024, creators are classified into three groups, Newbies, Explorers, and Veterans, to test how success factors vary across these stages. The results show that the relative importance of signals related to pledging conditions and project quality decreases with experience, while social capital becomes the dominant driver of success for more experienced crowdfunders. The findings suggest that crowdfunding success should be understood as a dynamic process in which the drivers of performance evolve along the entrepreneurial learning trajectory. This study contributes to crowdfunding research by providing evidence that success factors are contingent on experience, and by highlighting the central role of community building in sustaining long-term success for serial crowdfunders.
Awards
- Best PhD Paper Award, International Conference on Alternative Finance Research, Málaga, 2026
- Best Paper Award, 9th Crowdinvesting Symposium, Dresden, 2025
- “Economia Marche” Thesis Award, Fondazione Aristide Merloni, 2025
Selected conferences
- Conference on Field Experiments in Strategy (CFXS), Washington DC, July 2026
- Open and User Innovation Conference, Harvard Business School, July 2026
- Entrepreneurial Finance Conference (ENTFIN), Marseille, July 2026
- Babson College Entrepreneurship Research Conference (BCERC), University of Alabama, June 2026
- AI Plus Management Doctoral Consortium, UCL School of Management, 2026
- International Conference on Alternative Finance Research (ICAFR), Málaga, April 2026
- AiIG Scientific Meeting, Udine, February 2026
- SMS Annual Conference, San Francisco, October 2025
- Conference on Field Experiments in Strategy (CFXS), San Francisco, October 2025
- Crowdinvesting Symposium, Dresden, September 2025
Crowdfunding simulator
An artificial market in which synthetic investors, each with their own budget, risk preference and profile, evaluate real crowdfunding campaigns. Used for research and in the classroom.
Work in progress · coming soon
Teaching
- 2025 / 2026Course instructor, Business Economics and Organization, BSc in Software Engineering, Politecnico di Milano
- 2025 / 2026Supervisor of MBA and MSc theses, POLIMI Graduate School of Management
- 2023 / 2026Crowdfunding campaign design simulation, Politecnico di Milano, Università del Piemonte Orientale, Universidad de la Sabana
Contact
- giacomo.modugno@polimi.it
- School of Management, Politecnico di Milano, Via Lambruschini 4/B, Milan