Main research projects

More details on my Google Scholar page

The Narrative AI Advantage? A Field Experiment on Generative AI-Augmented Evaluations of Early-Stage Innovations (with Jacqueline Lane, Léonard Boussioux et al.) (Awarded 1st place in Wharton People Conference Competition 2025) 

Abstract:
The rise of generative artificial intelligence (AI) is transforming creative problem-solving, necessitating new approaches for evaluating innovative solutions. This study explores how human-AI collaboration can enhance early-stage evaluations, focusing on the interplay between objective criteria, which are quantifiable, and subjective criteria, which rely on personal judgment. We conducted a field experiment with MIT Solve, involving 72 experts and 156 community screeners who evaluated 48 solutions for the 2024 Global Health Equity Challenge. Screeners received assistance from GPT-4, offering recommendations and, in some cases, rationale. We compared a human-only control group with two AI-assisted treatments: a black box AI and a narrative AI with probabilistic explanations justifying its decisions. Our findings show that AI-assisted screeners were 9 percentage points more likely to fail a solution. For objective criteria, there was no significant difference between the black box and narrative AI conditions. However, for subjective criteria, screeners adhered to narrative AI’s recommendations 12 percentage points more often than the black box AI’s. These effects were consistent across both experts and non-experts. Mouse tracking data showed that deeper engagement with AI’s objective failure recommendations led to more overrides of the AI, particularly in the narrative AI condition, reflecting increased scrutiny. Conversely, deeper engagement with AI’s subjective failure recommendations led to greater alignment with AI, particularly in the black box condition. This research underscores the importance of developing AI interaction expertise in creative evaluation processes that combine human judgment with AI insights. While AI can standardize decision-making for objective criteria, human oversight and critical thinking remain indispensable in subjective assessments, where AI should complement, not replace, human judgment. 

Office at Offsite: How Temporary Colocation Shapes Communication in Fully-remote Organizations (with Prithwiraj Choudhury , Sujin Jang  and Victoria Sevcenko)
(Nominated for the EGOS Best Paper Award 2023 and 3rd place award in Wharton People Conference Competition 2025) 

Abstract:
This paper investigates the role of temporary colocation events and demographic diversity in shaping online communication in a fully remote organization. Motivated by the rise in remote work and the resulting increase in diversity within modern organizations, we examine the effectiveness of temporary colocation events, such as company retreats, as a tool to stimulate online interaction across demographic divides. Leveraging proprietary Slack communications data as well as data from an all-company retreat (including who shared taxi rides with whom) from a fully remote organization, we hypothesize and find that while these retreats enhance subsequent online communication, the benefits are less pronounced for pairs of demographically dissimilar employees. However, even brief periods of constrained temporary colocation with a limited number of interaction partners, such as sharing a taxi ride, can help overcome these barriers, increasing online interactions even for demographically dissimilar pairs of employees. This research contributes to the literature on remote work by highlighting the nuanced role of temporary colocation events in fostering communication. It also contributes to the literature on microgeography by highlighting important design features for the microgeography of temporary colocation events that help to maximize their benefits and foster a more inclusive communication culture in remote work environments.

Machine Learning in Healthcare: Mirage or Miracle for Breaking the Costs Dead-lock?(with Dominique Foray)

Abstract:
The ageing population in all developed economies and the limited productivity characterizing the healthcare sector are leading to alarmingly increasing costs. The current rapid advances in machine learn-ing (ML), a subfield of artificial intelligence (AI), offer new automation and prediction capabilities that could, if properly integrated, help address the healthcare costs deadlock. Are ML-driven solutions the ap-propriate ingredient to produce this necessary transformation, or are they condemned to face the same destiny as previous attempts to remodel healthcare delivery? This paper aims at bringing first elements to answer this question by providing both qualitative and quantitative evidence on the development of ML in healthcare and discussing the organizational and institutional conditions for the ML potential to be real-ized. Building on a novel search methodology for publications and patents in ML and on hospital surveys, our results reveal two major observations. On the one hand, while the publication rate in the field has tripled in the last decade, the level of patenting in ML applied to healthcare has so far been relatively low. This result has several potential explanations, such as the early stage of the technology, its rapid growth, and the emergence of new business models based on data accumulation and appropriation rather than patenting. On the other hand, the bulk of firms’ publications are produced by IT firms rather than by com-panies in healthcare. This last observation seems to be driven by the disruptiveness of the new ML tech-nology allowing the entry of new actors in healthcare. The technology producers benefit from their mas-tery of ML and the lack of investment and capabilities among health experts.

The Dual Effect of Intellectual Similarity: The Interplay of Partiality and Expertise in the Evaluation of Technological Problems (with Jacqueline Lane, Zoe Szajnfarber, and Karim Lakhani)

Abstract:
Technological innovations often arise from recombining knowledge from distant domains outside the problem to be solved. Yet, such recombinant innovations tend to be difficult to judge, given that evaluators need to possess knowledge not only of the problem but also knowledge of the distant domains that are being applied to create a novel solution. In this paper, we conceptualize how two dimensions of intellectual similarity—functional and industry—between an evaluator and solver shape evaluations given to technological solutions. We designed an experiment in collaboration with NASA and Freelancer.com to investigate the implications of intellectual similarity on the evaluation of crowdsourced technological solutions. Our research design randomly assigned 374 evaluators to 101 solutions, leading to 3,850 evaluator-solver pairs with a wide array of evaluator and solver expertise and backgrounds. We find that whereas greater functional intellectual similarity leads to more critical (lower) evaluation scores, greater industry overlap leads to more lenient (higher) evaluation scores. In addition, the scores given by the pairs with overlapping functional expertise were more closely aligned with those given by internal experts within NASA. Our findings reveal the competing effects of preference for familiar solutions and the discerning nature of expertise on evaluation outcomes.

Research Directions and Scientific Prizes: Evidence from a Prestigious Research Fellowship (with Sandra Barbosu, Michele Pezzoni, and Fabiana Visentin)

Abstract:
The decisions of scientific funding agencies and award granters greatly influence the direction of scientific research; however, our understanding of the factors influencing the selection of research directions remains limited. In this study, we investigate how a project’s coherence with a scientist’s previous work and its alignment with current scientific trends affect the probability of winning a renowned research fellowship. Employing a neural network algorithm to compute these indicators, we analyze the text of 2,494 research projects proposed by scholars competing for the Sloan Research Fellowship, one of the most prestigious prizes for early-career researchers in North America. Our findings reveal notable differences across fields in the effect of coherence and alignment on the probability of winning a fellowship, reflecting distinct organizational norms. For instance, applicants with coherent and aligned projects have higher chances of winning in life sciences and chemistry. In contrast, in physics, bibliometric measures such as the number of publications and citations play a major role in the selection process. In conclusion, this study offers valuable insights into the factors influencing scientific prize selections, helping to shape the research landscape for early-career scientists across various academic fields.

Pro-environmental behavior and morality: An economic model with heterogeneous preferences (with Boris Thurm)

Abstract:
Some individuals voluntarily engage in costly pro-environmental actions although their efforts have limited direct benefits. This paper proposes a novel economic model with heterogeneous agents explaining why. Each agent has a homo moralis type of preference, which combines selfishness and morality. Morality is modeled here as the payoff an agent receives if all other agents act like her. Our model builds on extant literature showing that homo moralis preferences have an evolutionary advantage to better evaluate the behavioral motives of agents. Shedding light on how people respond not only to economic but also moral incentives, we contribute to the ongoing policy debate on the design of efficient environmental policies.

Generating Innovation in the Lab: Experimental Evidence from the Life Sciences (with Jacqueline Lane, Kevin Boudreau, Eva Guinan, and Karim Lakhani)

Abstract:
High-value innovations in science and technology often come from especially novel or atypical recombinations of ideas and prior innovations. In this paper, we use a field experiment to investigate how exposure to new ideas affects the novelty and value of recombinant innovations. Partnering with a leading U.S. medical school and intervening in their grant evaluation process, we created exogenous variation in the new ideas that 142 research scientists were exposed to, by randomly assigning each researcher to evaluate 15 (out of 150) scientific proposals. We then leverage this unique approach to test key claims of existing theories of novel recombinations while also documenting a range of added patterns. Tracking the scientists’ research output over the following five years, we find that the randomized exposures led to papers that recombined ideas that are more novel to the scientist, while simultaneously broadening their expertise. In particular, the papers were 10% and 34% more likely to be published in journals and subfields that were new to the scientist, and also involved 0.5 new coauthors on average. That said, we find no evidence that the recombinations could be judged as more novel in relation to the prior literature. Further, recombinations that were less novel (i.e., more incremental) were more successful in terms of forward citations and journal impact factor. Our study offers insights into how successful recombinant innovations can be produced from new idea exposures, while highlighting the potential trade-offs between the novelty and value of innovative outputs.

Does It Pay to Do Novel Science? The Selectivity Patterns in Science Funding (SPP, 2021) (with Michele Pezzoni and Fabiana Visentin)

Abstract:
Public funding agencies aim to fund novel breakthrough research to promote the radical scientific discoveries of tomorrow. Identifying the profiles of scientists being financed to pursue their research is therefore crucial. This paper shows that the funding process is not always awarding the most novel scientists. Exploiting rich data on all applications to a leading Swiss research funding program, we find that novel scientists have a higher probability of applying for funds than non-novel scientists, but they get on average lower ratings by grant evaluators and have fewer chances of being funded. We discuss the implications for the allocation of scientific research spending.

COVID-19: Insights from Innovation Economists (SPP, 2020) (with George Abi Younes, Omar Ballester, Gabriele Cristelli, Gaétan de Rassenfosse, Dominique Foray, Patrick Gaulé, Gabriele Pellegrino, Matthias van den Heuvel, Elizabeth Webster and Ling Zhou)

Abstract:
This article provides the take of innovation economists on the COVID-19 pandemic. It is addressed to the general public and focuses on questions related to the Science, Technology, and Innovation (STI) ecosystem. It does not present new research findings. Instead, it provides a reading of current real-world developments using economic reasoning and relying on existing economic research. The first part of the report explains the root causes for a general underinvestment in Research and Development (R&D), with a particular focus on vaccines. The second part discusses several aspects related to current STI policy reactions. The third part of the report assesses some potential long-term impacts of the COVID-19 pandemic. The last part of the report offers some concluding thoughts. The STI policy response cannot be limited to the urgent need for ‘technological fixes.’ A second line of response involves the production of new knowledge to prevent outbreaks (ex-ante) or mitigate their effects (ex-post).

The Important Thing is not to Win, it is to Take Part: What if Scientists Benefit from Participating in Research Grant Competitions? (Research Policy, 2019) (with Michele Pezzoni and Fabiana Visentin)

Abstract:
“The important thing is not to win, it is to take part,” this famous saying by Pierre de Coubertin asserts that the value athletes draw from Olympic games lies in their participation in the event and not in the gold they collect during it. We find similar evidence for scientists involved in grant competitions. Relying on unique data from a Swiss funding program, we find that scientists taking part in a research grant competition boost their number of publications and average impact factor while extending their knowledge base and their collaboration network regardless of the result of the competition. Receiving the funds increases the probability of co-authoring with co-applicants but has no additional impact on the individual productivity.

At the Origins of Learning: Absorbing Knowledge Flows from Within the Team (JEBO, 2017) (with Michele Pezzoni and Fabiana Visentin)

Abstract:
Empirical studies document a positive effect of collaboration on team productivity. However, little has been done to assess how knowledge flows among team members. Our study addresses this issue by exploring unique rich data on a Swiss funding program promoting research team collaboration. We find that being involved in an established collaboration and team size foster the probability of an individual learning from the other team members. We also find that team members with limited experience are more likely to learn from experienced peers. Moreover, there is an inverted U-shaped effect of cognitive distance on the probability of learning from other team members.