33: Scientific Decision-Making in Entrepreneurship: Beliefs, Experimentation, and Learning Under Uncertainty (Chiara Spina)

Prerecorded session

https://video.wixstatic.com/video/ec1421_820be27f56ed4ca598d82a0622fcd51e/1080p/mp4/file.mp4

Live session

tba

Overview

Entrepreneurs regularly make high-stakes decisions under uncertainty, relying on limited data and prior beliefs. A growing body of research proposes a scientific approach can be a helpful way to make to these decisions. Using a scientific approach means treating the business model as a theory, articulating falsifiable hypotheses, testing them through structured experiments, and updating beliefs.

Evidence from randomized controlled trials suggests that this approach helps make informed decisions which leads to positive performance outcomes. Yet a central puzzle remains: if disciplined experimentation is so valuable, why is it not more widely or consistently adopted in practice? Why do founders persist in intuition-driven or confirmatory approaches to learning, even when structured alternatives are available?

This lecture uses that tension to examine what it actually takes to make entrepreneurial decision-making more scientific. Drawing on the Carnegie School tradition it focuses on the microfoundations of how entrepreneurs form theories and hypotheses, design tests, and update beliefs in practice. Combining theoretical knowledge and lessons learnt in the field, the session will discuss how cognitive and contextual constraints shape its effectiveness, and what this implies for both theory and future empirical work.

Required readings

Murray, F., & Tripsas, M. (2004). The exploratory processes of entrepreneurial firms: The role of purposeful experimentation.

Csaszar, F. A., & Levinthal, D. A. (2016). Mental representation and the discovery of new strategies. Strategic Management Journal, 37(10), 2031-2049.

Camuffo, A., Cordova, A., Gambardella, A., & Spina, C. (2020). A scientific approach to entrepreneurial decision making: Evidence from a randomized control trial. Management Science, 66(2), 564-586.

Spina, C. (2026). What Helps Entrepreneurs Learn: Reflections, New Directions, and Open Questions. Bayesian Entrepreneurship Book Chapter, available on SSRN: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5019371

Discussion questions

i. When and why does the scientific approach fail to produce positive effects?

Existing work emphasizes the positive effect (in terms of decision-making and performance) of the scientific approach, leaving its failure modes under-researched. A Carnegie perspective can be helpful to understand potential breakdowns at each stage: ill-formed or non-falsifiable theories, poorly designed or misinterpreted tests, and biased or incomplete belief updating. What key theoretical and empirical tools does the Carnegie tradition offer to diagnose and study them?

ii. What are the boundary conditions of the scientific approach?

We know relatively little about when, where, and for whom this approach is most effective. Key contingencies could include (but not be limited to) type of uncertainty, founder experience, and institutional context. More fundamentally, the approach may presume a world that is sufficiently learnable. In settings characterized by Knightian uncertainty, long feedback cycles, or ambiguous signals, disciplined experimentation may be infeasible or even misleading. When does scientific thinking improve decisions and when does it create false precision?

iii. How does scientific decision-making operate in teams?

Most work treats the entrepreneur as an individual, yet ventures are typically team-based. This raises questions about how priors are aggregated, which hypotheses get tested, and how disagreement is resolved. Does scientific thinking diffuse within teams, or generate friction? Under what conditions do team dynamics amplify or undermine structured learning?

iv. Is AI a substitute for, or a complement to, scientific thinking?

AI tools can now quickly generate hypotheses, design experiments, and synthesize feedback at low cost. This raises a fundamental question: does AI enhance learning, or bypass the cognitive effort that makes learning effective?

If testing becomes cheap and fast, the value of disciplined hypothesis formation may shift. Is the scientific approach primarily a tool for resource allocation under constraints, or a scaffold for belief formation that remains essential even when constraints weaken?

Group members

Faculty advisors

• tbd

Participants

• tbd