Recent, Ongoing, and Future Research Projects
My research focuses on non-classical levers to induce cooperation and innovation in a variety of settings. I have studied how innovative technologies are developed, how they can resolve existing cooperation problems, and how they may create new, emerging ways of working - such as crowdsourcing, anonymity, and risk-taking. In tackling this problem, my work is methodologically diverse, applying empirical methods to understand these problems in practice and analytical models to draw normative conclusions.
The future trajectory of my work is increasingly focused on innovation and the development of new machine learning and artificial intelligence models. Together with my co-authors, we are applying new tools to understand and improve the ‘supply chain’ that supports these models. This page describes my recent and ongoing research projects in this domain, with links to published or working papers as appropriate.
Citations: 104
H-Index: 4
Working Papers
Hold Me Accountable: Anonymity and Prosocial Behavior
Published at M&SOM (May 2025)
With Claire Senot (Tulane University). This paper focuses on the lever of accountability and how it can used to motivate prosocial behavior in a hierarchical setting. This paper uses a quasi-experimental method to investigate how service providers can use the names of their consumers to create an implicit incentive to engage in prosocial behavior by relying on social context. We study the case where a service provider removed the names of participants from test kits in a viral testing program. The effect was a surprising 21% reduction in participation, negative moderated by the size of the groups participants were randomly assigned to. We confirm the robustness of this result using difference-in-differences, regression discontinuity, and augmented local linear models. We examine further moderating effects and offer guidance to service providers can use accountability as an implicit lever to encourage prosocial behavior, both in volunteering settings and businesses.
Presented at: POMS 2023, INFORMS 2023, University of Toronto, Wharton Empirical Workshop 2024, POMS 2024.
Media Attention:
Cultures for Innovation
Under review at Management Science (August 2025)
With Jeremy Hutchison-Krupat (University of Cambridge). We examine how an organization can align its strategic objectives with its innovation culture. Where innovation culture varies according to the targets that are set, the autonomy that exists, and the tolerance management has for failed experiments. Recognizing that organizational culture varies, we build a model of a relational contract where a principal delegates the execution of a derivative and a novel innovation to an expert agent. Given a set of cultural traits, we determine the level of novelty that a principal could pursue. We then analyze how the principal may use their ability to influence culture in pursuit of the novel innovation that forms their strategic objective. We confirm that tolerance for failure can be beneficial to enable novel initiatives, but our results also provide an alternative, cautionary view of tolerance for failure: pursuit of the most novel innovations may require tolerating less failure. We verify that setting targets to constrain the amount of derivative innovation in the portfolio may increase the rate of novel innovation and reduce the likelihood that projects are terminated. We further extend our analysis to examine benefits of target setting and the principal's decision over cultural traits. For managers, we make specific recommendations to map organizational culture to successful innovation.
Presented at: INFORMS 2024, POMS 2025, University College Dublin, McGill University, INFORMS 2025
An Invisible Workforce: Coordinating Voluntary Effort in the AI Supply Chain
Submitted to Management Science (October 2025)
With Setareh Farahjollahzadeh (McGill University, Desautels Faculty of Management). The expansion of AI has increased demand for crowdsourced data labeling; however, constraints on scale and quality have become a bottleneck in this `AI supply chain' -- particularly when expert knowledge is required and ground truth is uncertain. We examine how a scalable operational lever—the minimum target score required for task submission—affects both quality and participation in a large-scale randomized field experiment, Borderlands Science, hosted within the video game Borderlands 3. Drawing on the psychological theory of learned industriousness, which posits that effort becomes intrinsically rewarding when repeatedly paired with positive outcomes, we test whether higher performance targets can condition greater discretionary effort and thereby increase participants’ intrinsic motivation, leading to higher task quality without additional incentives. We study 400,000 participants who completed 56.98 million tasks (≈1.17 million hours of work) under randomly assigned target scores. Holding extrinsic rewards constant while varying targets, we find that higher recent target exposure significantly increases subsequent discretionary quality; for example, a one–standard-deviation increase in the prior 10-task target average yields 9% higher quality. Our analysis confirms the causal relationship between higher quality on the current task and prior randomly assigned targets that condition effort. However, the effect of past target scores on quality exhibits diminishing returns and is reversible: exposure to lower targets accelerates the reduction of submission quality, consistent with behavioral conditioning rather than skill learning. Moreover, higher past target exposure not only increases the quality of completed tasks, but also leads participants to exert greater effort to explore more varied solutions.
Presented at: University College Dublin, University of Toronto
Minority Figures: Gender Inequality in Risk Taking
Under review at Manufacturing & Service Operations Management (October 2025)
With Claire Senot (Tulane University). Operational performance often depends on individuals disclosing private information that carries interpersonal risk -- such as errors, near-misses, or disease symptoms. Thus, individuals frequently withhold this information, with negative consequences for performance. We examine how group composition -- specifically gender diversity -- affects individuals' willingness to take these risks. Building on social identity theory, we argue that group diversity affects individuals' risk-taking behavior through tokenism and social recategorization. Low diversity leads to status concerns for the minority, while majority members face weaker pressure to disclose unless norms shift. We test this theory in a natural experiment: 5,000 students were randomly assigned to groups during a voluntary viral-screening initiative, where participation had both private costs and stigma risk, yet was required for operational planning. Under-represented gender groups were more likely to participate, and their participation did not induce a spillover effect to the gender majority at low levels of diversity. However, compared to single-gender groups, when the minority share exceeded 40% of the group composition, average group information disclosure increases, driven by an increase in the participation of the majority members. Our results show that gender composition affects the assignment of the interpersonal risks associated with disclosure. We clarify when diversity does (and does not) support operational goals, while also highlighting that the burden may be unevenly distributed in mixed gender groups. We inform the design of team structures and incentive schemes that support equitable, operationally-relevant information sharing.
View this paper in SSRN. (pre-revision version)
Presented at: POMS 2025
No Spoilers: Grocery supply chain traceability and food waste
Under Revision (October 2025), target Management Science
With Javad Nasiry (McGill University) & Elaheh Roshedinejad (University of Toronto). This paper examines the challenges posed to equity by food waste, where millions of tonnes of potential edible food is wasted while people go hungry. Reducing food waste is not typically in the best interests of retail businesses, who benefit from maximizing their coverage of demand and reducing sourcing costs, which can lead to unnecessary waste. In this paper, we build a model of a three-tier supply chain with perishable products and consumer demand that is sensitive to price and freshness. The retailer may make costly investments in monitoring the process conformance of suppliers with a traceability system, which provides information on the batch-level freshness and perishability. Traceability adoption improves ordering accuracy at the retailer, but may reduce the supplier’s incentive to invest in improving processes. The interplay of these effects means that traceability technology does not necessarily reduce food waste. We solve the societal planners problem and find that subsidies for technology investment are more effective in reducing food waste that penalties on the volume of waste produced at retailers. The balance between these forces depends critically on the consumers’ sensitivity to freshness. Our paper supports the wider adoption of traceability technology, but demonstrates that low cost technology is often more important for reducing food waste than highly accurate technology — even a little traceability information goes a long way to reducing food waste.
View this paper on SSRN. (pre-revision version)
Presented at: INFORMS 2024, POMS 2025, McGill University.
Media Attention:
Works in Progress
Adaptive Retention in Crowdsourced AI Data Labelling
With Setareh Farajollahzadeh (McGill University). Building on “An Invisible Workforce”, our large-scale field experiment revealed several important behavioral dynamics within open crowdsourced data annotation. This work builds on these behavioral dynamics with a normative model to develop an algorithm to recommend tasks to contributors based on their trajectory through tasks and their behavioral traits.
Managing Disruptions to Innovation Projects
With Jeremy Hutchison-Krupat (University of Cambridge). Disruptions — such as the actions of competitors, failed experiments, and global crises — threaten the success of innovation projects. In this work, we develop a model to determine how the managers of innovation projects can establish new directions and ways of workings after disruptions.