August 21-22, 2026
Carnegie Mellon University
Carnegie Mellon University
The YinzOR 2026 conference program features an exciting mix of keynote presentations from renowned Operations Research & Management Science scholars, competitive student flash talks, and an interactive poster session.
| 12:00 PM - 1:00 PM Tepper 3801 |
Registration |
| Come pick up your name badge and conference materials before the opening remarks. |
| 1:00 PM - 1:20 PM Tepper 3801 |
Opening Remarks |
| Welcome address from the YinzOR 2026 Co-Chairs. |
| 1:20 PM - 2:00 PM Tepper 3801 |
Column Elimination for Scheduling Problems |
| Vianney Coppé, Ph.D., Optimization Scientist, Hexaly | |
| Column elimination is a technique that has recently been introduced for solving optimization problems that aim to find an optimal set of sequences, such as vehicle routing problems. It relies on a compact representation of the problem in the form of a relaxed decision diagram, from which a relaxed solution can be obtained by solving a constrained network flow problem. As long as the solution contains sequences that are relaxed with respect to either the problem constraints or the cost function, the decision diagram is refined to eliminate the corresponding conflicts. In this talk, we describe the key ingredients required to apply column elimination to scheduling problems and present preliminary computational results. |
| 2:00 PM - 2:40 PM Tepper 3801 |
Congested Waiting Lists and Organ Allocation |
| Pengyu Qian, Assistant Professor, Boston University | |
| More than 25% of the kidneys that are recovered from deceased donors in the U.S. and are offered to patients on the national waiting list are not utilized. This paper shows that waiting list designs can suffer from a form of congestion that can lead to discarding valuable organs. Specifically, as organs have a limited "cold ischemia time" in which they can be offered before they expire, an organ may expire before being offered to a patient lower on the waiting list who would accept it. We develop a parsimonious framework to study equilibria of waiting lists with congestion, and show that congestion provides a strong enough externality to substantively affect welfare and wastage. The delegation of organ acceptance decisions to risk-averse doctors can worsen congestion and increase waiting times. We discuss how recent changes to waiting list designs affect congestion, and policies and market designs that can mitigate congestion. |
| 2:40 PM - 3:05 PM Tepper 3810 / 3807 |
Coffee Break & Networking |
| Network with peers and enjoy refreshments. |
| 3:05 PM - 3:45 PM Tepper 3801 |
Capacity Constrained ML Systems |
| Hannah Li, Assistant Professor, Columbia Business School | |
| AI tools increasingly guide targeted interventions in healthcare, education, and recruiting. Algorithms score individuals, trigger outreach to those above a threshold (e.g., high-risk or high-value), and encourage them to request service; then providers deliver service to those who request. Standard practice sets the threshold and selects the algorithm to maximize predictive accuracy, assuming that better predictions yield better outcomes. We show that this approach is suboptimal when limited service capacity and probabilistic behavioral responses influence who receives service. In such settings, the optimal score threshold must balance two effects: ensuring all capacity is filled (utilization) and ensuring high-value individuals are served despite competition between requests (cannibalization). We characterize the optimal threshold and prove that policies based solely on predictive accuracy are generally suboptimal. Further, because optimal thresholds vary with service capacity, algorithm selection metrics like AUC, which weight all thresholds equally, are misaligned with operational performance. We introduce a new metric—Operational AUC (OpAUC)—and show it leads to optimal algorithm selection. Finally, we conduct a case study on sepsis early warning data and illustrate the magnitude of improvement that can be achieved from improved threshold and algorithm selection. |
| 3:45 PM - 4:25 PM Tepper 3801 |
Voice AI in Firms: A Field Experiment on Automated Job Interviews |
| Brian Jabarian, Assistant Professor, Carnegie Mellon University | |
| This paper studies whether AI automation can improve organizational outcomes by reducing variance when collecting information. We conducted a large-scale natural field experiment in which 70,000 job applicants were randomly assigned to be interviewed by human recruiters or AI voice agents. In both conditions, human recruiters evaluate the interviews and make hiring decisions. Applicants interviewed by AI agents are 12% more likely to receive job offers, and these gains translate into higher job starts and worker retention, with no decline in the productivity of hired workers. Analyzing interview transcripts reveals that AI voice agents achieve controlled variance: their interviews are more structured and consistent while remaining responsive to individual applicants, which is associated with more hiring-relevant information collected. These results demonstrate that automating information collection with AI can enhance decision quality through standardization. Paper: SSRN |
| 4:25 PM - 4:50 PM Tepper 3810 / 3807 |
Coffee Break & Networking |
| Network with peers and enjoy refreshments. |
| 4:50 PM - 6:30 PM Tepper 3801 |
10-Minute Flash Talk Competition |
| PhD students present their research in a fast-paced format. Judged by a panel of faculty. Cash prizes will be awarded! |
| 6:30 PM - 9:00 PM Tepper PNC Room |
Conference Dinner & Reception |
| Catered dinner for all registered attendees. Excellent opportunity for informal discussion. |
| 9:30 AM - 10:30 AM Tepper PNC Room |
Breakfast & Welcome Coffee |
| Fuel up for Day 2 of the conference. |
| 10:30 AM - 11:15 AM Tepper 3801 |
Precise or Broad? The Reward-Learning Frontier in Algorithmic Advice Design |
| Park Sinchaisri, Assistant Professor, University of California, Berkeley | |
| Algorithmic tools increasingly guide operational decisions by telling users what action to take. Although precise recommendations can improve immediate execution, organizations may also care about how well users perform when guidance is unavailable or conditions change. We study how advice precision, whether the same policy is communicated as an executable action or as a qualitative rule the user must apply, affects both immediate performance and subsequent unsupported decision-making. We develop a rational-inattention model in which precise advice is easier to implement, whereas broader advice requires additional effort but may preserve more transferable decision strategies. We test these predictions in two preregistered online experiments involving sequential electric-vehicle charging decisions under uncertain traffic. All recommendations are generated by the same dynamic-programming policy; treatments vary only how the advice is presented. Precise advice produces the highest performance while guidance is visible. After advice is withdrawn, qualitative advice can improve performance relative to precise advice when users encounter related environments, but this advantage disappears in sufficiently unfamiliar settings. The results reveal a performance-learning frontier in algorithmic advice design: precise advice is preferable when immediate execution dominates, whereas broader advice may be valuable when users must later act independently in related situations. |
| 11:15 AM - 12:00 PM Tepper 3801 |
Splitting the Crowd? Platform Bifurcation and Community Outcomes |
| Xiaomeng Chen, Assistant Professor, University of Pittsburgh | |
| This paper studies platform bifurcation, the process in which a subgroup of users from an original platform launches an independent spin-off platform. We analyze a major Q&A platform and 50 of its bifurcated spin-off platforms to identify the effects of bifurcation on user engagement and knowledge exchange. Identification is based on a difference-in-differences approach that exploits the introduction of the spin-off platforms in an online platform incubator. We find that bifurcation leads to a strong overall increase in contributions. While contributions to the home platform decline, the two bifurcated platforms generate more combined user contributions and attract more new users than a single united platform. We further identify interconnectivity and topic expansion as key drivers of bifurcation outcomes at the platform level. High interconnectivity implies lower engagement with the new, specialized platform, and a greater share of contributions remains on the home platform. Moreover, our evidence suggests smaller improvements in knowledge exchange when interconnectivity is high while increased engagement coincides with new topic expansion on the specialized spin-off platform. This paper is the first to empirically analyze the strategic implications of platform separation at scale and to document the moderating role of interconnectivity empirically. |
| 12:00 PM - 1:45 PM Tepper 3808 available |
Lunch Break |
| Lunch on your own. Explore the many local dining options. Tepper 3808 is available for eating. |
| 1:45 PM - 3:30 PM Tepper PNC Room |
Interactive Poster Session & Popular Vote |
| Attendees browse the student research posters and vote for their favorite. Judges evaluate for the top prizes. ($400, $300, $200, $100) |
| 3:45 PM - 4:30 PM Tepper 3801 |
k-Plan Flexibility for Resource Allocation in Facility Logistics |
| Reem Khir, Assistant Professor, Purdue University | |
| Designing resource allocation policies that perform well across a wide range of possible demand conditions remains a central challenge in modern logistics and production systems. Fully dynamic approaches can, in principle, tailor decisions to each realized scenario but are often too complex to implement at scale. Static allocations, by contrast, are easy to deploy yet may perform poorly when demand deviates from expectations. We study k-plan flexibility as an intermediate approach: the system pre-computes a portfolio of k allocation plans before operations begin and commits to one once demand is observed, creating operational optionality without full dynamic complexity. We characterize when an individual plan is effective and when a portfolio creates value beyond any single alternative. These insights lead to a constructive framework for generating a k-plan portfolio whose members are both individually high-performing and mutually complementary. Computational experiments in parcel sorting and warehouse zone picking show that even small portfolios yield substantial improvements over static allocations and recover most of the benefits of fully dynamic policies, with the magnitude of these gains varying significantly with system structure and load conditions. |
|
4:30 PM - 5:15 PM Tepper 3801 |
Explaining GPS to Galileo On building toward something we don't quite yet see |
| Matt J. Milligan , Senior AI Consultant, Highmark Health | |
|
For most of recorded history, navigation was a specialty. You hired someone who knew the way and you paid them well. Getting lost wasn't about missing an exit; it meant you may not arrive at all. Today, virtually anybody can plot a destination to within a few feet, and hardly anyone finds it remarkable. Explain GPS to Galileo, and he'd conclude (reasonably) that you were unwell. Give him a few minutes and he'd wrap his head around the satellites; he'd already figured that part out using the moons of Jupiter. Now try explaining the atomic clock onboard. You'd blow his mind. Matt Milligan, Senior AI Consultant at Highmark Health, invites you to step into Galileo's shoes for half an hour and reconcile what the age of AI means for highly skilled professionals building with brand new tools toward a destination they can't quite yet see. Has there ever been a better time to be a curious person who likes to solve problems? Building has never been cheaper. Implementation has never been faster. Yet knowing where to point any of it has never been harder. That's where you come in... and you're driving. |
| 5:15 PM - 5:30 PM Tepper 3801 |
Closing Remarks |
| Presentation of Poster and Flash Talk competition winners and closing remarks. |
| 5:30 PM - 7:30 PM Tepper PNC Room |
Happy Hour & Social Event |
| Celebrate the completion of another successful YinzOR conference with drinks and snacks. |