Information Sharing in Design Teams under Competition: Effect on Decision Quality and Search Efficiency
Javadpour, N., Mutlu, K., Sha, Z., & Bayrak, A. E. ASME Journal of Mechanical Design. In press.
Loading recorded sessions...
Follow the search
Players could not see this curve or its peak.
Each test costs 10 tokens. Your team wins by finding a better-performing design than the opposing team.
What do the axes mean?
Imagine Sony and Bose testing noise-cancellation strength, or Airbus and Boeing testing wing shapes. Here, x is one design choice and y is its measured performance. Teams can share discoveries internally while seeing only part of their rivals’ progress. These are analogies for the lab experiment.
Selected player's testsNot started
Selected player's tokens left200
Selected player's best performanceNot sampled
Time since first own sampleNot started
Selected player
Teammate
Competitor best
Model comparison
Own best
Press Play or Next decision to follow the search.
Round results unlock when the visible search ends.
Can the model anticipate this player?
Reveal each prediction alongside the action it was made for.
Predicted choicePending
Stopping probabilityPending
Predicted actionPending
Position errorNo comparison yet
How the prediction works
An LSTM learns dependencies across the sequence of tests. We use 5-fold cross-validation and each player-round's held-out model. A per-scenario cutoff maximizes F1, balancing precision and recall for stopping. Predictions use the history available before the displayed action, starting after 3 own tests. Stopping includes submission and timeout.
Further reading
What changes when teams compete, share discoveries, or stop searching? Explore our research on human decisions and the rational benchmarks used to understand them.
Published
A Quantitative Analysis of Rational Decisions Under Uncertainty in Engineering Systems Design
Mutlu, K., Javadpour, N., Sha, Z., & Bayrak, A. E. (2025). Proceedings of the Design Society, 5, 249–258. DOI: 10.1017/pds.2025.10039.
Human search compared with a Bayesian expected-utility benchmark.
Impact of Information Sharing Between Members in Design Teams Under Competition in Unknown Design Space Exploration
Javadpour, N., Mutlu, K., Sha, Z., & Bayrak, A. E. (2025). Proceedings of the ASME 2025 IDETC-CIE, V03BT03A029. DOI: 10.1115/DETC2025-169672.
How sharing information changes search, performance and resource use.
Manuscripts
Predicting Designers’ Search Behavior and Stopping Decisions Under Competition with Sequence Learning
Mutlu, K., Javadpour, N., Sha, Z., & Bayrak, A. E. Manuscript in development.
Multi-task LSTM modeling of sequential next-action, outcome and stopping behavior.
Quantifying Bounded Rationality and Robustness in Design Decisions under Competition
Mutlu, K., Javadpour, N., Sha, Z., & Bayrak, A. E. Design Science. Revision under review.
Gaussian-process and Bayesian-optimization modeling of sequential decisions under uncertainty.