Deceptive Decision-Making under Uncertainty
Authors: Yagiz Savas, Christos K. Verginis, Ufuk Topcu5332-5340
AAAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We evaluate the performance of the proposed approach via comparative user studies and present a case study on the streets of Manhattan, New York, using real travel time distributions. ... We present three experiments. Firstly, we illustrate the effects of different parameters in the proposed approach on the agent s deceptive behavior. Secondly, we present online user studies and compare the proposed approach to two recently proposed deception methods (Masters and Sardina 2017; Dragan, Holladay, and Srinivasa 2015) as well as a baseline. Finally, we present a large-scale case study on the streets of Manhattan, New York with real travel time distributions and illustrate the use of deception in realistic scenarios under probabilistic constraints on travel time. |
| Researcher Affiliation | Academia | Yagiz Savas, Christos K. Verginis, Ufuk Topcu The University of Texas at Austin, Austin, TX yagiz.savas@utexas.edu, christos.verginis@austin.utexas.edu, utopcu@utexas.edu |
| Pseudocode | No | The paper provides mathematical formulations and algorithms described in text, but no explicitly labeled |
| Open Source Code | No | The paper states |
| Open Datasets | Yes | We utilize the real-world speed data provided in the open source database (Uber Technologies 2021) to express realistic travel times. |
| Dataset Splits | No | The paper describes user studies and experiments, but does not provide specific train/validation/test dataset splits (e.g., percentages or sample counts) for model training or evaluation. It describes experimental conditions and subject recruitment but not data partitioning for machine learning models. |
| Hardware Specification | Yes | We run all computations on a 3.2 GHz desktop with 8 GB RAM and employ the Gurobi solver (Gurobi Optimization 2021) for optimization. |
| Software Dependencies | Yes | We run all computations on a 3.2 GHz desktop with 8 GB RAM and employ the Gurobi solver (Gurobi Optimization 2021) for optimization. |
| Experiment Setup | Yes | We generate a continuous travel time distribution on each edge by assuming that the speed follows a lognormal distribution, which is a common assumption in transportation networks (Rakha, El-Shawarby, and Arafeh 2010). ... We choose the value 0.8 to clearly illustrate the effect of probabilistic time constraints on the deceptive behavior. Finally, we use the parameters c(s, a)=5 for all s S and a A, γo=0.95, α=1, and γa=1. |