(Georgia Institute of Technology, 2022-12-14)
Parmar, Sweta
Model-Based Decision Support Systems (MDSS) are prominent in many professional domains of high consequence, such as aeronautics, emergency management, military command and control, healthcare, nuclear operations, intelligence analysis, and maritime operations. An MDSS generally uses a simplified model of the task and the operator to impose structure to the decision-making situation and provide information cues to the operator that is useful for the decision-making task. Models are simplifications, can be misspecified, and have errors. Adoption and use of these errorful models can lead to the impoverished decision-making of users. I term this impoverished state of the decision-maker model blindness. A series of two experiments were conducted to investigate the consequences of model blindness on human decision-making and performance and how those consequences can be mitigated via an explainable AI (XAI) intervention. The experiments implemented a simulated route recommender system as an MDSS with a true data-generating model (unobservable world model). In Experiment 1, the true model generating the recommended routes was misspecified to different levels to impose model blindness on users. In Experiment 2, the same route-recommender system was employed with a mitigation technique to overcome the impact of model-misspecifications on decision-making. Overall, the results of both experiments provide little support for performance degradation due to model blindness imposed by misspecified systems. The XAI intervention provided valuable insights into how participants adjusted their decision-making to account for bias in the system and deviated from choosing the model-recommended alternatives. The participants' decision strategies revealed that they could understand model limitations from feedback and explanations and could adapt their strategy to account for those misspecifications. The results provide strong support for evaluating the role of decision strategies in the model blindness confluence model. These results help establish a need for carefully evaluating model blindness during the development, implementation, and usage stages of MDSS.
(Georgia Institute of Technology, 2020-03-31)
Parmar, Sweta
A tool commonly used to aid the navigational decisions of pilots to avoid weather hazards is Next Generation Radar (NEXRAD), which provides information about geographically referenced precipitation. However, this tool is limited because, when pilots use NEXRAD, they have to infer the uncertainty in the meteorological information for both understanding current hazards as well as extrapolating the impact of future conditions. Recent advancements in meteorology modeling afford the possibility of providing uncertainty information concerning hazardous weather for the current flight. Although probabilistic weather products do not exist in today’s cockpit, it is critical to evaluate how operators might use or misuse such products when incorporating uncertainty information in their decision-making. In addition, it is important to study how accurate a probabilistic decision aid needs to be for effective use by operators. Although there are systematic biases that plague professional’s use of uncertainty information, there is evidence that presenting forecast uncertainty can improve weather-related decision-making. The current study investigates a simulated probabilistic component of a decision aid that renders flight-path risk as a probability that the route will come within 20 nmi radius (FAA recommended safety distance) of hazardous weather within the next 45 minutes of flight. The study evaluates four NEXRAD displays integrated with Flight-Route Risk Estimates for Enhanced Decisions (FRREED) providing varying levels of support. The “no” support condition has no FRREED (the NEXRAD only condition). The “baseline” support condition employs a FRREED whose accuracy is consistent with current capability in meteorological modeling. The “moderate” support condition employs a FRREED whose accuracy is likely at the top of what is achievable in meteorology in the near future. The “high” support display provides a level of support that is likely unachievable in an aviation weather decision-making context without significant technological innovation. The results indicate that operators did rely on the FRREED to improve their performance over the no-support condition (NEXRAD only). The level of performance of the operators improved in terms of both calibration and resolution as the aids increased in accuracy. I will discuss the implications of the findings for the safe introduction of probabilistic decision aids in future general aviation cockpits.