In changing, dynamic environments, people must often make multiple, interdependent decisions in real time, while tracking external changes and the results of their own past decisions. At the [DDMLab][1], we use cognitive computational models and laboratory experiments to help explain and predict how people make such decisions. We also develop recommendations for how people can make such decisions better. **We study how humans make choices, learn and use their experiences to make decisions in dynamic environments.** - How does experience influence our decisions? - What kinds of experiences would produce better decisions and better adaptation? - How does experience transfer to new situations? Our research approach involves laboratory experiments and cognitive computational models. We study human decision making processes by observing and collecting human choices in dynamic tasks; we also develop cognitive-computational models that reproduce such human decision making process. Models are used to explain and predict new unobserved human behavior. The figure below represents our technical approach. Data are collected from two sources: a human interacting with a task, and a computational model interacting with the same task. These are compared at many different levels (e.g., over-time learning and dynamic effects, overall averages of optimal behavior, overall risky behavior, variance in behavior, etc.). From this comparison of human and model choices, we derive conclusions regarding the human decision making process and the accuracy of our computational representations. ![enter image description here][2] ---------------------------------- **We also study humans making decisions in a wide range of decision contexts that we bring to the laboratory in the form of dynamic simulations (MicroWorlds or DMGames)** - How do operators of complex industrial plants make dynamic allocation of limited resources? - How luggage screeners at the airport can be more successful at detecting possible threatening targets? - How cyber-security analysts may improve their detection of cyber-attacks? **Our driving theory is the Instance-Based Learning Theory (IBLT), which in essence proposes that people make choices by retrieving the best outcomes from their past experience. The process involves:** - Retrieve memories (instances) that resemble the current situation (instances are triplets: situation-decision-utility) - Filter memories according to their maximum experienced expected value (utility or blended value) - Evaluate and store new instances reflecting each possible option in the decision situation - Select the option with the maximum blended value ![enter image description here][3] ---------------------------------- [1]: http://ddmlab.com [2]: https://www.cmu.edu/dietrich/sds/ddmlab/imgs/research/methods2024.jpg [3]: https://www.cmu.edu/dietrich/sds/ddmlab/imgs/research/ibl2024.jpg