As a state of mind, Option Awareness relies on the ability to project a wide range of plausible future outcomes for potential options, which we call a “decision space”. Projecting a decision space is quite difficult for people to do unaided, especially as more options and criteria are involved. People can usually envision one option at a time and evaluate its potential outcomes, good and bad, before moving on through additional options until they arrive at one that is satisfactory. This describes a person with low OA.
MITRE’s Option Awareness research uses a three-step process to help people improve their OA and make better decisions. The first step, defining forecasting models, is done before any decisions need to be made, as it sets the foundations for the second two steps, which use exploratory methods to generate decision spaces around the current problem and apply interactive visualization techniques to help users develop a deep understanding of their options and potential outcomes.
1. Find or generate forecast models
The first stage of supporting OA is to find or generate models of the problem and its context that can be used to forecast plausible futures over any criteria that help discriminate better from worse options. Everyone has some ability to estimate what might happen because of different decisions — these are forecasting models in our heads. For example, if it looks like it might rain, you might think to yourself, “If I decide not to take my umbrella, I’ll probably get wet!”
However, even an expert’s mental model of a problem is not enough to sift through large numbers of options, with countless interdependencies and tradeoffs, over a wide range of uncertain futures and unknown forces. Exploratory modeling helps enhance this ability.
The exploratory modeling used for supporting OA is model-agnostic — we’ve worked with causal models, agent-based models, systems dynamics models, and more. Multiple models can be combined. In fact, the more models, the better! Not only do more models provide more information and perspectives on the problem, but discrepancies between models can reveal assumptions, limitations, and contradictions that might have otherwise gone unnoticed.
|Focus:||Acquiring knowledge of how the system in question works|
|Approaches:||Knowledge elicitation, cognitive mapping, task/work analysis, exploratory modeling|
2. Use exploratory modeling to generate decision spaces
The second stage of supporting OA is to use exploratory modeling to generate decision spaces. Exploratory modeling refers to running forecasting models repeatedly in computational experiments to evaluate complex problems over variable and uncertain conditions. The outputs of exploratory modeling produces decision spaces, which include estimated outcomes and the possible future conditions under which they occurred. In addition to model-generated data, decision spaces can also be constructed from historical data about options and outcomes, such as patient outcome data for different medical treatments.
|Focus:||Automated model-driven experiments to generate decision spaces|
|Approaches:||Modeling & simulation, AI, and data mining|
3. Interactively visualize the decision space
The third stage of supporting OA is to interactively visualize the decision space. MITRE’s Option Awareness research has developed many different decision space visualizations (DSVs) for different applications. The most effective way to display the distributions of potential outcomes for different options over different criteria depends on the unique types of decisions being made, the individuals making those decisions, and the social and technological environment in which they make those decisions.
|Focus:||Real-world application within interactive visualization environments|
|Approaches:||Data visualization, decision science, collaboration science, user-centered design, human machine teaming|
4a. Consider New or Modified Options & 4b. Consider New or Modified Models
Given new insights into options and outcomes, the decision maker is empowered to consider new or modified options and/or models. For example, given knowledge of the specific conditions under which an option is likely to fail, they can modify the option to avoid those conditions. Or, if exploring the decision space reveals new insights about how the options or the environment work, they may need to modify the model to include new features or criteria, or change relationships found to be untrue.
|Focus:||Creative enhancement of models and options|
|Approaches:||Option modification, model refinement|
© The MITRE Corporation. All Rights Reserved. Approved for Public Release; distribution unlimited; case #20-03307-1, #17-0960, #16-0265, #15-0039.
Sign up below to receive updates about our work!