TL;DR: We developed the Decisions Over Time (DOT) task to measure active inference, the cognitive skill of updating beliefs in uncertain situations. The task unexpectedly revealed two successful strategies for solving complex problems: Adapters repeatedly sought the correct underlying rules, while Satisficers found and used efficient, "good-enough" shortcuts.

Learning to orient in unfamiliar situations

Leaders face volatility, uncertainty, complexity, and ambiguity (VUCA). In these conditions, past experience can lead them to hold on to outdated models of the world.

Military doctrine has used Colonel John Boyd's OODA loop (Observe, Orient, Decide, Act) and similar models for decades to describe an ideal decision process. During the Orient phase, people make sense of a situation and form a mental model. It is the most important phase, but also the most misunderstood and hardest to train.

When a situation is unfamiliar, a checklist cannot tell someone how to orient. Training has to focus on the learning process itself.

Active inference and prediction error

We propose active inference as this missing competency. The theory comes from cognitive neuroscience and describes a core mechanism of the brain: it actively predicts what will happen, with the primary goal of minimizing "surprise," or prediction error. Prediction error is the mismatch between what someone expects and what happens.

The cycle runs constantly: people predict what happens next, act, observe the outcome, and update their model when the outcome surprises them. That update is the signal to learn. This cycle describes how people make sense of new situations that change quickly.

The Decisions Over Time task

To measure active inference, we developed the Decisions Over Time (DOT) task, a computerized test shown above.

Each trial presents three colored dots (red, green, and blue) in a row. The only instruction is to "click the correct dot". A hidden rule determines the correct choice based on the color of the middle dot. For example, under Rule 1, a blue middle dot might mean that the left dot is correct, red might mean the right dot, and green might mean the middle dot.

Players have limited feedback to work from: a correct click earns one point, and an incorrect click loses one. They must use that feedback to infer the hidden rule.

Once a player masters the rule and starts scoring perfectly, the task changes it without warning. The resulting errors are unexpected: the player's mental model no longer works, and they must inhibit the old belief to infer the new rule.

Two successful strategies

We found four distinct groups of performers. Two of these groups succeeded with very different strategies.

Adapters

Adapters used feedback to learn Rule 1, as we had expected. When the rule changed, they detected the errors, discarded their old model, and learned Rule 2, then Rule 3, and so on. They used logic and persistence to find the single correct solution and eliminate errors. This approach demanded considerable effort, focus, and executive functioning.

Satisficers

Satisficers surprised us by succeeding without ever learning the complex hidden rule. They quickly noticed a statistical shortcut: because of the task's design, one color was correct 66.6% of the time, regardless of its position. They clicked that color on every trial.

They willingly accepted a 33% error rate in exchange for a simple strategy that required little effort, was fast to execute, and still produced a high score. This probabilistic learning approach gave them a solution that was correct two-thirds of the time and could be executed 100% of the time.

Choosing a strategy for the situation

The distinction between Adapters and Satisficers also appears in real decisions. General George S. Patton's observation captures it: "A good plan violently executed Now, is better than a perfect plan next week."

Adapters pursue the "perfect plan" through detailed analysis. Their approach is essential for complex, high-stakes problems where a "good-enough" answer would be catastrophic and correctness is the only priority. Satisficers pursue the "good plan now." Their fast, efficient, probabilistic approach is needed when conditions change quickly and speed and resource management matter more than perfection.

Leaders often default to one style, although real situations call for both. Effective leadership requires knowing when to use each approach.

The DOT task could help identify a leader's natural tendency. With that knowledge, leaders can learn to switch strategies deliberately to suit the problem. The task gives us a concrete, measurable way to develop the Orient competency that doctrine has described for decades without operationalizing it.

Full Citation

Nye, J. M., Stothart, C., Graves, R., Francisco, A., & Peterson, J. (2024). Active inference: A competency for making decisions in uncertain situations (ARI-SS-SR 2024-09). U.S. Army Research Institute for the Behavioral and Social Sciences.