
Is this virtual jellyfish safe to touch? Image from a virtual reality task. Image credit: Mahajan et al. (CC BY 4.0)
Animals need to navigate a complex world where they gain rewards, such as finding food and shelter, while avoiding danger on a daily basis. This creates a ‘safety-efficiency dilemma’ in how to pursue rewards efficiently without increasing the risk of harm.
The human brain relies on two key systems to manage this challenge. The instrumental system learns from the consequences of our actions, weighing costs and benefits to maximize overall outcomes. In contrast, the Pavlovian fear system generates automatic defensive responses. For example, it learns to associate certain cues with danger and triggers instinctual reactions, like pulling the hand away from a hot stove. Sometimes these systems conflict—fear urges withdrawal even when careful action might lead to a greater reward—and the brain must resolve this tension. Mahajan et al. set out to explore how the brain flexibly manages the influence of this automatic fear system depending on how uncertain a situation is.
Is it computationally advantageous for the brain to adjust the impact of Pavlovian fear based on outcome uncertainty? And is there evidence that humans actually use this strategy? These questions are important because the influence of fear has often been treated as fixed. Recognizing its flexibility offers a more nuanced understanding of how we balance safety and efficiency in decision-making.
Mahajan et al. found that a flexible Pavlovian fear system, guided by uncertainty, promotes safer decisions without substantially sacrificing efficiency. The researchers first used computer simulations to show that a model in which fear’s influence decreases as the environment becomes more predictable resolves the safety–efficiency trade-off more effectively than models with a fixed fear response.
They then tested this prediction in a virtual reality experiment, where human participants decided whether to approach or withdraw from stimuli that could lead to a mild electric shock. The results showed that the flexible model best explained both people’s choices and their reaction times.
These findings may eventually benefit individuals with anxiety disorders or chronic pain, conditions often marked by excessive avoidance behaviors. The results suggest the core issue may not simply be an overactive fear system, but rather a difficulty in flexibly reducing its influence when situations become predictable. This perspective could inspire therapies that aim to help patients to better distinguish between controllable and uncontrollable threats, making their defensive responses more adaptable – a concept already present in some forms of cognitive behavioral therapy.