Why focus changes when it matters most

Sustaining attention over long periods affects everything from routine tasks to high-stakes decisions. Whether it is an umpire making a split-second call or a driver glancing at a navigation system, even small lapses in focus can have meaningful consequences. New research from Texas A&M University is shedding light on how people allocate their attention and why it fluctuates in ways that can lead to both errors and improved performance.

Lekhapriya Dheeraj Kashyap, a Ph.D. student in the Wm Michael Barnes ’64 Department of Industrial and Systems Engineering, is leading research that explores how individuals make strategic decisions about where and when to focus their attention. Her work challenges long-standing assumptions about how attention operates.

“At its core, we were trying to understand how and why our attention fluctuates,” Kashyap said. “When someone is doing a repetitive but important task, what drives them to pay more or less attention at any given moment?”

Traditional theories often frame attention as a limited resource that gradually depletes over time. Kashyap’s research offers a different perspective. Instead of viewing attention as something that simply runs out, her work suggests that people are constantly making decisions — often subconsciously — about how much effort to invest based on their perception of a situation.

“We wanted to build frameworks that treat attentional effort not as a fixed resource being drained, but as a strategic choice being made in response to an inferred situation,” she said.

Kashyap and her collaborators developed mathematical models that capture how individuals form internal beliefs about their environment and how those beliefs shape their actions. Rather than relying on controlled lab experiments alone, the research draws from rich, real-world data where attention plays a critical role.

The team explored this question in two very different settings: professional baseball and simulated driving. Despite their differences, both environments require individuals to sustain attention over time while responding to constantly changing conditions.

In the first study, researchers analyzed thousands of ball and strike calls made by Major League Baseball umpires. Using high-resolution tracking data, they compared each decision to an objective ground truth, allowing them to measure accuracy with remarkable precision.

The results showed that attention is not applied evenly across situations. Instead, it shifts depending on how important the moment feels to the individual, sometimes in ways that introduce bias.

“Our model showed that umpires tend to allocate different levels of attention when facing high-status pitchers,” Kashyap said. “When they are in a high-attention state, those biases almost disappear and accuracy is driven primarily by the physical location of the ball.”

This finding suggests that errors are not simply the result of fatigue or lack of skill. Instead, they can emerge from how a person interprets the context of a situation and adjusts their effort accordingly.

In a second study, Kashyap turned to the driver’s seat, examining how people divide their attention between the road and in-vehicle systems. Using a high-fidelity driving simulator equipped with eye-tracking technology, the team was able to measure exactly when and how long participants looked away from the road.

The study revealed striking differences in behavior. 

“We found that some drivers are very conservative and barely look away from the road,” Kashyap said. “Others place a higher value on completing secondary tasks and will engage with distractions even under high distraction.”

These patterns were not random. Instead, they reflected stable and measurable differences in how individuals evaluate trade-offs between effort, reward and risk.

“What I find most fascinating is how rational attentional effort is, even when it leads to outcomes we would call errors,” Kashyap said. “People are constantly making subconscious decisions about where to invest their mental energy based on what they think matters most in the moment.”

A key challenge in the research was capturing something that cannot be directly observed. Attention and context are internal states, meaning they must be inferred from behavior rather than measured outright. To address this, Kashyap used structural modeling techniques that reconstruct the decision-making process from observable data.

“Rather than asking what correlates with accuracy, we asked what process could generate the decisions we observe,” she said. “That allows us to build models that are not just descriptive, but actually predictive.”

The implications of this work extend far beyond sports and driving. Lapses in sustained attention contribute to errors in fields such as healthcare, aviation and security, where even small mistakes can carry significant consequences.

Most existing solutions, such as scheduled breaks or general training programs, take a broad approach. Kashyap’s research opens the door to something more precise. By identifying how attention is allocated across different contexts, her models can help pinpoint exactly where and when interventions are needed.

“Because we can infer how a specific individual is allocating attention, we can design interventions that address those gaps precisely,” she said.

In practical terms, this could lead to systems that adapt in real time. A driver assistance system, for example, could recognize when a person is likely to become distracted and intervene before a dangerous situation develops. In professional settings, targeted incentives or feedback mechanisms could reduce bias and improve decision making.

The research also highlights a broader shift in how attention is understood. Rather than being a passive process that fades over time, attention is an active and adaptive system shaped by perception and experience.

“We are not just passive observers whose focus fades over time,” Kashyap said. “We are actively interpreting our environment and adjusting our effort based on what we think the situation demands.”

As this work continues, it has the potential to reshape how researchers and practitioners approach human performance. By better understanding how attention is allocated and why it fluctuates, new tools and technologies can be developed to help people stay focused, make better decisions and perform more effectively in the moments that matter most.

“Intelligent engineered systems are rapidly entering domains that require sustained interaction with people,” said Dr. Alfredo Garcia, professor and Kashyap’s advisor. “Our work focusing on latent attentional states is a first step towards integrating latent-human state inference with ensuring safe and effective human-machine interaction.” 

This research was supported in part by the National Science Foundation and the U.S. Army Research Office.